1 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6 and -mstring-compare-inline-limit.
8 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
12 for stack clash protection in a register whenever we need it to be in
15 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
18 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
20 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
23 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
25 (mlong-double-): Ditto.
26 * config/rs6000/sysv4.opt (msdata=): Ditto.
29 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
31 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
33 "vector int vec_ldl (int, long int *)", and
34 "vector unsigned int vec_ldl (int, unsigned long int *)".
35 Add comments and entries for
36 "vector bool char vec_ldl (int, bool char *)",
37 "vector bool short vec_ldl (int, bool short *)",
38 "vector bool int vec_ldl (int, bool int *)",
39 "vector bool long long vec_ldl (int, bool long long *)",
40 "vector pixel vec_ldl (int, pixel *)",
41 "vector long long vec_ldl (int, long long *)",
42 "vector unsigned long long vec_ldl (int, unsigned long long *)".
43 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
44 type tree bool_long_long_type_node and correct definition of
45 bool_V2DI_type_node to make reference to this new type tree.
46 (rs6000_mangle_type): Replace erroneous reference to
47 bool_long_type_node with bool_long_long_type_node.
48 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
49 comments to emphasize sign distinctions for char and int types and
50 replace RS6000_BTI_bool_long constant with
51 RS6000_BTI_bool_long_long constant. Also add comment to restrict
52 use of RS6000_BTI_pixel.
53 (bool_long_type_node): Remove this macro definition.
54 (bool_long_long_type_node): New macro definition
56 2018-04-10 Jakub Jelinek <jakub@redhat.com>
58 PR rtl-optimization/85300
59 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
60 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
61 simplify_unary_operation fails.
63 2018-04-10 Martin Liska <mliska@suse.cz>
65 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
66 cgraph_edge and ipa_ref.
68 2018-04-10 Jakub Jelinek <jakub@redhat.com>
73 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
74 computation of the VEC_MERGE selector from mask.
75 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
76 Fix decoding of the VEC_MERGE selector into mask.
78 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
80 PR tree-optimization/85286
81 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
83 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
85 * final.c (final_1): Set insn_last_address as well as
88 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
91 * explow.c (emit_stack_probe): Call validize_mem on memory location
92 before passing it to gen_probe_stack. Create address operand and
93 legitimize it for the probe_stack_address case.
95 2018-04-09 Jan Hubicka <jh@suse.cz>
98 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
99 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
100 * tree.c (free_lang_data_in_type): Fix handling of binfos;
102 (free_lang_data): Rebuild type inheritance graph.
104 2018-04-09 Martin Sebor <msebor@redhat.com>
106 * invoke.texi (-finline-small-functions): Mention other optimization
108 (-findirect-inlining, -fpartial-inlining): Same.
109 (-finline-functions-called-once): Same.
110 (-freorder-blocks-and-partition): Same.
112 2018-04-09 Jan Hubicka <jh@suse.cz>
115 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
116 jumps; choose last target that matches the criteria (i.e.
117 no partition changes for non-crossing jumps).
118 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
119 support for redirecting crossing jumps to non-crossing.
121 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
123 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
124 also for naked functions.
126 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
128 * config/arc/arc.md (add_shift): New pattern.
129 (add_shift2): Likewise.
130 (sub_shift): Likewise.
131 (sub_shift_cmp0_noout): Likewise.
132 (compare_si_ashiftsi): Likewise.
133 (xbfu_cmp0_noout): New combine pattern.
134 (xbfu_cmp0"): Likewise.
135 (movsi_set_cc_insn): Place the predicable variant first.
136 (commutative_binary_cmp0_noout): Remove clobber.
137 (commutative_binary_cmp0): New pattern.
138 (noncommutative_binary_cmp0): Likewise.
139 (noncommutative_binary_cmp0_noout): Likewise.
140 (noncommutative_binary_comparison_result_used): Removed.
141 (rsub_cmp0): New pattern.
142 (rsub_cmp0_noout): Likewise.
143 (extzvsi): Changed, keep only meaningful variants.
144 (SQH, SEZ): New iterators.
145 (SQH_postfix): New mode attribute.
146 (SEZ_prefix): New code attribute.
147 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
148 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
149 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
151 (noncommutative_operator): Check the availability of barrel
154 2018-04-09 Richard Biener <rguenther@suse.de>
156 PR tree-optimization/85284
157 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
158 Only use the niter constraining form of simple_iv when the exit
161 2018-04-09 Tom de Vries <tom@codesourcery.com>
164 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
165 (define_expand "*memory_barrier"): New define_expand.
166 (define_insn "memory_barrier"): New insn.
168 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
170 PR rtl-optimization/80463
171 PR rtl-optimization/83972
172 PR rtl-optimization/83480
174 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
175 correct producer for the insn.
176 (tidy_control_flow): Fixup seqnos in case of debug insns.
178 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
180 PR rtl-optimization/83913
182 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
183 different sched-times when merging exprs.
185 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
187 PR rtl-optimization/83962
189 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
190 tidy_fallthru_edge and tidy_control_flow.
192 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
194 PR rtl-optimization/83530
196 * sel-sched.c (force_next_insn): New global variable.
197 (remove_insn_for_debug): When force_next_insn is true, also leave only
198 next insn in the ready list.
199 (sel_sched_region): When the region wasn't scheduled, make another pass
200 over it with force_next_insn set to 1.
202 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
204 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
206 * config/nds32/constants.md (unspec_volatile_element): Add enum values
207 for interrupt control.
208 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
209 functions for interrupt control.
210 * config/nds32/nds32-intrinsic.md: Likewise.
211 * config/nds32/nds32_intrinsic.h: Likewise.
212 * config/nds32/nds32.h (nds32_builtins): Likewise.
214 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
216 * config/nds32/nds32.c (nds32_init_machine_status,
217 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
218 strict_aligned_p field.
219 (nds32_expand_to_rtl_hook): New function.
220 (TARGET_EXPAND_TO_RTL_HOOK): Define.
221 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
223 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
224 Chung-Ju Wu <jasonwucj@gmail.com>
226 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
227 * config/nds32/nds32-n7.md: New file.
228 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
229 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
231 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
232 * config/nds32/nds32.md (pipeline_model): Add n7.
233 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
234 * config/nds32/pipelines.md: Include n7 settings.
236 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
237 Chung-Ju Wu <jasonwucj@gmail.com>
239 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
240 * config/nds32/nds32-e8.md: New file.
241 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
242 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
244 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
245 * config/nds32/nds32.md (pipeline_model): Add e8.
246 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
247 * config/nds32/pipelines.md: Include e8 settings.
249 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
250 Chung-Ju Wu <jasonwucj@gmail.com>
252 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
253 * config/nds32/nds32-n8.md: New file.
254 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
255 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
257 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
258 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
259 * config/nds32/nds32.md (pipeline_model): Add n8.
260 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
261 * config/nds32/pipelines.md: Include n8 settings.
263 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
264 Chung-Ju Wu <jasonwucj@gmail.com>
266 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
267 * config/nds32/nds32-n9-2r1w.md: New file.
268 * config/nds32/nds32-n9-3r2w.md: New file.
269 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
270 nds32_register_ports): New or modify for cpu n9.
271 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
273 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
274 * config/nds32/nds32-utils.c: New file.
275 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
276 TARGET_MUL_SLOW): Define.
277 * config/nds32/nds32.md (pipeline_model): New attribute.
278 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
279 New options that support cpu n9.
280 * config/nds32/pipelines.md: Include n9 settings.
281 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
283 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
285 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
286 information if necessary.
287 (output_cond_branch_compare_zero): Likewise.
288 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
289 (nds32_target_alignment): Refine for alignment.
290 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
291 (FUNCTION_BOUNDARY): Modify.
292 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
294 * config/nds32/nds32.opt (malways-align, malign-functions): New.
296 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
298 * config/nds32/constants.md (unspec_volatile_element): Add values for
299 TLB operation and data prefetch.
300 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
301 functions for TLB operation and data prefetch.
302 * config/nds32/nds32-intrinsic.md: Likewise.
303 * config/nds32/nds32_intrinsic.h: Likewise.
304 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
305 (nds32_print_operand): Likewise.
306 * config/nds32/nds32.h (nds32_builtins): Likewise.
308 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
309 Andrew Pinski <pinsika@gcc.gnu.org>
312 * match.pd: Use constant_boolean_node of correct type instead of
313 boolean_true_node or boolean_false_node for simplifying
314 pointer comparisons to zero.
316 2018-04-07 Jakub Jelinek <jakub@redhat.com>
318 PR tree-optimization/80021
319 * tree.c (verify_type_variant): Make error call in verify_variant_match
320 translatable and remove final full stop.
322 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
324 * config/nds32/constants.md (unspec_volatile_element): Add
325 UNSPEC_VOLATILE_EH_RETURN.
326 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
327 nds32_output_stack_pop): Support dwarf exception handling process.
328 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
329 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
330 exception handling process.
331 (nds32_compute_stack_frame): Likewise.
332 (nds32_return_addr_rtx): Likewise.
333 (nds32_initial_elimination_offset): Likewise.
334 (nds32_expand_prologue): Likewise.
335 (nds32_expand_epilogue): Likewise.
336 (nds32_dynamic_chain_address): New function.
337 * config/nds32/nds32.h (machine_function): Add fields for dwarf
339 (DYNAMIC_CHAIN_ADDRESS): Define.
340 (EH_RETURN_DATA_REGNO): Define.
341 (EH_RETURN_STACKADJ_RTX): Define.
342 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
343 patterns for dwarf exception handling.
345 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
347 * config/nds32/nds32.h: Clean up obsolete macros.
349 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
351 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
352 Add enum values for particular instructions.
353 * config/nds32/nds32-intrinsic.c: Implementation of expanding
354 particular intrinsic functions.
355 * config/nds32/nds32-intrinsic.md: Likewise.
356 * config/nds32/nds32_intrinsic.h: Likewise.
357 * config/nds32/nds32.h (nds32_builtins): Likewise.
358 * config/nds32/nds32.md (type): Add pbsad and pbsada.
359 (btst, ave): New patterns for particular instructions.
361 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
363 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
364 Add enum values for atomic load/store and memory sync.
365 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
367 * config/nds32/nds32-intrinsic.md: Likewise.
368 * config/nds32/nds32_intrinsic.h: Likewise.
369 * config/nds32/nds32.h (nds32_builtins): Likewise.
371 2018-04-07 Jakub Jelinek <jakub@redhat.com>
373 PR tree-optimization/85257
374 * fold-const.c (native_encode_vector): If not all elts could fit
375 and off is -1, return 0 rather than offset.
376 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
377 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
378 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
379 adjust buffer in native_interpret_expr call.
381 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
383 * config/nds32/constants.md (unspec_volatile_element): Add cache
385 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
386 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
387 * config/nds32/nds32.c (nds32_cctl_names): New.
388 (nds32_print_operand): Handle cache control register names.
389 * config/nds32/nds32.h (nds32_builtins): New enum values.
390 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
392 * config/nds32/nds32.md (type): Add mmu.
393 * config/nds32/pipelines.md (simple_insn): Add mmu.
395 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
397 * config/nds32/nds32.md (type): Remove call.
398 * config/nds32/pipelines.md (simple_insn): Likewise.
400 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
402 * config/nds32/constants.md (unspec_volatile_element): Add
403 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
404 UNSPEC_VOLATILE_FMFCFG.
405 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
406 description for fmfcfg and fmfcsr.
407 (bdesc_1arg): Add fmtcsr.
408 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
409 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
410 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
411 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
412 unspec_fmfcfg): New patterns.
413 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
414 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
415 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
416 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
417 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
418 __nds32__fmfcfg): Define.
420 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
422 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
423 intrinsic register names.
424 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
425 intrinsic register enum values and macros.
427 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
429 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
430 for load/store addressing form.
431 (nds32_print_operand_address): Likewise.
433 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
436 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
437 based on LABEL_REF. Remove useless assertion.
438 (pic_address_needs_scratch): Fix formatting.
439 (sparc_legitimize_pic_address): Minor tweaks.
440 (sparc_delegitimize_address): Adjust assertion accordingly.
441 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
442 into symbolic_operand.
443 (movsi_high_pic_label_ref): Likewise.
444 (movsi_lo_sum_pic_label_ref): Likewise.
445 (movdi_pic_label_ref): Likewise.
446 (movdi_high_pic_label_ref): Likewise.
447 (movdi_lo_sum_pic_label_ref): Likewise.
449 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
451 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
452 custom LIB_SPEC setup.
454 2018-04-06 Ruslan Bukin <br@bsdpad.com>
455 Kito Cheng <kito.cheng@gmail.com>
457 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
458 * config/riscv/freebsd.h: New.
460 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
462 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
463 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
466 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
467 Kito Cheng <kito.cheng@gmail.com>
469 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
470 nds32_output_call, nds32_symbol_binds_local_p): New functions.
471 * config/nds32/nds32-protos.h (nds32_output_call,
472 nds32_output_return): Declare.
473 * config/nds32/nds32.md: Refine all the call and return patterns.
475 2018-04-06 Jakub Jelinek <jakub@redhat.com>
478 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
479 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
481 PR rtl-optimization/84872
482 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
483 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
486 2018-04-06 Tamar Christina <tamar.christina@arm.com>
488 * expr.c (copy_blkmode_to_reg): Revert 254862.
489 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
491 2018-04-06 Richard Biener <rguenther@suse.de>
494 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
495 after seeing a component reference with an adjacent field. Treat
496 refs to arrays at struct end of external decls similar to
497 refs to unconstrained commons.
499 2018-04-06 Jakub Jelinek <jakub@redhat.com>
502 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
503 look through SAVE_EXPRs with non-side-effects argument. Adjust
505 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
508 2018-04-06 Richard Biener <rguenther@suse.de>
511 * alias.c (find_base_term): New wrapper around find_base_term
512 unwinding CSELIB_VAL_PTR changes.
513 (find_base_term): Do not restore CSELIB_VAL_PTR during the
516 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
518 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
520 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
521 constant definitions.
522 ("nop"): lr 0,0 -> nopr r0
523 ("nop_lr0", "nop_lr1"): New insn definitions.
525 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
527 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
528 NDS32_V3PUSH_AVAILABLE_P macro.
530 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
531 Chung-Ju Wu <jasonwucj@gmail.com>
533 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
534 (nds32*-*-*): Add float and fpu_config into supported_defaults.
535 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
536 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
537 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
538 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
539 * config/nds32/constraints.md: New constraints and checking for hard
541 * config/nds32/iterators.md: New mode iterator and attribute for hard
543 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
545 * config/nds32/nds32-fpu.md: New file.
546 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
547 deal with hard float code generation.
548 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
550 (abi_type, float_reg_number): New enum type.
551 * config/nds32/nds32-predicates.c: New predicates for hard float.
552 * config/nds32/nds32-protos.h: Declare functions for hard float.
553 * config/nds32/nds32.c: Implementation for hard float configuration.
554 * config/nds32/nds32.h: Definitions for hard float configuration.
555 * config/nds32/nds32.md: Include hard float machine description and
556 modify patterns for hard float configuration.
557 * config/nds32/nds32.opt: New options for hard float configuration.
558 * config/nds32/predicates.md: New predicates for hard float
561 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
563 * common/config/nds32/nds32-common.c
564 (nds32_option_optimization_table): Enable -mreleax-hint by default.
566 2018-04-05 Jakub Jelinek <jakub@redhat.com>
569 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
570 CONSTRUCTOR_ELT (ctor, ...)->value.
572 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
575 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
577 2018-04-05 Tom de Vries <tom@codesourcery.com>
580 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
583 2018-04-05 Shiva Chen <shiva0217@gmail.com>
584 Kito Cheng <kito.cheng@gmail.com>
586 * config/nds32/constraints.md (U33): Fine-tune checking condition.
587 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
588 * config/nds32/nds32.h (nds32_16bit_address_type): Add
589 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
591 2018-04-05 Shiva Chen <shiva0217@gmail.com>
592 Kito Cheng <kito.cheng@gmail.com>
594 * config/nds32/constraints.md (Ufe): New memory constraint.
595 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
596 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
597 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
599 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
600 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
602 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
604 * config/nds32/nds32.md: Use optimize_size in the condition for
605 alu-shift instructions.
607 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
609 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
611 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
613 * config/nds32/nds32.md (negsi2): Refine pattern.
615 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
616 Chung-Ju Wu <jasonwucj@gmail.com>
618 * config/nds32/iterators.md (shift_rotate): New code iterator.
619 (shift): New code attribute.
620 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
621 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
622 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
623 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
625 (andsi3, *andsi3): Ditto.
626 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
627 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
628 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
629 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
630 nds32_ior_operand, nds32_xor_operand): New predicates.
632 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
634 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
635 (addsi3, subsi3): ... this.
637 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
639 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
641 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
643 * config/nds32/nds32.md: Adjust indention.
645 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
647 * config/nds32/nds32.md (feature): New attribute.
649 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
651 * config/nds32/nds32.md (subtype): New attribute.
653 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
656 * config/arm/arm-builtins.c (arm_expand_builtin): Change
657 expansion to perform a bitwise AND of the argument followed by a
658 boolean negation of the result.
660 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
662 PR rtl-optimization/84878
663 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
664 the basic block. Assert the use reference is not artificial and that
665 it has an associated insn.
667 2018-04-04 Michael Matz <matz@suse.de>
669 * builtins.c (compute_objsize): Pass correct operand
670 to array_at_struct_end_p.
672 2018-04-04 Richard Biener <rguenther@suse.de>
675 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
676 from contexts for DINFO_LEVEL_TERSE and below.
678 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
680 * config/nds32/nds32-doubleword.md (move_<mode>): Require
681 resiter_operand condition.
682 * config/nds32/nds32.md (*move<mode>): Ditto.
684 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
685 Monk Chiang <sh.chiang04@gmail.com>
687 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
689 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
691 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
693 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
694 Kito Cheng <kito.cheng@gmail.com>
696 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
697 nds32_cond_code_str, output_cond_branch,
698 output_cond_branch_compare_zero, nds32_expand_cbranch,
699 nds32_expand_cstore, nds32_expand_movcc,
700 nds32_output_cbranchsi4_equality_zero,
701 nds32_output_cbranchsi4_equality_reg,
702 nds32_output_cbranchsi4_equality_reg_or_const_int,
703 nds32_output_cbranchsi4_greater_less_zero: New functions.
704 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
705 nds32_expand_cstore, nds32_expand_movcc,
706 nds32_output_cbranchsi4_equality_zero,
707 nds32_output_cbranchsi4_equality_reg,
708 nds32_output_cbranchsi4_equality_reg_or_const_int,
709 nds32_output_cbranchsi4_greater_less_zero): Declare.
710 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
711 nds32_rimm11s_operand): New predicates.
712 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
713 * config/nds32/nds32.md: Rewrite all the branch and conditional move
716 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
718 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
719 * config/nds32/nds32.md: Ditto.
720 * config/nds32/pipelines.md: Ditto.
722 2018-04-04 Richard Biener <rguenther@suse.de>
724 PR tree-optimization/85168
725 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
726 propagating abnormals.
728 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
730 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
732 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
733 Kito Cheng <kito.cheng@gmail.com>
735 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
736 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
737 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
738 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
739 * config/nds32/nds32.md (sibcall_internal): New.
740 (sibcall_register): Remove.
741 (sibcall_immediate): Remove.
742 (sibcall_value_internal): New.
743 (sibcall_value_register): Remove.
744 (sibcall_value_immediate): Remove.
745 * config/nds32/predicates.md (nds32_general_register_operand): New.
746 (nds32_call_address_operand): New.
748 2018-04-03 Jakub Jelinek <jakub@redhat.com>
750 PR rtl-optimization/85167
751 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
752 bb_defs if *split_p, instead preinitialize it to NULL.
754 PR tree-optimization/85156
755 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
756 evaluating the argument multiple times.
758 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
760 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
762 (_mm_cvtpd_ps): Likewise.
763 (_mm_cvttpd_epi32): Likewise.
764 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
765 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
766 vector, pixel, and bool following altivec.h include.
768 2018-04-03 Martin Sebor <msebor@redhat.com>
770 * doc/extend.texi (Common Function Attributes): Clarify.
771 (const attribute): Likewise.
772 (pure attribute): Likewise.
774 2018-04-03 Jakub Jelinek <jakub@redhat.com>
777 * config/i386/i386.c (ix86_expand_vector_set): Use
778 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
780 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
782 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
783 instructions when changing rounding bits to preserve precision bits
784 in the x87 control word.
786 2018-04-03 Martin Liska <mliska@suse.cz>
788 PR tree-optimization/82491
789 * rtl.h (strip_offset_and_add): Replace += suboffset with
790 poly_uint64 () + suboffset.
792 2018-03-29 Martin Liska <mliska@suse.cz>
793 Martin Jambor <mjambor@suse.cz>
796 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
797 param_type is not an integral or pointer type.
799 2018-04-03 Richard Biener <rguenther@suse.de>
801 * sese.h (recompute_all_dominators): Remove.
803 2018-04-02 Martin Sebor <msebor@redhat.com>
805 * doc/invoke.texi (-Wrestrict): Fix typos.
807 2018-04-02 Jim Wilson <jimw@sifive.com>
809 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
810 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
811 (<optab>di3, <optab>si3_extend): Likewise.
812 (<optab>si3_mask, <optab>si3_mask_1): New.
813 (<optab>di3_mask, <optab>di3_mask_1): New.
814 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
815 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
816 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
818 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
820 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
823 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
825 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
826 (nds32_canonicalize_comparison): New function.
828 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
829 Kito Cheng <kito.cheng@gmail.com>
830 Kuan-Lin Chen <kuanlinchentw@gmail.com>
832 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
833 * config/nds32/constants.md (unspec_volatile_element): Add
834 UNSPEC_VOLATILE_RELAX_GROUP.
835 * config/nds32/nds32-relax-opt.c: New file.
836 * config/nds32/nds32-predicates.c
837 (nds32_symbol_load_store_p): New function.
838 * config/nds32/nds32-protos.h
839 (nds32_symbol_load_store_p): Declare function.
840 (make_pass_nds32_relax_opt): Declare new rtl pass function.
841 * config/nds32/nds32.c
842 (nds32_register_pass): New function to register pass.
843 (nds32_register_passes): New function to register passes.
844 * config/nds32/nds32.md (relax_group): New pattern.
845 * config/nds32/nds32.opt (mrelax-hint): New option.
846 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
848 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
850 * config/nds32/t-nds32: Modify files dependency.
852 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
854 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
855 (PROFILE_HOOK): Define its implementation.
857 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
859 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
860 type and 32-bit size.
862 2018-04-01 Jakub Jelinek <jakub@redhat.com>
865 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
866 (V_128_256): New mode iterator.
867 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
868 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
869 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
871 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
872 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
874 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
877 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
878 NaN inputs correctly.
880 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
883 * config/rs6000/vsx.md (??r): New mode attribute.
884 (*vsx_mov<mode>_64bit): Use it.
885 (*vsx_mov<mode>_32bit): Likewise.
887 2018-03-30 Martin Sebor <msebor@redhat.com>
889 PR tree-optimization/84818
890 * builtins.c (check_access): Use warning_n.
892 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
895 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
897 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
900 2018-03-30 Julia Koval <julia.koval@intel.com>
903 * x86-tune.def (movx, partial_reg_dependency): Enable for
906 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
909 * lra-constraints.c (process_alt_operands): Move setting
910 this_alternative_matches below.
912 2018-03-29 Martin Liska <mliska@suse.cz>
915 * doc/invoke.texi: Document how LTO works with debug info.
916 Describe auto-load support of binutils. Mention 'x86-64'
917 as valid option value of -march option.
919 2018-03-29 Jakub Jelinek <jakub@redhat.com>
921 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
924 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
925 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
926 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
929 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
932 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
933 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
934 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
935 for RS6000_BTM_POWERPC64.
936 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
937 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
938 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
943 2018-03-28 Carl Love <cel@us.ibm.com>
946 2017-09-27 Carl Love <cel@us.ibm.com>
948 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
949 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
950 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
953 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
955 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
956 instead of __vector bool.
957 (_mm_max_pu8): Likewise.
958 (_mm_min_pi16): Likewise.
960 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
963 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
967 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
969 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
970 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
971 (div_extend): Likewise.
972 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
974 (__builtin_divweuo): Likewise.
975 (__builtin_divdeo): Likewise.
976 (__builtin_divdeuo): Likewise.
978 2018-03-28 Jakub Jelinek <jakub@redhat.com>
981 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
982 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
984 PR tree-optimization/82004
985 * gimple-match-head.c (optimize_pow_to_exp): New function.
986 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
987 Don't fold to exp if optimize_pow_to_exp is false.
989 2018-03-28 Martin Liska <mliska@suse.cz>
992 * calls.c (initialize_argument_information): Fix trailing space.
993 * common.opt: Fix typo and provide better explanation for
994 -fsanitize-coverage option.
995 * config/i386/i386.opt: Fix typo.
997 2018-03-28 Jakub Jelinek <jakub@redhat.com>
998 Martin Liska <mliska@suse.cz>
1001 * gimplify.c (asan_poison_variable): Don't do the check for
1002 gimplify_omp_ctxp here.
1003 (gimplify_decl_expr): Do it here.
1004 (gimplify_target_expr): Likewise.
1006 2018-03-28 Martin Liska <mliska@suse.cz>
1009 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1010 chkp_type_bounds_count if MPX is not enabled.
1012 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1014 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1016 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1019 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1020 function to create the function decl for complex long double
1021 multiply and divide for -mabi=ieeelongdouble.
1022 (init_float128_ieee): Call it.
1024 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1027 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1028 -fcf-protection=branch -mibt.
1029 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1031 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1034 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1036 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1039 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1040 extern array declarations.
1042 2018-03-27 Richard Biener <rguenther@suse.de>
1045 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1046 explicit single_use checks.
1048 2018-03-27 Richard Biener <rguenther@suse.de>
1050 PR tree-optimization/85082
1051 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1054 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1056 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1057 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1058 Turn on fasynchronous-unwind-tables and funwind-tables.
1060 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1063 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1064 (*bmi_blsr_<mode>_ccz): Ditto.
1066 2018-03-26 Tom de Vries <tom@codesourcery.com>
1068 PR tree-optimization/85063
1069 * omp-general.c (offloading_function_p): New function. Factor out
1071 * omp-offload.c (pass_omp_target_link::gate): ... here.
1072 * omp-general.h (offloading_function_p): Declare.
1073 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1074 with attribute omp declare target for offloading functions.
1076 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1078 PR tree-optimization/84005
1079 * tree-data-ref.h (get_base_for_alignment): Declare.
1080 * tree-data-ref.c (get_base_for_alignment_1): New function.
1081 (get_base_for_alignment): Likewise.
1082 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1083 get_base_for_alignment to find a suitable base object, instead
1084 of always using drb->base_address.
1086 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1089 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1090 known size by default.
1092 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1095 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1096 and non BLKmode operands.
1098 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1101 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1102 Clean up attributes.
1104 2018-03-23 Richard Biener <rguenther@suse.de>
1107 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1108 we are going to emit early debug for LTO.
1110 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1113 * function.c (match_asm_constraints_1): Don't optimize if input
1114 doesn't satisfy general_operand predicate for output's mode.
1117 * alias.c (write_dependence_p): Don't require for x_canonicalized
1118 non-VOIDmode if x has VOIDmode.
1121 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1122 just don't try to optimize it rather than assert it never happens.
1124 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1126 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1127 macro expansions for definition of ST_INTERNAL_<mode> and
1128 LD_INTERNAL_<mode> builtins.
1129 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1131 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1133 (altivec_expand_st_builtin): Likewise.
1134 (altivec_expand_builtin): Remove calls to deleted functions.
1135 (rs6000_address_for_altivec): Delete this function.
1136 * config/rs6000/vector.md: Remove expands for
1137 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1139 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1142 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1143 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1144 re-computing once computed.
1145 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1146 (arm_init_machine_status): Initialize
1147 machine->static_chain_stack_bytes.
1149 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1152 * doc/extend.texi: Add four new prototypes for vec_ld.
1153 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1154 definitions for more logical presentation.
1155 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1156 entries for V1TI variants of __builtin_altivec_ld builtin.
1157 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1158 handling of V1TI variant of LVX icode pattern.
1159 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1160 (rs6000_gimple_fold_builtin): Likewise.
1161 (altivec_init_builtins): Add code to define
1162 __builtin_altivec_lvx_v1ti function.
1164 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1167 * function.c (match_asm_constraints_1): Don't do the optimization
1168 if input isn't a REG, SUBREG, MEM or constant.
1170 2018-03-22 Tom de Vries <tom@codesourcery.com>
1172 PR tree-optimization/84956
1173 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1174 bb_has_abnormal_pred.
1176 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1179 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1180 DECL_INITIAL (decl) to decl at the end.
1181 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1184 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1186 * doc/extend.texi (__builtin_tgmath): Document when complex
1187 integer types are treated as _Complex _Float64.
1189 2018-03-21 Tom de Vries <tom@codesourcery.com>
1191 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1193 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1195 PR tree-optimization/84960
1196 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1197 if it is ENTRY block, move them into single succ of ENTRY in that case.
1199 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1201 PR tree-optimization/84811
1202 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1203 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1204 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1205 (known_subrange_p): Update accordingly. Cast each value involved
1206 in the size comparison, rather than casting the result of the
1209 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1211 PR tree-optimization/84982
1212 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1213 by flipping the least significant bit rather than all bits from
1214 bitpos to bitpos + bitsize - 1.
1216 2018-03-21 Nathan Sidwell <nathan@acm.org>
1218 * doc/extend.texi (Deprecated Features): Remove mention of
1219 long-deleted deprecations.
1221 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1224 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1225 * configure: Regenerate.
1227 2018-03-21 Tom de Vries <tom@codesourcery.com>
1229 PR tree-optimization/83126
1230 * tree-parloops.c (num_phis): New function.
1231 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1233 2018-03-21 Nathan Sidwell <nathan@acm.org>
1235 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1236 mention anon-struct/union members and trailing attributes.
1238 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1240 PR tree-optimization/84969
1241 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1242 builtin memset partitions if they set different rhs values.
1244 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1246 PR rtl-optimization/84989
1247 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1248 VEC_DUPLICATE with scalar result mode.
1250 2018-03-21 Martin Liska <mliska@suse.cz>
1253 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1254 not intended return statement.
1256 2018-03-21 Martin Liska <mliska@suse.cz>
1259 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1260 (chkp_find_bound_slots_1): Limit number of iterations.
1262 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1265 * Minor grammar fixes for x86 options.
1267 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1270 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1271 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1273 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1276 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1277 (altivec_lvx_<mode>_1op): Likewise.
1278 (altivec_stvx_<mode>_2op): Likewise.
1279 (altivec_stvx_<mode>_1op): Likewise.
1280 (altivec_lvx_<VM2:mode>): New define_expand.
1281 (altivec_stvx_<VM2:mode>): Likewise.
1282 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1283 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1284 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1285 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1286 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1287 (rs6000_gen_lvx): Likewise.
1288 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1289 (altivec_expand_stv_builtin): Likewise.
1290 (altivec_expand_builtin): Likewise.
1291 * config/rs6000/vector.md: Likewise.
1293 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1296 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1299 2018-03-20 Richard Biener <rguenther@suse.de>
1302 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1303 sign-conversions as zero, fall back to standard scalar_stmt
1306 2018-03-20 Martin Liska <mliska@suse.cz>
1309 * predict.c (rebuild_frequencies): Handle case when we have
1310 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1312 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1315 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1316 flag_section_anchors.
1317 * varasm.c (use_blocks_for_decl_p): Remove hack for
1318 dw2_force_const_mem.
1321 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1323 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1324 be created, use lowpart_subreg of operands[0] rather than operands[0]
1326 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1327 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1328 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1329 and n constraint instead of aarch64_shift_imm_di and Usd.
1330 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1331 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1333 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1336 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1337 to favor GPR over NEON registers.
1338 (<shift>di3_neon): Likewise.
1340 2018-03-20 Tom de Vries <tom@codesourcery.com>
1343 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1344 (nvptx_process_pars): Emit bar.sync asap and alap.
1346 2018-03-20 Tom de Vries <tom@codesourcery.com>
1349 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1350 seen_label if seen_label is already set.
1352 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1355 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1356 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1357 Use 1U instead of 1. Formatting fixes.
1360 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1361 instead of TREE_TYPE (s1) for the return value.
1363 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1365 PR tree-optimization/84946
1366 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1367 bitsize + bitsize in poly_uint64 rather than poly_int64.
1370 * dwarf2asm.c: Include fold-const.c.
1371 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1372 of decl rather than decl itself.
1374 PR rtl-optimization/84643
1375 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1377 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1380 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1381 calling assemble_variable.
1383 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1386 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1387 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1389 2018-03-19 Jim Wilson <jimw@sifive.com>
1392 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1393 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1394 (riscv_first_stack_step): Likewise.
1395 (riscv_option_override): Use STACK_BOUNDARY instead of
1397 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1399 (BIGGEST_ALIGNMENT): Set to 128.
1400 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1401 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1404 2018-03-19 Richard Biener <rguenther@suse.de>
1406 PR tree-optimization/84933
1407 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1408 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1410 2018-03-19 Richard Biener <rguenther@suse.de>
1412 PR tree-optimization/84859
1413 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1414 (cond_if_else_store_replacement): Perform sinking operation on
1415 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1416 Generalize what a BB with a single eligible store is.
1418 2018-03-19 Richard Biener <rguenther@suse.de>
1420 PR tree-optimization/84929
1421 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1422 chrec_is_positive against non-chrec arg.
1424 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1427 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1429 2018-03-18 Martin Liska <mliska@suse.cz>
1431 PR rtl-optimization/84635
1432 * regrename.c (build_def_use): Use matches_mode only when
1435 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1437 PR tree-optimization/84913
1438 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1439 vectorize chains of COND_EXPRs.
1441 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1443 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1445 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1447 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1449 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1451 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1453 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1454 Kito Cheng <kito.cheng@gmail.com>
1456 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1457 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1458 (nds32_adjust_reg_alloc_order): New function.
1459 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1461 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1463 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1464 nds32_print_operand, nds32_print_operand_address): Use
1465 HOST_WIDE_INT_PRINT_DEC instead.
1467 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1469 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1471 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1474 * config/i386/i386.c (initial_ix86_tune_features,
1475 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1477 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1478 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1479 rather than 1u << ix86_tune. Formatting fix.
1480 (ix86_option_override_internal): Change ix86_arch_mask from
1481 unsigned int to unsigned HOST_WIDE_INT, initialize to
1482 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1483 (ix86_function_specific_restore): Likewise.
1485 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1488 * postreload.c (reload_combine_recognize_pattern): Perform
1489 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1490 truncate_int_for_mode the result for the destination's mode.
1493 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1494 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1496 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1499 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1502 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1503 regno_allocno_class_array and sorted_pseudos.
1504 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1505 insns where regno is used.
1507 2018-03-16 Martin Liska <mliska@suse.cz>
1510 * multiple_target.c (create_dispatcher_calls): Redirect
1511 reference in the symbol table.
1513 2018-03-16 Martin Liska <mliska@suse.cz>
1516 * multiple_target.c (create_dispatcher_calls): Redirect also
1519 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1523 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1524 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1525 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1527 2018-03-16 Julia Koval <julia.koval@intel.com>
1529 * doc/invoke.texi (Skylake Server): Add CLWB.
1530 Cannonlake): Remove CLWB.
1532 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1534 PR tree-optimization/84841
1535 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1537 (FLOAT_ONE_CONST_TYPE): Define.
1538 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1539 (sort_by_operand_rank): Put entries with higher constant_type last
1540 rather than first to match comments.
1542 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1544 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1547 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1550 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1551 check and use address of target always.
1553 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1556 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1557 (indirect_thunk_bnd_needed): Likewise.
1558 (indirect_thunks_used): Likewise.
1559 (indirect_thunks_bnd_used): Likewise.
1560 (indirect_return_needed): New.
1561 (indirect_return_bnd_needed): Likewise.
1562 (output_indirect_thunk_function): Add a bool argument for
1564 (output_indirect_thunk_function): Don't generate alias for
1565 function return thunk.
1566 (ix86_code_end): Call output_indirect_thunk_function to generate
1567 function return thunks.
1568 (ix86_output_function_return): Set indirect_return_bnd_needed
1569 and indirect_return_needed instead of indirect_thunk_bnd_needed
1570 and indirect_thunk_needed.
1572 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1574 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1575 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1576 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1578 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1579 Paul Hua <paul.hua.gm@gmail.com>
1582 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1584 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1586 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1587 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1590 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1593 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1594 instead of GET_MODE_SIZE when comparing Units.
1596 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1599 * varasm.c (hash_section): Return an unchangeble hash value
1600 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1601 Return !aarch64_can_use_per_function_literal_pools_p ().
1603 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1606 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1607 rather than address of cmode as last argument to prepare_cmp_insn.
1609 2018-03-15 Julia Koval <julia.koval@intel.com>
1611 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1612 F_AVX512VNNI, F_AVX512BITALG): New.
1614 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1617 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1618 insn for floating-point loads and stores.
1620 2018-03-14 Carl Love <cel@us.ibm.com>
1622 * config/rs6000/rs6000-c.c: Add macro definitions for
1623 ALTIVEC_BUILTIN_VEC_PERMXOR.
1624 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1625 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1626 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1627 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1629 * config/doc/extend.texi: Add prototypes for vec_permxor.
1631 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1634 * diagnostic-show-locus.c (class layout_point): Convert m_line
1635 from int to linenum_type.
1636 (line_span::comparator): Use linenum "compare" function when
1637 comparing line numbers.
1638 (test_line_span): New function.
1639 (layout_range::contains_point): Convert param "row" from int to
1641 (layout_range::intersects_line_p): Likewise.
1642 (layout::will_show_line_p): Likewise.
1643 (layout::print_source_line): Likewise.
1644 (layout::should_print_annotation_line_p): Likewise.
1645 (layout::print_annotation_line): Likewise.
1646 (layout::print_leading_fixits): Likewise.
1647 (layout::annotation_line_showed_range_p): Likewise.
1648 (struct line_corrections): Likewise for field m_row.
1649 (line_corrections::line_corrections): Likewise for param "row".
1650 (layout::print_trailing_fixits): Likewise.
1651 (layout::get_state_at_point): Likewise.
1652 (layout::get_x_bound_for_row): Likewise.
1653 (layout::print_line): Likewise.
1654 (diagnostic_show_locus): Likewise for locals "last_line" and
1656 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1657 * input.c (selftest::test_linenum_comparisons): New function.
1658 (selftest::input_c_tests): Call it.
1659 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1660 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1661 * selftest.h (ASSERT_GT): New macro.
1662 (ASSERT_GT_AT): New macro.
1663 (ASSERT_LT): New macro.
1664 (ASSERT_LT_AT): New macro.
1666 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1668 PR rtl-optimization/84780
1669 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1671 2018-03-14 Martin Liska <mliska@suse.cz>
1673 * tree.c (record_node_allocation_statistics): Use
1674 get_stats_node_kind.
1675 (get_stats_node_kind): New function extracted from
1676 record_node_allocation_statistics.
1677 (free_node): Use get_stats_node_kind.
1679 2018-03-14 Richard Biener <rguenther@suse.de>
1681 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1682 that the value-set of ANTIC_IN doesn't grow.
1685 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1687 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1688 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1690 2018-03-14 Julia Koval <julia.koval@intel.com>
1692 * config.gcc (icelake-client, icelake-server): New.
1694 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1695 (initial_ix86_arch_features): Ditto.
1696 (PTA_SKYLAKE): Add SGX.
1697 (PTA_ICELAKE): Remove.
1698 (PTA_ICELAKE_CLIENT): New.
1699 (PTA_ICELAKE_SERVER): New.
1700 (ix86_option_override_internal): Split up icelake on icelake client and
1702 (get_builtin_code_for_version): Ditto.
1703 (fold_builtin_cpu): Ditto.
1704 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1705 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1706 * config/i386/i386.h (processor_type): Ditto.
1707 * doc/invoke.texi: Ditto.
1709 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1712 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1713 INTEGER_CST offset, add it together with bitpos / 8 and
1714 sign extend based on POINTER_SIZE.
1718 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1721 * config/i386/constraints.md (Yc): New register constraint.
1722 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1723 Use Yc constraint for alternative 2 of operand 0. Remove
1724 preferred_for_speed attribute.
1726 2018-03-14 Richard Biener <rguenther@suse.de>
1728 PR tree-optimization/84830
1729 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1730 with the old one to avoid oscillations.
1732 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1735 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1737 (assign_by_spills): Return a flag of reload assignment failure.
1738 Do not process the reload assignment failures. Do not spill other
1739 reload pseudos if they has the same reg class. Update n if
1741 (lra_assign): Add a return arg. Set up from the result of
1742 assign_by_spills call.
1743 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1744 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1745 usage_insns if it is not NULL.
1746 (spill_hard_reg_in_range): New function.
1747 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1748 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1749 function prototypes.
1750 (lra_assign): Change prototype.
1751 * lra.c (lra): Add code to deal with fails by splitting hard reg
1754 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1756 * config/riscv/riscv.opt (mrelax): New option.
1757 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1758 "norelax" when riscv_mrelax is disabled.
1759 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1761 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1764 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1765 reassociation for int modes.
1767 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1769 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1770 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1772 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1773 * config/aarch64/aarch64-sve.md
1774 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1775 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1776 (*extend<mode><Vwide>2): Rename to...
1777 (aarch64_sve_extend<mode><Vwide>2): ...this.
1778 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1779 renaming the old pattern to...
1780 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1782 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1783 define_expand, renaming the old pattern to...
1784 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1785 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1786 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1787 account when deciding which SVE instruction the optab should use.
1788 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1790 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1792 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1793 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1794 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1795 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1796 (tlsdesc_small_<mode>): Turn a define_expand and use
1797 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1798 (tlsdesc_small_advsimd_<mode>): ...this.
1799 (tlsdesc_small_sve_<mode>): New pattern.
1801 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1803 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1804 (UNSPEC_UMUL_HIGHPART): New constants.
1805 (MUL_HIGHPART): New int iteraor.
1806 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1807 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1809 (*<su>mul<mode>3_highpart): New define_insn.
1811 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1814 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1817 2018-03-13 Martin Liska <mliska@suse.cz>
1820 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1822 (sem_item_optimizer::~sem_item_optimizer): Release it.
1823 (sem_item_optimizer::merge_classes): Register variable aliases.
1824 (sem_item_optimizer::fixup_pt_set): New function.
1825 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1826 * ipa-icf.h: Declare new variables and functions.
1828 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1831 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1832 integer_pow2p@2 and test integer_pow2p in condition.
1833 (A < 0 ? C : 0): Similarly for @1.
1836 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1837 characters starting at p contain '\0' character, don't look beyond
1841 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1842 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1845 * reg-stack.c (change_stack): Change update_end var from int to
1846 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1847 also call set_block_for_insn on the newly added insns and rescan.
1850 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1851 on the last operand.
1854 * tree.c (stabilize_reference_1): Return save_expr (e) for
1855 STATEMENT_LIST even if it doesn't have side-effects.
1857 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1859 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1861 2018-03-12 Renlin Li <renlin.li@arm.com>
1863 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1864 aarch64_output_scalar_simd_mov_immediate.
1866 2018-03-12 Martin Sebor <msebor@redhat.com>
1868 PR tree-optimization/83456
1869 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1870 for perfectly overlapping calls to memcpy.
1871 (gimple_fold_builtin_memory_chk): Same.
1872 (gimple_fold_builtin_strcpy): Handle no-warning.
1873 (gimple_fold_builtin_stxcpy_chk): Same.
1874 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1876 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1878 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1879 parameter. Use it for SFmode.
1880 (rs6000_function_arg_advance_1): Adjust.
1881 (rs6000_function_arg): Adjust.
1882 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1884 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1886 PR rtl-optimization/84169
1887 PR rtl-optimization/84780
1888 * combine.c (can_combine_p): Check for a 2-insn combination whether
1889 the destination register is used between the two insns, too.
1891 2018-03-12 Richard Biener <rguenther@suse.de>
1893 PR tree-optimization/84803
1894 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1895 for refs DR analysis didn't process.
1897 2018-03-12 Richard Biener <rguenther@suse.de>
1899 PR tree-optimization/84777
1900 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1901 force-vectorize loops ignore whether we are optimizing for size.
1903 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1905 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1906 (TARGET_MD_ASM_ADJUST): Define.
1908 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1909 Kito Cheng <kito.cheng@gmail.com>
1910 Chung-Ju Wu <jasonwucj@gmail.com>
1912 * config/nds32/nds32.c (nds32_compute_stack_frame,
1913 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1914 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1915 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1916 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1917 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1918 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1919 * config/nds32/nds32.md (prologue, epilogue): Use macro
1920 NDS32_V3PUSH_AVAILABLE_P to do checking.
1922 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1925 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1926 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1927 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1928 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1929 addition of most attributes on !orig_type_die or the attribute not
1930 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1932 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1933 Chung-Ju Wu <jasonwucj@gmail.com>
1935 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1937 * config/nds32/nds32.opt (mvh): New option.
1939 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1940 Chung-Ju Wu <jasonwucj@gmail.com>
1942 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1944 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1945 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1948 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1949 Chung-Ju Wu <jasonwucj@gmail.com>
1951 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1953 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1954 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1956 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1957 Kito Cheng <kito.cheng@gmail.com>
1958 Chung-Ju Wu <jasonwucj@gmail.com>
1960 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1961 UNSPEC_FFMISM and UNSPEC_FLMISM.
1962 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1963 for ffb, ffmism and flmism.
1964 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1965 (unspec_ffmism): Ditto.
1966 (unspec_flmism): Ditto.
1967 (nds32_expand_builtin_impl): Check if string extension is available.
1968 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1969 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1971 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1974 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1977 * lra-assigns.c (assign_by_spills): Return a flag of reload
1978 assignment failure. Do not process the reload assignment
1979 failures. Do not spill other reload pseudos if they has the same
1981 (lra_assign): Add a return arg. Set up from the result of
1982 assign_by_spills call.
1983 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1984 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1985 usage_insns if it is not NULL.
1986 (spill_hard_reg_in_range): New function.
1987 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1988 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1989 function prototypes.
1990 (lra_assign): Change prototype.
1991 * lra.c (lra): Add code to deal with fails by splitting hard reg
1994 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1997 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1999 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2002 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2003 (dw_val_node): Add val_symbolic_view.
2004 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2005 (symview_upper_bound): New.
2006 (new_line_info_table): Initialize symviews_since_reset.
2007 (dwarf2out_source_line): Count symviews_since_reset and set
2008 symview_upper_bound.
2009 (dw_val_equal_p): Handle symview.
2010 (add_AT_symview): New.
2011 (print_dw_val): Handle symview.
2012 (attr_checksum, attr_checksum_ordered): Likewise.
2013 (same_dw_val_p, size_of_die): Likewise.
2014 (value_format, output_die): Likewise.
2015 (add_high_low_attributes): Use add_AT_symview for entry_view.
2016 (dwarf2out_finish): Reset symview_upper_bound, clear
2019 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2022 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2023 Add strict argument and use it.
2024 (rs6000_split_multireg_move): Update for new strict argument.
2025 (mem_operand_gpr): Disallow all non-offsettable addresses.
2026 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2028 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2031 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2032 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2033 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2036 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2037 decl, use remap_type if we want to use the type.
2039 2018-03-09 Martin Sebor <msebor@redhat.com>
2041 PR tree-optimization/84526
2042 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2044 (builtin_access::generic_overlap): Be prepared to handle non-array
2047 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2049 PR rtl-optimization/84682
2050 * lra-constraints.c (process_address_1): Check is_address flag
2051 for address constraints.
2052 (process_alt_operands): Likewise.
2053 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2054 preprocess_constraints.
2055 * recog.h (preprocess_constraints): Add oploc parameter.
2057 * recog.c (preprocess_constraints): Test address_operand for
2058 CT_ADDRESS constraints.
2060 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2063 * lra-assigns.c (assign_by_spills): Return a flag of reload
2064 assignment failure. Do not process the reload assignment
2065 failures. Do not spill other reload pseudos if they has the same
2067 (lra_assign): Add a return arg. Set up from the result of
2068 assign_by_spills call.
2069 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2070 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2071 usage_insns if it is not NULL.
2072 (spill_hard_reg_in_range): New function.
2073 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2074 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2075 function prototypes.
2076 (lra_assign): Change prototype.
2077 * lra.c (lra): Add code to deal with fails by splitting hard reg
2080 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2083 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2084 Accept complain bool parameter. Only emit errors if it is true.
2085 (arm_parse_cpu_option_name): Likewise.
2086 (arm_target_thumb_only): Adjust callers of the above.
2087 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2088 prototype to take a default true bool parameter.
2089 (arm_parse_arch_option_name): Likewise.
2091 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2092 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2096 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2097 * configure: Regenerate.
2098 * configure.ac ("linker --version-script option"): New.
2099 ("linker soname option"): New.
2101 2018-03-09 Richard Biener <rguenther@suse.de>
2103 PR tree-optimization/84775
2104 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2105 immediate uses of predicate stmts and mark them modified.
2108 PR tree-optimization/84178
2109 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2111 (version_loop_for_if_conversion): Delay update_ssa call.
2112 (tree_if_conversion): Delay update_ssa until after predicate
2115 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2118 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2119 when the function accesses prior frames.
2121 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2124 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2125 gen_llsym, otherwise call maybe_gen_llsym.
2128 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2129 has ',' character inside of it.
2131 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2134 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2135 as clobbering CC_REGNUM.
2137 2018-03-08 Richard Biener <rguenther@suse.de>
2140 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2141 (follow_copies_to_constant): Do not follow SSA names registered
2144 2018-03-08 Richard Biener <rguenther@suse.de>
2146 PR tree-optimization/84178
2147 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2149 (version_loop_for_if_conversion): Delay update_ssa call.
2150 (tree_if_conversion): Delay update_ssa until after predicate
2153 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2155 PR tree-optimization/84178
2156 * tree-if-conv.c (release_bb_predicate): Remove the
2157 the assertion that the stmts have NULL use_ops.
2158 Discard the statements, asserting that they haven't
2159 yet been added to a BB.
2161 2018-03-08 Richard Biener <rguenther@suse.de>
2163 PR tree-optimization/84746
2164 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2165 (phi_translate): Pass in destination ANTIC_OUT set.
2166 (phi_translate_1): Likewise. For a simplified result lookup
2167 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2168 (phi_translate_set): Adjust.
2169 (do_pre_regular_insertion): Likewise.
2170 (do_pre_partial_partial_insertion): Likewise.
2172 2018-03-08 Martin Liska <mliska@suse.cz>
2174 PR gcov-profile/84735
2175 * doc/gcov.texi: Document usage of profile files.
2176 * gcov-io.h: Document changes in the format.
2178 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2182 * dwarf2out.c (struct dw_line_info_table): Update comments for
2184 (FORCE_RESET_NEXT_VIEW): New.
2185 (FORCE_RESETTING_VIEW_P): New.
2186 (RESETTING_VIEW_P): Check for -1 too.
2187 (ZERO_VIEW_P): Likewise.
2188 (new_line_info_table): Force-reset next view.
2189 (dwarf2out_begin_function): Likewise.
2190 (dwarf2out_source_line): Simplify zero_view_p initialization.
2191 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2192 view directly. Omit view when omitting .loc at line 0.
2194 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2196 PR tree-optimization/84740
2197 * tree-switch-conversion.c (process_switch): Call build_constructors
2198 only if info.phi_count is non-zero.
2200 PR tree-optimization/84739
2201 * tree-tailcall.c (find_tail_calls): Check call arguments against
2202 DECL_ARGUMENTS (current_function_decl) rather than
2203 DECL_ARGUMENTS (func) when checking for tail recursion.
2205 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2207 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2208 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2209 Volker Reichelt's entry and add entries for people that perform
2210 GCC fuzzy testing and report numerous bugs.
2212 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2215 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2216 readonly data in sdata, if that is disabled.
2217 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2218 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2219 -mreadonly-in-sdata option.
2221 2018-03-07 Martin Sebor <msebor@redhat.com>
2223 PR tree-optimization/84468
2224 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2225 basic block when looking for nul assignment.
2227 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2230 * except.h (output_function_exception_table): Adjust prototype.
2231 * except.c (output_function_exception_table): Remove FNNAME parameter
2232 and add SECTION parameter. Ouput one part of the table at a time.
2233 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2234 the first part of the exception table and emit unwind directives.
2235 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2236 (i386_pe_seh_cold_init): Likewise.
2237 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2238 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2239 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2240 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2241 * config/i386/winnt.c: Include except.h.
2242 (struct seh_frame_state): Add reg_offset, after_prologue and
2243 in_cold_section fields.
2244 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2245 (i386_pe_seh_cold_init): New function.
2246 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2247 to seh->in_cold_section.
2248 (seh_emit_push): Record the offset of the push.
2249 (seh_emit_save): Record the offet of the save.
2250 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2251 Test seh->after_prologue to disregard the epilogue.
2252 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2253 (i386_pe_end_cold_function): New function.
2255 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2258 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2259 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2262 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2264 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2265 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2268 * multiple_target.c: Include tree-inline.h and intl.h.
2269 (expand_target_clones): Diagnose and fail if node->definition and
2270 !tree_versionable_function_p (node->decl).
2272 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2274 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2276 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2277 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2278 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2280 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2283 * combine.c (try_combine): Use reg_or_subregno instead of handling
2284 just paradoxical SUBREGs and REGs.
2286 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2288 * config/arc/arc.c (arc_finalize_pic): Remove function.
2289 (arc_must_save_register): We use single base PIC register, remove
2290 checks to save/restore the PIC register.
2291 (arc_expand_prologue): Likewise.
2292 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2294 (arc_verify_short): Likewise.
2295 (arc_attr_type): Likewise.
2296 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2297 (walk_stores): Likewise.
2298 (arc_address_cost): Make it static.
2299 (arc_verify_short): Likewise.
2300 (branch_dest): Likewise.
2301 (arc_attr_type): Likewise.
2302 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2303 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2304 (arc_final_prescan_insn): Remove inserting the nops due to
2305 hardware hazards. It is done in reorg step.
2306 (insn_length_variant_t): Remove.
2307 (insn_length_parameters_t): Likewise.
2308 (arc_insn_length_parameters): Likewise.
2309 (arc_get_insn_variants): Likewise.
2310 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2312 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2315 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2318 PR tree-optimization/84687
2319 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2321 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2323 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2325 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2326 Rename to ppc_speculation_barrier.
2327 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2328 __builtin_ppc_speculation_barrier.
2330 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2333 * combine.c (combine_simplify_rtx): Don't try to simplify if
2334 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2337 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2339 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2340 to 32 bytes when compiling for POWER9.
2342 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2345 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2346 regparm >= 3 with no arg reg available also for calls with
2347 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2350 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2352 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2354 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2357 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2359 2018-03-05 Richard Biener <rguenther@suse.de>
2361 PR tree-optimization/84486
2362 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2363 When inserting a __builtin_assume_aligned call set the LHS
2364 SSA name alignment info accordingly.
2366 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2368 PR tree-optimization/84114
2369 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2370 Avoid reassociation of FLOAT_MODE addition.
2372 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2374 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2375 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2376 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2377 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2378 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2379 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2380 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2382 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2383 __builtin_ia32_wbinvd): New builtins.
2384 (SPECIAL_ARGS2): New.
2385 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2386 (SPECIAL_ARGS2): New.
2387 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2388 (ix86_valid_target_attribute_inner_p): Ditto.
2389 (ix86_init_mmx_sse_builtins): Add special_args2.
2390 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2391 TARGET_WBNOINVD_P): New.
2392 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2393 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2394 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2395 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2396 * config/i386/pconfigintrin.h: New file.
2397 * config/i386/wbnoinvdintrin.h: Ditto.
2398 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2399 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2401 2018-03-05 Richard Biener <rguenther@suse.de>
2403 PR tree-optimization/84670
2404 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2406 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2407 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2408 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2409 doesn't grow if all successors (recursively) were visited at least
2412 2018-03-05 Richard Biener <rguenther@suse.de>
2414 PR tree-optimization/84650
2415 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2416 if executed in the loop pipeline.
2418 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2420 * doc/configfiles.texi (Configuration Files): Move info about
2421 conditionalizing $target-protos.h to...
2422 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2423 differs from $target-protos.h.
2425 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2426 Chung-Ju Wu <jasonwucj@gmail.com>
2428 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2429 * config/nds32/nds32-multiple.md (setmemsi): Define.
2430 * config/nds32/nds32-memory-manipulation.c
2431 (nds32_gen_dup_4_byte_to_word_value): New.
2432 (emit_setmem_word_loop): New.
2433 (emit_setmem_byte_loop): New.
2434 (nds32_expand_setmem_loop): New.
2435 (nds32_expand_setmem_loop_v3m): New.
2436 (nds32_expand_setmem_unroll): New.
2437 (nds32_expand_setmem): New.
2439 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2440 Chung-Ju Wu <jasonwucj@gmail.com>
2442 * config/nds32/nds32-memory-manipulation.c
2443 (nds32_emit_load_store): New.
2444 (nds32_emit_post_inc_load_store): New.
2445 (nds32_emit_mem_move): New.
2446 (nds32_emit_mem_move_block): New.
2447 (nds32_expand_movmemsi_loop_unknown_size): New.
2448 (nds32_expand_movmemsi_loop_known_size): New.
2449 (nds32_expand_movmemsi_loop): New.
2450 (nds32_expand_movmemsi_unroll): New.
2451 (nds32_expand_movmemqi): Rename ...
2452 (nds32_expand_movmemsi): ... to this.
2453 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2454 (movmemsi): ... to this.
2455 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2456 (nds32_expand_movmemsi): ... to this.
2458 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2459 Monk Chiang <sh.chiang04@gmail.com>
2460 Chung-Ju Wu <jasonwucj@gmail.com>
2462 * config/nds32/nds32-protos.h
2463 (nds32_expand_load_multiple): New arguments.
2464 (nds32_expand_store_multiple): Ditto.
2465 (nds32_valid_multiple_load_store): Rename ...
2466 (nds32_valid_multiple_load_store_p): ... to this.
2467 * config/nds32/nds32-memory-manipulation.c
2468 (nds32_expand_load_multiple): Refine implementation.
2469 (nds32_expand_store_multiple): Ditto.
2470 * config/nds32/nds32-multiple.md
2471 (load_multiple): Update nds32_expand_load_multiple interface.
2472 (store_multiple): Update nds32_expand_store_multiple interface.
2473 * config/nds32/nds32-predicates.c
2474 (nds32_valid_multiple_load_store): Rename ...
2475 (nds32_valid_multiple_load_store_p): ... to this and refine
2477 * config/nds32/predicates.md
2478 (nds32_load_multiple_and_update_address_operation): New predicate.
2479 (nds32_store_multiple_and_update_address_operation): New predicate.
2481 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2482 Chung-Ju Wu <jasonwucj@gmail.com>
2484 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2485 (combo): New attribute.
2486 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2488 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2490 * config/nds32/nds32.opt: Change -mcmodel= default value.
2492 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2493 Monk Chiang <sh.chiang04@gmail.com>
2494 Chung-Ju Wu <jasonwucj@gmail.com>
2496 * config/nds32/constants.md (unspec_element): New enum.
2497 * config/nds32/constraints.md (Umw): New constraint.
2498 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2499 * config/nds32/nds32-intrinsic.md: Likewise.
2500 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2501 (nds32_valid_smw_lwm_base_p): New.
2502 (nds32_output_smw_single_word): New.
2503 (nds32_output_lmw_single_word): New.
2504 (nds32_expand_unaligned_load): New.
2505 (nds32_expand_unaligned_store): New.
2506 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2507 (nds32_output_smw_single_word): Declare.
2508 (nds32_output_lmw_single_word): Declare.
2509 (nds32_expand_unaligned_load): Declare.
2510 (nds32_expand_unaligned_store): Declare.
2511 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2512 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2513 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2514 NDS32_BUILTIN_UASTORE_DW.
2515 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2518 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2519 Kito Cheng <kito.cheng@gmail.com>
2520 Chung-Ju Wu <jasonwucj@gmail.com>
2522 * config/nds32/nds32-intrinsic.c
2523 (nds32_expand_builtin_null_ftype_reg): Delete.
2524 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2525 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2526 (nds32_read_argument): New.
2527 (nds32_legitimize_target): Ditto.
2528 (nds32_legitimize_argument): Ditto.
2529 (nds32_check_constant_argument): Ditto.
2530 (nds32_expand_unop_builtin): Ditto.
2531 (nds32_expand_unopimm_builtin): Ditto.
2532 (nds32_expand_binop_builtin): Ditto.
2533 (nds32_builtin_decl_impl): Ditto.
2534 (builtin_description): Ditto.
2535 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2536 (nds32_init_builtins_impl): Ditto.
2537 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2538 (nds32_builtin_decl): New.
2539 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2540 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2542 2018-03-02 Jeff Law <law@redhat.com>
2544 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2545 (redundant_insn, fill_simple_delay_slots): Likewise.
2546 (fill_slots_from_thread): Likewise.
2547 * resource.c (mark_referenced_resources): Likewise.
2548 (mark_set_resources, find_dead_or_set_registers): Likewise.
2550 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2552 * substring-locations.h (format_warning_va): Formatting fix for
2554 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2556 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2557 * substring-locations.c: Include intl.h.
2558 (format_warning_va): Turned into small wrapper around
2559 format_warning_n_va, renamed to ...
2560 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2561 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2563 (format_warning_at_substring_n): New function.
2564 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2565 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2566 format_warning_at_substring with just a shorter name instead of
2567 const function pointer.
2568 (fmtwarn_n): New function.
2569 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2570 appropriate, get rid of all the fmtstr temporaries, move conditionals
2571 with G_() wrapped string literals directly into fmtwarn arguments,
2572 cast dir.len to (int), formatting fixes.
2574 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2576 * doc/invoke.texi: Remove "Cilk Plus" references.
2578 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2579 Richard Biener <rguenther@suse.de>
2582 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2583 for error or warning attributes if CALL_FROM_THUNK_P is set.
2586 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2589 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2590 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2593 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2594 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2596 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2597 instead of -1U in last predictors element's probability member.
2599 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2602 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2603 arguments if they are comparable.
2605 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2607 PR tree-optimization/84634
2608 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2609 masks and masked_loop_p with a single loop_masks, making sure it's
2610 null for bb vectorization.
2612 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2614 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2615 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2616 loop->force_vectorize to check whether there is no alias.
2618 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2621 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2623 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2625 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2626 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2627 instead of a loop around prev_real_insn.
2628 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2632 * config/i386/i386.c (ix86_print_operand): Use conditional
2633 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2636 2018-03-02 Richard Biener <rguenther@suse.de>
2638 PR tree-optimization/84427
2639 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2640 (bitmap_set_subtract_values): Rewrite to handle multiple
2643 (prune_clobbered_mems): Likewise.
2644 (phi_translate): Take edge instead of pred/phiblock.
2645 (phi_translate_1): Likewise.
2646 (phi_translate_set): Likewise. Insert all translated
2647 exprs for a value into the set, keeping possibly multiple
2648 expressions per value.
2649 (compute_antic_aux): Adjust for phi_translate changes.
2650 When intersecting union the expressions and prune those
2651 not in the final value set, keeping possibly multiple
2652 expressions per value. Do not use value-insertion
2653 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2654 all expressions. Add verification that the value-sets
2655 only shrink during iteration.
2656 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2657 (do_pre_regular_insertion): Likewise.
2658 (do_pre_partial_partial_insertion): Likewise.
2660 2018-03-02 Richard Biener <rguenther@suse.de>
2663 * config/darwin.c (saved_debug_info_level): New static global.
2664 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2665 (darwin_asm_lto_end): Restore debug info generation settings.
2667 2018-03-01 Martin Liska <mliska@suse.cz>
2670 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2673 2018-03-01 Richard Biener <rguenther@suse.de>
2676 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2677 type annotation with LTO when debug was disabled at compile-time.
2679 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2681 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2683 (mips_final_postscan_insn): Likewise.
2685 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2687 PR rtl-optimization/84528
2688 * alias.c (init_alias_target): Add commentary.
2689 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2690 a unique base value if the frame pointer is not eliminated
2691 to the stack pointer.
2693 2018-03-01 Tom de Vries <tom@codesourcery.com>
2695 PR rtl-optimization/83327
2696 * lra-int.h (hard_regs_spilled_into): Declare.
2697 * lra.c (hard_regs_spilled_into): Define.
2698 (init_reg_info): Init hard_regs_spilled_into.
2699 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2700 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2701 (process_bb_lives): Handle hard_regs_spilled_into.
2702 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2703 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2705 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2707 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2708 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2709 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2710 * config/rs6000/aix72.h: New file.
2712 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2714 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2715 instead of warning_at with conditional singular and plural messages
2719 * stor-layout.c (update_alignment_for_field): For
2720 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2721 && !DECL_PACKED (field), do the alignment update, just use
2722 only desired_align instead of MAX (type_align, desired_align)
2724 (place_field): Don't do known_align < desired_align handling
2725 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2726 is non-NULL, instead do it after rli->prev_field handling and
2727 only if not within a bitfield word. For DECL_PACKED (field)
2728 use type_align of BITS_PER_UNIT.
2730 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2732 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2733 superfluous parentheses and trailing spaces.
2735 2018-02-28 Richard Biener <rguenther@suse.de>
2737 PR tree-optimization/84584
2738 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2739 SCoPs with fake exit edge.
2741 2018-02-28 Martin Liska <mliska@suse.cz>
2744 * timevar.c (timer::print): Fix format to properly print 100%
2747 2018-02-28 Richard Biener <rguenther@suse.de>
2750 * genmatch.c (capture_info::walk_match): Do not mark
2751 captured expressions without operands as expr_p given
2752 they act more like predicates and should be subject to
2753 "lost tail" side-effect preserving.
2755 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2757 PR rtl-optimization/81611
2758 * auto-inc-dec.c (attempt_change): Move dead note from
2759 mem_insn if it's the next use of regno
2760 (find_address): Take address use of reg holding
2761 non-incremented value. Add parm to limit search to the named
2763 (merge_in_block): Attempt to use a mem insn that is the next
2764 use of the original regno.
2766 2018-02-27 Martin Sebor <msebor@redhat.com>
2769 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2770 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2772 2018-02-27 Martin Sebor <msebor@redhat.com>
2774 PR translation/84207
2775 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2776 n argument to unsigned HOST_WIDE_INT.
2777 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2778 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2779 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2780 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2782 2018-02-27 Richard Biener <rguenther@suse.de>
2784 PR tree-optimization/84512
2785 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2786 Do not use the estimate returned from record_stmt_cost for
2787 the scalar iteration cost but sum properly using add_stmt_cost.
2789 2018-02-27 Richard Biener <rguenther@suse.de>
2791 PR tree-optimization/84466
2792 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2793 Adjust last change to less strictly validate use operands.
2795 2018-02-27 Martin Liska <mliska@suse.cz>
2797 PR gcov-profile/84548
2798 * gcov.c (process_file): Allow partial overlap and consider it
2799 also as group functions.
2800 (output_lines): Properly calculate range of lines for a group.
2802 2018-02-27 Martin Liska <mliska@suse.cz>
2804 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2805 'ggc' suffixes. Change first column width.
2806 (timer::print): Fix formatting of the column.
2808 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2810 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2811 preserve inline entry blocks for the sake of debug inline
2812 entry point markers alone.
2813 (remove_unused_locals): Suggest in comments a better place to
2814 force the preservation of inline entry blocks that are
2815 otherwise unused, but do not preserve them.
2817 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2819 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2821 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2824 * config/i386/constraints.md (Bs): Replace
2825 ix86_indirect_branch_register with
2826 TARGET_INDIRECT_BRANCH_REGISTER.
2828 * config/i386/i386.md (indirect_jump): Likewise.
2829 (tablejump): Likewise.
2830 (*sibcall_memory): Likewise.
2831 (*sibcall_value_memory): Likewise.
2832 Peepholes of indirect call and jump via memory: Likewise.
2833 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2834 (*sibcall_value_GOT_32): Likewise.
2835 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2836 (GOT_memory_operand): Likewise.
2837 (call_insn_operand): Likewise.
2838 (sibcall_insn_operand): Likewise.
2839 (GOT32_symbol_operand): Likewise.
2840 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2842 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2844 PR rtl-optimization/83496
2845 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2846 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2847 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2848 redundant insn, if any.
2849 (relax_delay_slots): Likewise.
2850 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2852 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2854 PR tree-optimization/83965
2855 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2856 that grouped statements are part of a reduction chain. Return
2857 true if the statement is not marked as a reduction itself but
2859 (vect_recog_dot_prod_pattern): Don't check whether the statement
2860 is part of a group here.
2861 (vect_recog_sad_pattern): Likewise.
2862 (vect_recog_widen_sum_pattern): Likewise.
2864 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2867 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2869 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2872 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2874 (ix86_output_indirect_function_return): New prototype.
2875 (ix86_split_simple_return_pop_internal): Likewise.
2876 * config/i386/i386.c (indirect_return_via_cx): New.
2877 (indirect_return_via_cx_bnd): Likewise.
2878 (indirect_thunk_name): Handle return va CX_REG.
2879 (output_indirect_thunk_function): Create alias for
2880 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2881 (ix86_output_indirect_jmp): Remove the bool argument.
2882 (ix86_output_indirect_function_return): New function.
2883 (ix86_split_simple_return_pop_internal): Likewise.
2884 * config/i386/i386.md (*indirect_jump): Don't pass false
2885 to ix86_output_indirect_jmp.
2886 (*tablejump_1): Likewise.
2887 (simple_return_pop_internal): Change it to define_insn_and_split.
2888 Call ix86_split_simple_return_pop_internal to split it for
2890 (simple_return_indirect_internal): Call
2891 ix86_output_indirect_function_return instead of
2892 ix86_output_indirect_jmp.
2894 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2897 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2898 memset and value initialization afterwards.
2900 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2902 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2904 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2907 * common/config/aarch64/aarch64-common.c
2908 (aarch_option_optimization_table[]): Switch
2909 off fomit-frame-pointer
2911 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2912 Chung-Ju Wu <jasonwucj@gmail.com>
2914 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2916 (store_multiple): Ditto.
2918 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2920 * config.gcc: Add --with-cpu support for nds32 target.
2921 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2922 * config/nds32/nds32.opt: Add -mcpu= option.
2924 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2926 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2927 isel=yes): Warn for these deprecated options.
2929 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2931 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2932 ISA_2_5_MASKS_EMBEDDED.
2934 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2936 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2937 p->max as pointers rather than using iterative_hash_expr.
2939 2018-02-23 Carl Love <cel@us.ibm.com>
2941 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2942 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2944 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2945 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2946 P8V_BUILTIN_VEC_VUNSIGNED2.
2948 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2951 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2952 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2954 * lra-constraints.c (curr_insn_transform): Set up
2955 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2957 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2958 * lra-eliminations.c (spill_pseudos): Ditto.
2959 (process_insn_for_elimination): Ditto.
2960 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2961 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2962 LRA_NON_CLOBBERED_ALT.
2964 2018-02-22 Martin Sebor <msebor@redhat.com>
2966 PR tree-optimization/84480
2967 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2968 to maybe_diag_stxncpy_trunc. Call it.
2969 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2970 from gimple_fold_builtin_strcpy. Print inlining stack.
2971 (handle_builtin_stxncpy): Print inlining stack.
2972 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2974 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2977 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2978 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2979 and -fcheck-pointer-bounds are used together.
2980 (indirect_thunk_prefix): New enum.
2981 (indirect_thunk_need_prefix): New function.
2982 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2983 "_nt" instead of "_bnd" for NOTRACK prefix.
2984 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2985 (output_indirect_thunk_function): Likewise.
2987 (ix86_code_end): Update output_indirect_thunk_function calls.
2988 (ix86_output_indirect_branch_via_reg): Replace
2989 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2990 (ix86_output_indirect_branch_via_push): Likewise.
2991 (ix86_output_function_return): Likewise.
2992 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2993 incompatible with -fcf-protection=branch and
2994 -fcheck-pointer-bounds.
2996 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2999 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3000 Change gcc_assert call to output_operand_lossage.
3002 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3004 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3006 2018-02-22 DJ Delorie <dj@redhat.com>
3007 Sebastian Perta <sebastian.perta@renesas.com>
3008 Oleg Endo <olegendo@gcc.gnu.org>
3010 * config/rx/rx.c (rx_rtx_costs): New function.
3011 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3013 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3015 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3017 2018-02-22 Martin Liska <mliska@suse.cz>
3020 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3021 Add "native" as a possible value.
3023 2018-02-22 Martin Liska <mliska@suse.cz>
3026 * config/i386/i386.c (ix86_option_override_internal):
3027 Add "native" as a possible value for -march and -mtune.
3029 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3032 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3033 to all type variants.
3035 PR tree-optimization/84503
3036 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3037 width as info->bitpos + info->bitsize - start.
3038 (merged_store_group::merge_overlapping): Simplify width computation.
3039 (check_no_overlap): New function.
3040 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3041 start + width and last_order of the group, fail if check_no_overlap
3043 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3044 to group if check_no_overlap fails.
3046 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3048 * config/rs6000/altivec.md: Delete contraint arguments to
3049 define_expand, define_split, and define_peephole2, and in
3050 define_insn_and_split if always unused.
3051 * config/rs6000/darwin.md: Ditto.
3052 * config/rs6000/dfp.md: Ditto.
3053 * config/rs6000/rs6000.md: Ditto.
3054 * config/rs6000/sync.md: Ditto.
3055 * config/rs6000/vector.md: Ditto.
3056 * config/rs6000/vsx.md: Ditto.
3058 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3060 * config/rs6000/altivec.md: Write output control strings as braced
3061 blocks instead of double-quoted strings.
3062 * config/rs6000/darwin.md: Ditto.
3063 * config/rs6000/rs6000.md: Ditto.
3064 * config/rs6000/vector.md: Ditto.
3065 * config/rs6000/vsx.md: Ditto.
3067 2018-02-21 Jason Merrill <jason@redhat.com>
3069 PR c++/84314 - ICE with templates and fastcall attribute.
3070 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3072 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3074 * ipa-cp.c (determine_versionability): Fix comment typos.
3076 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3079 * ipa-cp.c (determine_versionability): Do not version functions caling
3082 2018-02-21 Martin Liska <mliska@suse.cz>
3085 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3086 Add "native" as a possible value.
3087 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3088 the macro when native cpu detection is available.
3090 2018-02-21 Martin Liska <mliska@suse.cz>
3093 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3094 Add "native" as a possible value.
3095 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3096 when native cpu detection is available.
3098 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3099 Martin Sebor <msebor@redhat.com>
3101 PR tree-optimization/84478
3102 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3104 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3105 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3106 support which is conservatively correct, for 2 only stay conservative
3107 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3108 argument to the 2 argument get_range_strlen, adjust 6 arg
3109 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3111 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3112 (gimple_fold_builtin_strlen): Pass true as last argument to
3115 2018-02-20 Martin Sebor <msebor@redhat.com>
3118 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3119 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3120 (builtin_memref::builtin_memref): Factor out parts into
3121 set_base_and_offset and call it.
3123 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3126 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3127 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3128 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3129 search at the associated MODE_INT.
3131 2018-02-20 Jeff Law <law@redhat.com>
3134 PR tree-optimization/81592
3136 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3137 for range data rather than using global data.
3138 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3139 range data rather than using global data.
3140 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3141 pass it to children as needed.
3142 (struct directive::fmtresult): Similarly.
3143 (struct directive::set_width): Similarly.
3144 (struct directive::set_precision): Similarly.
3145 (format_integer, format_directive, parse_directive): Similarly.
3146 (format_none): Accept unnamed vr_values parameter.
3147 (format_percent, format_floating, format_character): Similarly.
3148 (format_string, format_plain): Similarly.
3149 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3150 the EVRP range analyzer for range data rather than using global data.
3151 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3152 gimple-ssa-evrp-analyze.h
3153 (class sprintf_dom_walker): Add after_dom_children member function.
3154 Add evrp_range_analyzer member.
3155 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3156 range analyzer as needed.
3157 (sprintf_dom_walker::after_dom_children): New member function.
3158 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3160 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3161 (evrp_range_analyzer::pop_to_marker): Likewise.
3163 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3165 PR tree-optimization/84419
3166 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3167 with the required type if its current type is compatible but
3170 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3173 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3174 after vectorization.
3176 2018-02-20 Martin Liska <mliska@suse.cz>
3179 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3180 possible values if we don't have a hint.
3182 2018-02-20 Martin Liska <mliska@suse.cz>
3186 * final.c (shorten_branches): Build align_tab array with one
3188 * opts.c (finish_options): Add alignment option limit check.
3189 (MAX_CODE_ALIGN): Likewise.
3190 (MAX_CODE_ALIGN_VALUE): Likewise.
3191 * doc/invoke.texi: Document maximum allowed option value for
3192 all -falign-* options.
3194 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3197 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3198 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3199 * var-tracking.c (emit_note_insn_var_location): Remove all references
3200 to NOTE_INSN_CALL_ARG_LOCATION.
3201 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3202 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3203 Use copy_rtx_if_shared.
3204 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3205 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3206 (dwarf2out_var_location): Remove handling of
3207 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3209 * final.c (final_scan_insn): Remove all references to
3210 NOTE_INSN_CALL_ARG_LOCATION.
3211 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3212 before dumping final insns.
3213 * except.c (emit_note_eh_region_end): Remove all references to
3214 NOTE_INSN_CALL_ARG_LOCATION.
3215 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3216 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3217 * config/arc/arc.c (hwloop_optimize): Likewise.
3218 * config/arm/arm.c (create_fix_barrier): Likewise.
3219 * config/s390/s390.c (s390_chunkify_start): Likewise.
3220 * config/sh/sh.c (find_barrier): Likewise.
3221 * config/i386/i386.c (rest_of_insert_endbranch,
3222 ix86_seh_fixup_eh_fallthru): Likewise.
3223 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3224 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3225 * config/frv/frv.c (frv_function_prologue): Likewise.
3226 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3228 (note_outside_basic_block_p): Remove all references to
3229 NOTE_INSN_CALL_ARG_LOCATION.
3230 * gengtype.c (adjust_field_rtx_def): Likewise.
3231 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3233 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3234 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3237 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3240 PR tree-optimization/84452
3241 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3242 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3245 2018-02-19 Martin Liska <mliska@suse.cz>
3248 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3250 2018-02-19 Martin Liska <mliska@suse.cz>
3251 Richard Sandiford <richard.sandiford@linaro.org>
3253 PR tree-optimization/82491
3254 * gimple-fold.c (get_base_constructor): Make earlier bail out
3257 2018-02-19 Carl Love <cel@us.ibm.com>
3259 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3260 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3262 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3263 P8V_BUILTIN_VEC_NEG.
3265 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3267 * config/rl78/rl78.md (movdf): New define expand.
3269 2018-02-19 Martin Liska <mliska@suse.cz>
3272 * doc/invoke.texi: Fix typo.
3273 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3275 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3277 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3278 handle rs6000_single_float and rs6000_double_float specially for
3281 2018-02-16 Jeff Law <law@redhat.com>
3283 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3286 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3289 * ipa-inline.c (inline_small_functions): Fix a typo.
3291 2018-02-16 Nathan Sidwell <nathan@acm.org>
3293 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3295 2018-02-16 Carl Love <cel@us.ibm.com>
3297 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3298 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3299 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3300 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3301 expansion to P8V_BUILTIN_VEC_FLOAT2.
3303 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3305 PR rtl-optimization/70023
3306 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3307 src_regno into account.
3309 2018-02-16 Carl Love <cel@us.ibm.com>
3311 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3312 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3313 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3314 * config/rs6000/rs6000.c: Remove case statements for
3315 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3316 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3317 and P9V_BUILTIN_VEC_VINSERT4B.
3318 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3319 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3320 * config/rs6000/vsx.md:
3321 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3324 2018-02-16 Carl Love <cel@us.ibm.com>
3326 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3328 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3330 * config/rs6000/rs6000-c.c: Add the definitions for
3331 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3332 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3333 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3334 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3335 definition for insert4b and define insn *insert3b_internal.
3336 * doc/extend.texi: Add documentation for vec_extract4b.
3338 2018-02-16 Nathan Sidwell <nathan@acm.org>
3340 * doc/extend.texi (Backwards Compatibility): Mention friend
3341 injection. Note for-scope is deprecated.
3342 * doc/invoke.texi (-ffriend-injection): Deprecate.
3344 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3346 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3347 that moved to I2, also allow destinations that are a paradoxical
3348 subreg (instead of a normal reg).
3350 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3353 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3356 2018-02-16 Richard Biener <rguenther@suse.de>
3358 PR tree-optimization/84037
3359 PR tree-optimization/84016
3361 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3362 Adjust vec_construct for the fact we need additional higher latency
3363 128bit inserts for AVX256 and AVX512 vector builds.
3364 (ix86_add_stmt_cost): Scale vector construction cost for
3367 2018-02-16 Richard Biener <rguenther@suse.de>
3369 PR tree-optimization/84417
3370 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3371 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3372 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3374 2018-02-16 Martin Liska <mliska@suse.cz>
3377 * internal-fn.def (ASAN_CHECK): Set proper flags.
3378 (ASAN_MARK): Likewise.
3380 2018-02-16 Julia Koval <julia.koval@intel.com>
3382 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3383 from PTA_CANNONLAKE.
3385 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3388 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3389 Use ++iter rather than iter++ for std::list iterators.
3390 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3391 defer deleting them until all nodes in the forest are processed. Do
3392 free even leaf nodes. Change to_process into auto_vec.
3395 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3396 * vec.h (vec_default_construct): Use memset instead of placement new
3397 if BROKEN_VALUE_INITIALIZATION is defined.
3398 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3399 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3402 PR rtl-optimization/83723
3403 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3404 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3405 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3407 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3409 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3411 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3413 PR rtl-optimization/81443
3414 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3415 from inner REGs to paradoxical SUBREGs.
3417 2018-02-16 Richard Biener <rguenther@suse.de>
3419 PR tree-optimization/84399
3420 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3421 For operands we can analyze at their definition make sure we can
3422 analyze them at each use as well.
3424 2018-02-16 Richard Biener <rguenther@suse.de>
3426 PR tree-optimization/84190
3427 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3428 volatile accesses if the decl isn't volatile.
3430 2018-02-15 Jason Merrill <jason@redhat.com>
3432 PR c++/84314 - ICE with templates and fastcall attribute.
3433 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3434 TYPE_CANONICAL on an existing type.
3436 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3438 PR tree-optimization/84383
3439 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3440 dstoff nor call operand_equal_p if dstbase is NULL.
3442 PR tree-optimization/84334
3443 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3444 also a CONSTANT_CLASS_P, punt.
3446 2018-02-14 Jim Wilson <jimw@sifive.com>
3448 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3449 first SMALL_OPERAND check. New local min_second_step. Move assert
3450 to where locals are set. Add TARGET_RVC support.
3451 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3453 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3455 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3457 2018-02-14 Martin Sebor <msebor@redhat.com>
3459 PR tree-optimization/83698
3460 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3461 arrays constrain the offset range to their bounds.
3462 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3463 (builtin_access::overlap): Avoid setting the size of overlap if it's
3465 (maybe_diag_overlap): Also consider arrays when deciding what values
3466 of offsets to include in diagnostics.
3468 2018-02-14 Martin Sebor <msebor@redhat.com>
3471 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3472 that correspond to the kind of a declaration.
3474 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3477 * config/pa/pa.md: Load address of PIC label using the linkage table
3478 if the label is nonlocal.
3480 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3482 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3483 warning message if user requests -maltivec=be.
3484 * doc/invoke.texi: Document deprecation of -maltivec=be.
3486 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3489 * config/rs6000/rs6000-c.c: Update definitions for
3490 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3491 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3493 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3496 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3497 add _get_ssp intrinsics. Remove argument from
3498 __builtin_ia32_rdssp[d|q].
3499 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3500 * config/i386/i386-builtin.def: Remove argument from
3501 __builtin_ia32_rdssp[d|q].
3502 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3503 ix86_expand_special_args_builtin for _rdssp[d|q].
3504 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3505 Clear register before usage.
3506 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3507 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3509 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3511 PR tree-optimization/84357
3512 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3513 operand 1 of an ARRAY_REF too.
3515 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3518 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3519 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3521 (set_of_reg): New struct.
3522 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3523 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3524 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3526 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3527 Split into bitclr, bitset, bitinvert patterns if appropriate.
3528 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3529 use rx_fuse_in_memory_bitop.
3530 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3531 to named insn, correct maximum insn length.
3533 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3536 * machmode.def: Define a complex mode for PARTIAL_INT.
3537 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3539 * doc/rtl.texi: Document CSPImode.
3540 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3542 (msp430_hard_regno_nregs_with_padding): Likewise.
3544 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3547 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3549 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3551 PR rtl-optimization/84169
3552 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3553 we generated a parallel as new i3 and we split that to new i2 and i3
3554 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3555 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3556 those to i2, not i1. Partially rewrite this scan code.
3558 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3561 * stor-layout.c (place_field): For variable length fields, adjust
3562 offset_align afterwards not just based on the field's alignment,
3563 but also on the size.
3566 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3567 of exps and logs in the use_exp2 case.
3569 2018-02-13 Jeff Law <law@redhat.com>
3571 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3574 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3577 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3581 * common.opt (gas-loc-support, gas-locview-support): New.
3582 (ginline-points, ginternal-reset-location-views): New.
3583 * doc/invoke.texi: Document them. Use @itemx where intended.
3584 (gvariable-location-views): Adjust.
3585 * target.def (reset_location_view): New.
3586 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3587 (TARGET_RESET_LOCATION_VIEW): New.
3588 * doc/tm.texi: Rebuilt.
3589 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3590 (dwarf2out_default_as_locview_support): New.
3591 (output_asm_line_debug_info): Use option variables.
3592 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3593 (output_loc_list): Likewise.
3594 (add_high_low_attributes): Check option variables.
3595 Don't output entry view attribute in strict mode.
3596 (gen_inlined_subroutine_die): Check option variables.
3597 (dwarf2out_inline_entry): Likewise.
3598 (init_sections_and_labels): Likewise.
3599 (dwarf2out_early_finish): Likewise.
3600 (maybe_reset_location_view): New, from...
3601 (dwarf2out_var_location): ... here. Call it.
3602 * debug.h (dwarf2out_default_as_loc_support): Declare.
3603 (dwarf2out_default_as_locview_support): Declare.
3604 * hooks.c (hook_int_rtx_insn_0): New.
3605 * hooks.h (hook_int_rtx_insn_0): Declare.
3606 * toplev.c (process_options): Take -gas-loc-support and
3607 -gas-locview-support from dwarf2out. Enable
3608 -gvariable-location-views by default only with locview
3609 assembler support. Enable -ginternal-reset-location-views by
3610 default only if the target defines the corresponding hook.
3611 Enable -ginline-points by default if location views are
3612 enabled; force it disabled if statement frontiers are
3614 * tree-inline.c (expand_call_inline): Check option variables.
3615 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3617 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3619 PR tree-optimization/84321
3620 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3621 handling. Also check whether the anti-range contains any values
3622 that satisfy the mask; switch to a VR_RANGE if not.
3624 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3627 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3629 2018-02-13 Martin Jambor <mjambor@suse.cz>
3632 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3633 of call statements, also set location of a load to a temporary.
3635 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3637 * config/rl78/rl78.c (add_vector_labels): New function.
3638 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3639 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3640 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3641 which checks that no arguments are passed.
3642 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3643 * doc/extend.texi: Documentation for the new attribute.
3645 2018-02-13 Andreas Schwab <schwab@suse.de>
3647 * config/riscv/linux.h (CPP_SPEC): Define.
3649 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3652 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3653 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3654 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3655 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3656 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3657 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3658 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3659 temporarily for AES and PCLMUL builtins.
3661 PR tree-optimization/84339
3662 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3663 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3667 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3668 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3669 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3671 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3673 * omp-simd-clone.h: New file.
3674 * omp-simd-clone.c: Include omp-simd-clone.h.
3675 (expand_simd_clones): No longer static.
3676 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3677 cgraph.h and omp-simd-clone.h.
3678 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3679 (vect_recog_widen_shift_pattern): Formatting fix.
3680 (vect_pattern_recog_1): Don't check optab for calls.
3683 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3684 operands[2] into a REG before using gen_lowpart on it.
3686 2018-02-12 Jeff Law <law@redhat.com>
3689 * config/sh/sh.c (find_barrier): Consider a sibling call
3692 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3693 successfully back substituting a reg.
3695 2018-02-12 Richard Biener <rguenther@suse.de>
3697 PR tree-optimization/84037
3698 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3699 parameter, move visited init to caller.
3700 (vect_slp_analyze_operations): Separate cost from validity
3701 check, initialize visited once for all instances.
3702 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3704 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3706 (vect_model_promotion_demotion_cost): Likewise.
3707 (vectorizable_bswap): Guard cost modeling with !slp_node
3708 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3710 (vectorizable_call): Likewise.
3711 (vectorizable_conversion): Likewise.
3712 (vectorizable_assignment): Likewise.
3713 (vectorizable_shift): Likewise.
3714 (vectorizable_operation): Likewise.
3715 (vectorizable_store): Likewise.
3716 (vectorizable_load): Likewise.
3717 (vectorizable_condition): Likewise.
3718 (vectorizable_comparison): Likewise.
3720 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3723 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3724 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3725 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3727 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3730 * params.def (inline-min-speedup): Increase from 8 to 15.
3731 (max-inline-insns-auto): Decrease from 40 to 30.
3732 * ipa-split.c (consider_split): Add some buffer for function to
3733 be considered inlining candidate.
3734 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3737 2018-02-12 Richard Biener <rguenther@suse.de>
3739 PR tree-optimization/84037
3740 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3741 matched stmts if we cannot swap the non-matched ones.
3743 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3745 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3746 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3747 _mm_maskz_scalef_round_ss): New intrinsics.
3748 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3749 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3750 __builtin_ia32_scalefss_round): Remove.
3751 (__builtin_ia32_scalefsd_mask_round,
3752 __builtin_ia32_scalefss_mask_round): New intrinsics.
3753 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3754 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3755 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3756 "<round_constraint>")): Changed to ...
3757 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3758 "<round_scalar_constraint>")): ... this.
3759 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3760 %0, %1, %2<round_op3>}"): Changed to ...
3761 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3762 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3763 %2<round_scalar_mask_op3>}"): ... this.
3764 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3766 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3768 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3769 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3770 (_mm_maskz_sqrt_round_ss): New intrinsics.
3771 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3772 (__builtin_ia32_sqrtsd_mask_round)
3773 (__builtin_ia32_sqrtss_mask_round): New builtins.
3774 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3775 (__builtin_ia32_sqrtss_round): Remove.
3776 (__builtin_ia32_sqrtsd_mask_round)
3777 (__builtin_ia32_sqrtss_mask_round): New builtins.
3778 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3779 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3780 ((match_operand:VF_128 1 "vector_operand"
3781 "xBm,<round_constraint>")): Changed to ...
3782 ((match_operand:VF_128 1 "vector_operand"
3783 "xBm,<round_scalar_constraint>")): ... this.
3784 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3785 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3786 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3787 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3788 %<iptr>1<round_scalar_mask_op3>}): ... this.
3789 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3790 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3792 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3795 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3796 Cast vec_cmpeq result to correct type.
3797 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3798 Cast vec_cmpgt result to correct type.
3800 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3802 * final.c (final_scan_insn_1): Renamed from...
3803 (final_scan_insn): ... this. New wrapper, to recover
3804 seen from the outermost call in recursive ones.
3805 * config/sparc/sparc.c (output_return): Drop seen from call.
3806 (output_sibcall): Likewise.
3807 * config/visium/visium.c (output_branch): Likewise.
3809 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3811 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3814 2018-02-10 Alan Modra <amodra@gmail.com>
3817 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3818 Specify LR as an input.
3820 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3823 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3824 remove_member_access_dummy_vars): New functions.
3825 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3826 lower_omp_1, execute_lower_omp): Use them.
3828 PR rtl-optimization/84308
3829 * shrink-wrap.c (spread_components): Release todo vector.
3831 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3833 PR rtl-optimization/57193
3834 * ira-color.c (struct allocno_color_data): Add member
3835 conflict_allocno_hard_prefs.
3836 (update_conflict_allocno_hard_prefs): New.
3837 (bucket_allocno_compare_func): Add a preference based on
3838 conflict_allocno_hard_prefs.
3839 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3840 (color_allocnos): Remove a dead code. Initiate
3841 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3843 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3846 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3847 constraint from =wa to wa. Avoid a subreg on the output operand,
3848 instead use a pseudo and subreg it in a move.
3849 (p9_xxbrd_<mode>): Changed to ...
3850 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3851 (p9_xxbrd_v2df): New expander.
3852 (p9_xxbrw_<mode>): Changed to ...
3853 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3854 (p9_xxbrw_v4sf): New expander.
3856 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3858 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3860 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3863 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3864 multiply in 32-bit mode.
3865 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3866 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3869 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3871 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3872 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3873 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3874 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3876 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3879 * dwarf2out.c (is_trivial_indirect_ref): New function.
3880 (dwarf2out_late_global_decl): Do not generate a location
3881 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3882 and that are not defined in the current unit.
3884 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3886 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3887 instead of a libcall for UNORDERED.
3889 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3892 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3893 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3895 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3898 * config/s390/s390.c (s390_set_current_function): Invoke
3899 s390_indirect_branch_settings also if fndecl didn't change.
3901 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3903 * config/rs6000/rs6000.md (blockage): Set length to zero.
3905 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3907 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3909 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3912 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3913 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3917 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3918 PARALLEL incoming that failed vt_get_decl_and_offset check.
3921 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3922 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3924 (get_variable_section): For decls in named .bss* sections pass true as
3925 second argument to bss_initializer_p.
3927 2018-02-09 Marek Polacek <polacek@redhat.com>
3928 Jakub Jelinek <jakub@redhat.com>
3931 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3932 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3933 Sync some changes from cxx_fold_indirect_ref.
3935 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3937 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3939 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3940 (BLOCK_INLINE_ENTRY_LABEL): New.
3941 (dwarf2out_var_location): Disregard inline entry markers.
3942 (inline_entry_data): New struct.
3943 (inline_entry_data_hasher): New hashtable type.
3944 (inline_entry_data_hasher::hash): New.
3945 (inline_entry_data_hasher::equal): New.
3946 (inline_entry_data_table): New variable.
3947 (add_high_low_attributes): Add DW_AT_entry_pc and
3948 DW_AT_GNU_entry_view attributes if a pending entry is found
3949 in inline_entry_data_table. Add old entry_pc attribute only
3950 if debug nonbinding markers are disabled.
3951 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3952 markers are enabled.
3953 (block_within_block_p, dwarf2out_inline_entry): New.
3954 (dwarf2out_finish): Check that no entries remained in
3955 inline_entry_data_table.
3956 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3957 (final_scan_insn, notice_source_line): Likewise.
3958 (rest_of_clean_state): Skip inline entry markers.
3959 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3961 * gimple.c (gimple_build_debug_inline_entry): New.
3962 * gimple.h (enum gimple_debug_subcode): Add
3963 GIMPLE_DEBUG_INLINE_ENTRY.
3964 (gimple_build_debug_inline_entry): Declare.
3965 (gimple_debug_inline_entry_p): New.
3966 (gimple_debug_nonbind_marker_p): Adjust.
3967 * insn-notes.def (INLINE_ENTRY): New.
3968 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3969 inline entry marker notes.
3970 (print_insn): Likewise.
3971 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3972 (INSN_DEBUG_MARKER_KIND): Likewise.
3973 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3974 * tree-inline.c (expand_call_inline): Build and insert
3975 debug_inline_entry stmt.
3976 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3977 inline entry blocks early, if nonbind markers are enabled.
3978 (dump_scope_block): Dump fragment info.
3979 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3980 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3981 (gimple_build_debug_inline_entry): New.
3982 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3983 Enable/disable inline entry points too.
3984 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3985 (DEBUG_INSN): Describe inline entry markers.
3987 * common.opt (gvariable-location-views): New.
3988 (gvariable-location-views=incompat5): New.
3989 * config.in: Rebuilt.
3990 * configure: Rebuilt.
3991 * configure.ac: Test assembler for view support.
3992 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3993 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3994 * dwarf2out.c (var_loc_view): New typedef.
3995 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3996 (dwarf2out_locviews_in_attribute): New.
3997 (dwarf2out_locviews_in_loclist): New.
3998 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3999 (enum dw_line_info_opcode): Add LI_adv_address.
4000 (struct dw_line_info_table): Add view.
4001 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4002 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4003 (zero_view_p): New variable.
4004 (ZERO_VIEW_P): New macro.
4005 (output_asm_line_debug_info): New.
4006 (struct var_loc_node): Add view.
4007 (add_AT_view_list, AT_loc_list): New.
4008 (add_var_loc_to_decl): Add view param. Test it against last.
4009 (new_loc_list): Add view params. Record them.
4010 (AT_loc_list_ptr): Handle loc and view lists.
4011 (view_list_to_loc_list_val_node): New.
4012 (print_dw_val): Handle dw_val_class_view_list.
4013 (size_of_die): Likewise.
4014 (value_format): Likewise.
4015 (loc_list_has_views): New.
4016 (gen_llsym): Set vl_symbol too.
4017 (maybe_gen_llsym, skip_loc_list_entry): New.
4018 (dwarf2out_maybe_output_loclist_view_pair): New.
4019 (output_loc_list): Output view list or entries too.
4020 (output_view_list_offset): New.
4021 (output_die): Handle dw_val_class_view_list.
4022 (output_dwarf_version): New.
4023 (output_compilation_unit_header): Use it.
4024 (output_skeleton_debug_sections): Likewise.
4025 (output_rnglists, output_line_info): Likewise.
4026 (output_pubnames, output_aranges): Update version comments.
4027 (output_one_line_info_table): Output view numbers in asm comments.
4028 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4029 Call maybe_gen_llsym.
4030 (loc_list_from_tree_1): Adjust.
4031 (add_AT_location_description): Create view list attribute if
4032 needed, check it's absent otherwise.
4033 (convert_cfa_to_fb_loc_list): Adjust.
4034 (maybe_emit_file): Call output_asm_line_debug_info for test.
4035 (dwarf2out_var_location): Reset views as needed. Precompute
4036 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4037 attribute. Set view.
4038 (new_line_info_table): Reset next view.
4039 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4040 (dwarf2out_source_line): Likewise. Output view resets and labels to
4041 the assembler, or select appropriate line info opcodes.
4042 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4043 (optimize_string_length): Catch it. Adjust.
4044 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4045 dw_val_class_view_list, and remove it if no longer needed.
4046 (hash_loc_list): Hash view numbers.
4047 (loc_list_hasher::equal): Compare them.
4048 (optimize_location_lists): Check whether a view list symbol is
4049 needed, and whether the locview attribute is present, and
4050 whether they match. Remove the locview attribute if no longer
4052 (index_location_lists): Call skip_loc_list_entry for test.
4053 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4054 Use output_dwarf_version.
4055 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4056 (struct dw_val_node): Add val_view_list.
4057 * final.c (SEEN_NEXT_VIEW): New.
4058 (set_next_view_needed): New.
4059 (clear_next_view_needed): New.
4060 (maybe_output_next_view): New.
4061 (final_start_function): Rename to...
4062 (final_start_function_1): ... this. Take pointer to FIRST,
4063 add SEEN parameter. Emit param bindings in the initial view.
4064 (final_start_function): Reintroduce SEEN-less interface.
4065 (final): Rename to...
4066 (final_1): ... this. Take SEEN parameter. Output final pending
4067 next view at the end.
4068 (final): Reintroduce seen-less interface.
4069 (final_scan_insn): Output pending next view before switching
4070 sections or ending a block. Mark the next view as needed when
4071 outputting variable locations. Notify debug backend of section
4072 changes, and of location view changes.
4073 (rest_of_handle_final): Adjust.
4074 * toplev.c (process_options): Autodetect value for debug variable
4075 location views option. Warn on incompat5 without -gdwarf-5.
4076 * doc/invoke.texi (gvariable-location-views): New.
4077 (gvariable-location-views=incompat5): New.
4078 (gno-variable-location-views): New.
4080 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4082 PR tree-optimization/84136
4083 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4084 that the result of find_edge is non-NULL.
4086 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4089 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4090 storing integer register in SImode. Fix cost of 256 and 512
4091 byte aligned SSE register store.
4093 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4095 * config/i386/i386.c (ix86_multiplication_cost): Fix
4096 multiplication cost for TARGET_AVX512DQ.
4098 2018-02-08 Marek Polacek <polacek@redhat.com>
4100 PR tree-optimization/84238
4101 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4104 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4106 PR tree-optimization/84265
4107 * tree-vect-stmts.c (vectorizable_store): Don't treat
4108 VMAT_CONTIGUOUS accesses as grouped.
4109 (vectorizable_load): Likewise.
4111 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4113 PR tree-optimization/81635
4114 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4115 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4116 (test_round_for_mask): New functions.
4117 (wide_int_cc_tests): Call test_round_for_mask.
4118 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4119 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4120 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4121 range returned by get_range_info.
4123 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4126 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4127 * symtab.c: Include builtins.h
4128 (symtab_node::output_to_lto_symbol_table_p): Move here
4129 from lto-streamer-out.c:output_symbol_p.
4130 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4131 (output_symbol_p): Move all logic to symtab.c
4132 (produce_symtab): Update.
4134 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4136 * config/s390/s390-opts.h (enum indirect_branch): Define.
4137 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4138 (s390_indirect_branch_via_thunk)
4139 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4140 (enum s390_indirect_branch_type): Define.
4141 * config/s390/s390.c (struct s390_frame_layout, struct
4142 machine_function): Remove.
4143 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4144 (indirect_branch_table_label_no, indirect_branch_table_name):
4146 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4147 (enum s390_indirect_branch_option): Define.
4148 (s390_return_addr_from_memory): New function.
4149 (s390_handle_string_attribute): New function.
4150 (s390_attribute_table): Add new attribute handler.
4151 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4152 (s390_indirect_branch_via_thunk): New function.
4153 (s390_indirect_branch_via_inline_thunk): New function.
4154 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4155 sibling call optimization for non z10 compiles.
4156 (s390_emit_call): Force indirect branch target to be a single
4157 register. Add r1 clobber for non-z10 compiles.
4158 (s390_emit_epilogue): Emit return jump via return_use expander.
4159 (s390_reorg): Handle JUMP_INSNs as execute targets.
4160 (s390_option_override_internal): Perform validity checks for the
4161 new command line options.
4162 (s390_indirect_branch_attrvalue): New function.
4163 (s390_indirect_branch_settings): New function.
4164 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4165 (s390_output_indirect_thunk_function): New function.
4166 (s390_code_end): Implement target hook.
4167 (s390_case_values_threshold): Implement target hook.
4168 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4170 * config/s390/s390.h (struct s390_frame_layout)
4171 (struct machine_function): Move here from s390.c.
4172 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4173 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4174 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4175 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4176 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4177 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4178 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4179 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4180 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4181 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4182 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4183 (mnemonic attribute): Add values which aren't recognized
4185 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4186 pattern for branch conversion. Fix mnemonic attribute.
4187 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4188 indirect branch via thunk if requested.
4189 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4190 ("*indirect_jump"): Disable for branch conversion using out of
4192 ("indirect_jump_via_thunk<mode>_z10")
4193 ("indirect_jump_via_thunk<mode>")
4194 ("indirect_jump_via_inlinethunk<mode>_z10")
4195 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4196 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4197 ("casesi_jump_via_inlinethunk<mode>_z10")
4198 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4199 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4200 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4201 ("*indirect2_jump"): Disable for branch conversion.
4202 ("casesi_jump"): Turn into expander and expand patterns for branch
4204 ("return_use"): New expander.
4205 ("*return"): Emit return via thunk and rename it to ...
4206 ("*return<mode>"): ... this one.
4207 * config/s390/s390.opt: Add new options and and enum for the
4210 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4212 * lra-constraints.c (match_reload): Unconditionally use
4213 gen_lowpart_SUBREG, rather than selecting between that
4214 and equivalent gen_rtx_SUBREG code.
4216 2018-02-08 Richard Biener <rguenther@suse.de>
4218 PR tree-optimization/84233
4219 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4220 changed flag instead of boguously re-using phi_inserted.
4222 2018-02-08 Martin Jambor <mjambor@suse.cz>
4224 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4225 static local variables.
4227 2018-02-08 Richard Biener <rguenther@suse.de>
4229 PR tree-optimization/84278
4230 * tree-vect-stmts.c (vectorizable_store): When looking for
4231 smaller vector types to perform grouped strided loads/stores
4232 make sure the mode is supported by the target.
4233 (vectorizable_load): Likewise.
4235 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4237 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4238 Increase LDP/STP opportunities by adding adjacent callee-saves.
4240 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4242 PR rtl-optimization/84068
4243 PR rtl-optimization/83459
4244 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4246 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4248 PR tree-optimization/84224
4249 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4250 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4253 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4256 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4257 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4258 count, remove one USE.
4260 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4262 * doc/install.texi (Configuration): Document the
4263 --with-long-double-format={ibm,ieee} PowerPC configuration
4267 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4268 Convert from define_expand to be define_insn_and_split. Rework
4269 float/double/_Float128 conversions to QI/HI/SImode to work with
4270 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4271 conversions to QI/HImode types did a store and then a load to
4272 truncate the value. For conversions to VSX registers, don't split
4273 the insn, instead emit the code directly. Use the code iterator
4274 any_fix to combine signed and unsigned conversions.
4275 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4276 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4277 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4278 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4279 (fix_<mode>di2_hw): Likewise.
4280 (fixuns_<mode>di2_hw): Likewise.
4281 (fix_<mode>si2_hw): Likewise.
4282 (fixuns_<mode>si2_hw): Likewise.
4283 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4284 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4285 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4286 fix<uns>_trunc<SFDF:mode>si2_p8.
4287 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4289 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4290 (fix<uns>_<mode>_mem): Likewise.
4291 (fctiw<u>z_<mode>_mem): Likewise.
4292 (fix<uns>_<mode>_mem): Likewise.
4293 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4294 the register allocator from doing a direct move to the GPRs to do
4295 a store, and instead use the ISA 3.0 store byte/half-word from
4296 vector register instruction. For IEEE 128-bit floating point,
4297 also optimize stores of 32-bit ints.
4298 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4300 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4302 * genextract.c (push_pathstr_operand): New function to support
4304 (walk_rtx): Call push_pathstr_operand.
4305 (print_path): Support [a-zA-Z].
4307 2018-02-07 Richard Biener <rguenther@suse.de>
4309 PR tree-optimization/84037
4310 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4311 (cse_and_gimplify_to_preheader): Declare.
4312 (vect_get_place_in_interleaving_chain): Likewise.
4313 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4315 (_loop_vec_info::~_loop_vec_info): Delete it.
4316 (cse_and_gimplify_to_preheader): New function.
4317 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4318 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4319 (vectorizable_load): Likewise. For grouped stores always base
4320 the IV on the first element.
4321 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4322 condition before gimplifying.
4324 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4326 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4327 *DIV_EXPR and *MOD_EXPR.
4329 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4332 * config/i386/i386.c (ix86_option_override_internal): Mask out
4333 the CF_SET bit when checking -fcf-protection.
4335 2018-02-07 Tom de Vries <tom@codesourcery.com>
4338 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4341 2018-02-07 Richard Biener <rguenther@suse.de>
4343 PR tree-optimization/84204
4344 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4347 PR tree-optimization/84205
4348 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4349 special-case isl_ast_op_zdiv_r.
4351 PR tree-optimization/84223
4352 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4353 Only add conditions from within the region.
4354 (gather_bbs::after_dom_children): Adjust.
4356 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4359 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4360 * config/avr/avr.md: Only post-reload split REG-REG moves if
4361 either register is GENERAL_REG_P.
4363 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4365 PR tree-optimization/84235
4366 * tree-ssa-scopedtables.c
4367 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4368 if the subtraction is performed in floating point type where NaNs are
4369 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4370 build 1. Formatting fix.
4372 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4375 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4376 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4377 and skip it regardless of bb boundaries. Use CALL_P macro,
4378 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4379 unnecessarily, formatting fix.
4381 2018-02-06 Michael Collison <michael.collison@arm.com>
4383 * config/arm/thumb2.md:
4384 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4385 (*thumb_mov_notscc): Ditto.
4387 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4390 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4391 unsigned_fix, not "s".
4393 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4395 * configure.ac (gcc_fn_eh_frame_ro): New function.
4396 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4397 correct .eh_frame permissions.
4398 * configure: Regenerate.
4400 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4402 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4405 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4407 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4408 Display warning message for -mno-speculate-indirect-jumps.
4410 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4412 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4414 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4416 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4418 PR tree-optimization/84225
4419 * tree-eh.c (find_trapping_overflow): Only call
4420 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4422 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4425 * config/i386/i386.c: Reimplement the check of possible options
4426 -mibt/-mshstk conbination. Change error messages.
4427 * doc/invoke.texi: Fix a typo: remove extra '='.
4429 2018-02-06 Marek Polacek <polacek@redhat.com>
4431 PR tree-optimization/84228
4432 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4434 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4437 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4438 emitted arch directives.
4439 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4440 __ARM_FEATURE_COPROC before changing architectures.
4442 2018-02-06 Richard Biener <rguenther@suse.de>
4444 * config/i386/i386.c (print_reg): Fix typo.
4445 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4447 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4449 * configure: Regenerate.
4451 2018-02-05 Martin Sebor <msebor@redhat.com>
4453 PR tree-optimization/83369
4454 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4457 2018-02-05 Martin Liska <mliska@suse.cz>
4459 * doc/invoke.texi: Cherry-pick upstream r323995.
4461 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4463 * ira.c (ira_init_register_move_cost): Adjust comment.
4465 2018-02-05 Martin Liska <mliska@suse.cz>
4467 PR gcov-profile/84137
4468 * doc/gcov.texi: Fix typo in documentation.
4470 2018-02-05 Martin Liska <mliska@suse.cz>
4472 PR gcov-profile/83879
4473 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4474 using dlopen functionality.
4476 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4478 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4479 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4480 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4481 _mm_maskz_range_round_ss): New intrinsics.
4482 (__builtin_ia32_rangesd128_round)
4483 (__builtin_ia32_rangess128_round): Remove.
4484 (__builtin_ia32_rangesd128_mask_round,
4485 __builtin_ia32_rangess128_mask_round): New builtins.
4486 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4487 __builtin_ia32_rangess128_round): Remove.
4488 (__builtin_ia32_rangesd128_mask_round,
4489 __builtin_ia32_rangess128_mask_round): New builtins.
4490 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4491 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4492 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4493 "<round_saeonly_constraint>")): Changed to ...
4494 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4495 "<round_saeonly_scalar_constraint>")): ... this.
4496 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4497 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4498 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4499 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4500 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4502 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4504 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4506 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4507 Remove all values except native, 8540 and 8548.
4509 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4511 * config/i386/i386.c (ix86_output_function_return): Pass
4512 INVALID_REGNUM, instead of -1, as invalid register number to
4513 indirect_thunk_name and output_indirect_thunk.
4515 2018-02-02 Julia Koval <julia.koval@intel.com>
4517 * config.gcc: Add -march=icelake.
4518 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4519 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4520 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4521 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4522 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4523 (processor_target_table): Add icelake.
4524 (ix86_option_override_internal): Handle new PTAs.
4525 (get_builtin_code_for_version): Handle icelake.
4526 (M_INTEL_COREI7_ICELAKE): New.
4527 (fold_builtin_cpu): Handle icelake.
4528 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4529 * doc/invoke.texi: Add -march=icelake.
4531 2018-02-02 Julia Koval <julia.koval@intel.com>
4533 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4534 to wide_int_bitmask.
4535 * wide-int-bitmask.h: New.
4537 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4540 * config/i386/i386.md: Replace Pmode with word_mode in
4541 builtin_setjmp_setup and builtin_longjmp to support x32.
4543 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4547 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4549 (rs6000_supported_cpu_names): New static variable.
4550 (linux_cpu_translation_table): Likewise.
4551 (elf_platform) <cpu>: Define new static variable and use it.
4552 Translate kernel AT_PLATFORM name to canonical name if needed.
4553 Error if platform name is unknown.
4555 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4558 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4560 2018-02-01 Jeff Law <law@redhat.com>
4563 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4564 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4565 the scratch if RELEASE_VIA_POP is false.
4566 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4567 If we have to save a temporary register, decrement SIZE appropriately.
4568 Pass new arguments to release_scratch_register_on_entry.
4569 (ix86_adjust_stack_and_probe): Likewise.
4570 (ix86_emit_probe_stack_range): Pass new arguments to
4571 release_scratch_register_on_entry.
4573 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4575 PR rtl-optimization/84157
4576 * combine.c (change_zero_ext): Use REG_P predicate in
4577 front of HARD_REGISTER_P predicate.
4579 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4581 * config/avr/avr.c (avr_option_override): Move disabling of
4582 -fdelete-null-pointer-checks to...
4583 * common/config/avr/avr-common.c (avr_option_optimization_table):
4586 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4588 PR tree-optimization/81635
4589 * tree-data-ref.c (split_constant_offset_1): For types that
4590 wrap on overflow, try to use range info to prove that wrapping
4593 2018-02-01 Renlin Li <renlin.li@arm.com>
4596 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4598 (aarch64_register_move_cost): Likewise.
4599 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4601 (REG_CLASS_NAMES): Likewise.
4602 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4603 TAILCALL_ADDR_REGS. Remove IP registers.
4604 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4606 2018-02-01 Richard Biener <rguenther@suse.de>
4608 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4609 for specifying RPO order and allow NULL for that.
4610 * domwalk.c (dom_walker::dom_walker): Likewise.
4611 (dom_walker::walk): Handle NULL RPO order.
4612 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4614 (rewrite_update_dom_walker): Likewise.
4615 (mark_def_dom_walker): Likewise.
4617 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4619 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4620 (aarch64_maybe_expand_sve_subreg_move): Declare.
4621 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4622 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4624 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4625 that are semantically a reverse operation.
4626 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4627 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4628 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4630 (aarch64_can_change_mode_class): For big-endian, forbid changes
4631 between two SVE modes if they have different element sizes.
4633 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4635 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4636 the TImode handling for big-endian targets.
4638 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4640 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4641 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4643 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4644 Remove BSWAP handing for big-endian targets and use the form of
4645 LD1RQ appropariate for the mode.
4647 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4649 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4650 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4653 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4656 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4657 check for operands that need to go through aarch64_sve_reload_be.
4659 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4661 PR tree-optimization/81661
4662 PR tree-optimization/84117
4663 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4664 * tree-eh.c: Include gimplify.h.
4665 (find_trapping_overflow, replace_trapping_overflow,
4666 rewrite_to_non_trapping_overflow): New functions.
4667 * tree-vect-loop.c: Include tree-eh.h.
4668 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4669 * tree-data-ref.c: Include tree-eh.h.
4670 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4672 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4674 PR rtl-optimization/84123
4675 * combine.c (change_zero_ext): Check if hard register satisfies
4676 can_change_dest_mode before calling gen_lowpart_SUBREG.
4678 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4681 * ira.c (ira_init_register_move_cost): Remove assert.
4683 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4685 PR rtl-optimization/84071
4686 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4687 * doc/tm.texi: Regenerate.
4689 2018-01-31 Richard Biener <rguenther@suse.de>
4691 PR tree-optimization/84132
4692 * tree-data-ref.c (analyze_miv_subscript): Properly
4693 check whether evolution_function_is_affine_multivariate_p
4694 before calling gcd_of_steps_may_divide_p.
4696 2018-01-31 Julia Koval <julia.koval@intel.com>
4699 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4700 * config/i386/i386.md (rdpid_rex64) New.
4701 (rdpid): Make 32bit only.
4703 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4706 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4707 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4709 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4712 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4714 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4716 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4718 PR rtl-optimization/84071
4719 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4720 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4722 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4724 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4725 (arc_attribute_table): Add 'aux' attribute.
4726 (arc_in_small_data_p): Consider aux like variables.
4727 (arc_is_aux_reg_p): New function.
4728 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4729 (arc_get_aux_arg): New function.
4730 (prepare_move_operands): Handle aux-register access.
4731 (arc_handle_aux_attribute): New function.
4732 * doc/extend.texi (ARC Variable attributes): Add subsection.
4734 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4736 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4737 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4738 (arc_attribute_table): Add 'uncached' attribute.
4739 (arc_print_operand): Print '.di' flag for uncached memory
4741 (arc_in_small_data_p): Do not consider for small data the uncached
4743 (arc_is_uncached_mem_p): New function.
4744 * config/arc/predicates.md (compact_store_memory_operand): Check
4745 for uncached memory accesses.
4746 (nonvol_nonimm_operand): Likewise.
4747 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4749 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4752 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4753 falign-loops=): Add Optimization flag.
4755 2018-01-30 Jeff Law <law@redhat.com>
4758 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4759 INT_REGISTERS_SAVED. Check it prior to calling
4760 get_scratch_register_on_entry.
4761 (ix86_adjust_stack_and_probe): Similarly.
4762 (ix86_emit_probe_stack_range): Similarly.
4763 (ix86_expand_prologue): Corresponding changes.
4765 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4768 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4769 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4771 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4774 * lra-constraints.c (curr_insn_transform): Process AND in the
4777 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4779 PR rtl-optimization/83986
4780 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4781 dependence against last_pending_memory_flush in addition to
4784 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4786 PR tree-optimization/81611
4787 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4790 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4793 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4796 2018-01-30 Richard Biener <rguenther@suse.de>
4797 Jakub Jelinek <jakub@redhat.com>
4799 PR tree-optimization/84111
4800 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4801 inner loops added during recursion, as they don't have up-to-date
4804 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4807 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4808 (can_inline_edge_by_limits_p): ... here.
4809 (can_early_inline_edge_p, check_callers,
4810 update_caller_keys, update_callee_keys, recursive_inlining,
4811 add_new_edges_to_heap, speculation_useful_p,
4812 inline_small_functions,
4813 inline_small_functions, flatten_function,
4814 inline_to_all_callers_1): Update.
4816 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4818 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4821 2018-01-30 Richard Biener <rguenther@suse.de>
4823 PR tree-optimization/83008
4824 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4825 invariant and constant vector uses in stmts when they need
4828 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4831 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4832 * configure: Regenerate.
4834 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4836 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4838 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4839 Use gen_rtx_REG rather than gen_lowpart.
4841 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4843 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4844 rather than 0 when creating partial subregs.
4846 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4848 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4851 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4854 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4855 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4856 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4857 flags. This restores the settings used before the 2017-07-24.
4858 Turning off pre increment/decrement/modify allows IVOPTS to
4859 optimize DF/SF loops where the index is an int.
4861 2018-01-29 Richard Biener <rguenther@suse.de>
4862 Kelvin Nilsen <kelvin@gcc.gnu.org>
4865 * tree-vect-stmts.c (vectorizable_call): Don't call
4866 targetm.vectorize_builtin_md_vectorized_function if callee is
4869 2018-01-22 Carl Love <cel@us.ibm.com>
4871 * doc/extend.tex: Fix typo in second arg in
4872 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4874 2018-01-29 Richard Biener <rguenther@suse.de>
4876 PR tree-optimization/84086
4877 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4878 (flush_ssaname_freelist): When SSA names were released reset
4879 the SCEV hash table.
4881 2018-01-29 Richard Biener <rguenther@suse.de>
4883 PR tree-optimization/84057
4884 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4885 removed paths when removing edges.
4887 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4889 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4890 -mfunction-return=@var{choice}.
4892 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4895 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4897 (layout::print_source_line): Likewise.
4898 (test_get_line_width_without_trailing_whitespace): Add test cases.
4900 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4903 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4906 2018-01-26 Jim Wilson <jimw@sifive.com>
4908 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4910 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4913 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4915 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4916 and CMP + SUB-immediate -> SUBS.
4918 2018-01-26 Martin Sebor <msebor@redhat.com>
4920 PR tree-optimization/83896
4921 * tree-ssa-strlen.c (get_string_len): Rename...
4922 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4923 Avoid assuming length is constant.
4924 (handle_char_store): Use HOST_WIDE_INT for string length.
4926 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4929 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4930 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4932 2018-01-26 Richard Biener <rguenther@suse.de>
4934 PR rtl-optimization/84003
4935 * dse.c (record_store): Only record redundant stores when
4936 the earlier store aliases at least all accesses the later one does.
4938 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4940 PR rtl-optimization/83985
4941 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4942 REG_CFA_RESTORE insns.
4943 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4944 REG_CFA_RESTORE insns here.
4947 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4948 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4950 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4952 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4953 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4954 (arc_init): Likewise.
4955 (arc_override_options): Likewise.
4956 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4958 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4960 * config/arc/arc.h (TARGET_DBNZ): Define.
4961 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4962 properly set the tune attribute.
4963 (dbnz): Use TARGET_DBNZ guard.
4964 * config/arc/arc.opt (mtune): Add core3 option.
4966 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4968 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4969 recognize new pic like addresses.
4970 (arc_delegitimize_address): Clean up.
4972 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4974 * config/arc/arc-arches.def: Option mrf16 valid for all
4976 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4977 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4978 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4979 * config/arc/arc-tables.opt: Regenerate.
4980 * config/arc/arc.c (arc_conditional_register_usage): Handle
4981 reduced register file case.
4982 (arc_file_start): Set must have build attributes.
4983 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4985 * config/arc/arc.opt (mrf16): Add new option.
4986 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4987 * config/arc/genmultilib.awk: Handle new mrf16 option.
4988 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4989 * config/arc/t-multilib: Regenerate.
4990 * doc/invoke.texi (ARC Options): Document mrf16 option.
4992 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4994 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4995 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4996 (arc_attribute_table): Add 'secure_call' attribute.
4997 (arc_print_operand): Print secure call operand.
4998 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5000 (arc_is_secure_call_p): New function. * config/arc/arc.md
5001 (call_i): Add support for sjli instruction.
5002 (call_value_i): Likewise.
5003 * config/arc/constraints.md (Csc): New constraint.
5005 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5006 John Eric Martin <John.Martin@emmicro-us.com>
5008 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5009 * config/arc/arc.c (_arc_jli_section): New struct.
5010 (arc_jli_section): New type.
5011 (rc_jli_sections): New static variable.
5012 (arc_handle_jli_attribute): New function.
5013 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5014 (arc_file_end): New function.
5015 (TARGET_ASM_FILE_END): Define.
5016 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5017 (arc_add_jli_section): New function.
5018 (jli_call_scan): Likewise.
5019 (arc_reorg): Call jli_call_scan.
5020 (arc_output_addsi): Remove 'S' from printing asm operand.
5021 (arc_is_jli_call_p): New function.
5022 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5024 (movhi_insn): Likewise.
5025 (movsi_insn): Likewise.
5026 (movsi_set_cc_insn): Likewise.
5027 (loadqi_update): Likewise.
5028 (load_zeroextendqisi_update): Likewise.
5029 (load_signextendqisi_update): Likewise.
5030 (loadhi_update): Likewise.
5031 (load_zeroextendhisi_update): Likewise.
5032 (load_signextendhisi_update): Likewise.
5033 (loadsi_update): Likewise.
5034 (loadsf_update): Likewise.
5035 (movsicc_insn): Likewise.
5036 (bset_insn): Likewise.
5037 (bxor_insn): Likewise.
5038 (bclr_insn): Likewise.
5039 (bmsk_insn): Likewise.
5040 (bicsi3_insn): Likewise.
5041 (cmpsi_cc_c_insn): Likewise.
5042 (movsi_ne): Likewise.
5043 (movsi_cond_exec): Likewise.
5044 (clrsbsi2): Likewise.
5055 (call_i): Remove 'S' asm letter, add jli instruction.
5056 (call_value_i): Likewise.
5057 * config/arc/arc.op (mjli-always): New option.
5058 * config/arc/constraints.md (Cji): New constraint.
5059 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5061 (subsf3_fpx): Likewise.
5062 (mulsf3_fpx): Likewise.
5063 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5065 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5067 * doc/invoke.texi (ARC): Document mjli-always option.
5069 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5071 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5072 avoid addition with 0 and use incw and decw where possible.
5074 2018-01-26 Richard Biener <rguenther@suse.de>
5076 PR tree-optimization/81082
5077 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5078 association if it requires casting to unsigned.
5079 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5080 from fold_plusminus_mult_expr to catch important cases late when
5081 range info is available.
5083 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5085 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5086 * configure.ac (hidden_linkonce): New test.
5087 * configure: Regenerate.
5088 * config.in: Regenerate.
5090 2018-01-26 Julia Koval <julia.koval@intel.com>
5092 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5093 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5094 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5095 _mm_mask_bitshuffle_epi64_mask): Fix type.
5096 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5097 USI_FTYPE_V4DI_V4DI_USI): Remove.
5098 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5099 __builtin_ia32_vpshufbitqmb256_mask,
5100 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5101 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5102 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5104 2018-01-26 Alan Modra <amodra@gmail.com>
5107 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5108 UNSPEC_VBPERMQ. Sort other unspecs.
5110 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5112 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5114 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5117 * predict.c (drop_profile): Do not push/pop cfun; update also
5119 (handle_missing_profiles): Fix logic looking for zero profiles.
5121 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5124 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5125 on functions with #pragma omp declare simd or functions with simd
5127 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5128 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5129 Remove trailing \n from warning_at calls.
5131 2018-01-25 Tom de Vries <tom@codesourcery.com>
5134 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5135 for neutered workers.
5137 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5140 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5141 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5143 2018-01-24 Jeff Law <law@redhat.com>
5146 * i386.c (get_probe_interval): Move to earlier point.
5147 (ix86_compute_frame_layout): If -fstack-clash-protection and
5148 the frame is larger than the probe interval, then use pushes
5149 to save registers rather than reg->mem moves.
5150 (ix86_expand_prologue): Remove conditional for int_registers_saved
5153 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5156 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5157 min/max for never referenced object.
5159 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5162 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5164 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5165 attributes from DECL_ATTRIBUTES (decl) without affecting
5166 DECL_ATTRIBUTES (current_function_decl).
5167 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5168 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5170 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5172 PR tree-optimization/83979
5173 * fold-const.c (fold_comparison): Use constant_boolean_node
5174 instead of boolean_{true,false}_node.
5176 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5178 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5181 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5183 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5184 Simplify the clause that sets the length attribute.
5185 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5186 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5187 clause that sets the length attribute.
5188 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5190 2018-01-24 Tom de Vries <tom@codesourcery.com>
5193 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5194 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5195 Add strict parameter.
5196 (prevent_branch_around_nothing): Insert dummy insn between branch to
5197 label and label with no ptx insn inbetween.
5198 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5200 2018-01-24 Tom de Vries <tom@codesourcery.com>
5203 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5204 for neutered threads in warp.
5205 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5207 2018-01-24 Richard Biener <rguenther@suse.de>
5209 PR tree-optimization/83176
5210 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5213 2018-01-24 Richard Biener <rguenther@suse.de>
5215 PR tree-optimization/82819
5216 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5217 code generating pluses that are no-ops in the target precision.
5219 2018-01-24 Richard Biener <rguenther@suse.de>
5222 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5224 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5226 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5227 to merge probabilities.
5228 * predict.c (probably_never_executed): Also mark as cold functions
5229 with global 0 profile and guessed local profile.
5230 * profile-count.c (profile_probability::combine_with_count): New
5232 * profile-count.h (profile_probability::operator*,
5233 profile_probability::operator*=, profile_probability::operator/,
5234 profile_probability::operator/=): Reduce precision to adjusted
5235 and set value to guessed on contradictory divisions.
5236 (profile_probability::combine_with_freq): Remove.
5237 (profile_probability::combine_wiht_count): Declare.
5238 (profile_count::force_nonzero):: Set to adjusted.
5239 (profile_count::probability_in):: Set quality to adjusted.
5240 * tree-ssa-tail-merge.c (replace_block_by): Use
5243 2018-01-23 Andrew Waterman <andrew@sifive.com>
5244 Jim Wilson <jimw@sifive.com>
5246 * config/riscv/riscv.c (riscv_stack_boundary): New.
5247 (riscv_option_override): Set riscv_stack_boundary. Handle
5248 riscv_preferred_stack_boundary_arg.
5249 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5250 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5251 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5252 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5253 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5254 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5256 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5259 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5260 of struct ix86_frame.
5261 (ix86_expand_epilogue): Likewise. Add a local variable for
5262 the reg_save_offset field in struct ix86_frame.
5264 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5266 PR tree-optimization/82604
5267 * tree-loop-distribution.c (enum partition_kind): New enum item
5268 PKIND_PARTIAL_MEMSET.
5269 (partition_builtin_p): Support above new enum item.
5270 (generate_code_for_partition): Ditto.
5271 (compute_access_range): Differentiate cases that equality can be
5272 proven at all loops, the innermost loops or no loops.
5273 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5274 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5275 (finalize_partitions, distribute_loop): Don't fuse partition of
5276 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5277 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5280 2018-01-23 Martin Liska <mliska@suse.cz>
5282 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5283 order to ignore the predictor.
5284 (PRED_POLYMORPHIC_CALL): Likewise.
5285 (PRED_RECURSIVE_CALL): Likewise.
5287 2018-01-23 Martin Liska <mliska@suse.cz>
5289 * tree-profile.c (tree_profiling): Print function header to
5290 aware reader which function we are working on.
5291 * value-prof.c (gimple_find_values_to_profile): Do not print
5292 not interesting value histograms.
5294 2018-01-23 Martin Liska <mliska@suse.cz>
5296 * profile-count.h (enum profile_quality): Add
5297 profile_uninitialized as the first value. Do not number values
5298 as they are zero based.
5299 (profile_count::verify): Update sanity check.
5300 (profile_probability::verify): Likewise.
5302 2018-01-23 Nathan Sidwell <nathan@acm.org>
5304 * doc/invoke.texi (ffor-scope): Deprecate.
5306 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5308 PR tree-optimization/83510
5309 * domwalk.c (set_all_edges_as_executable): New function.
5310 (dom_walker::dom_walker): Convert bool param
5311 "skip_unreachable_blocks" to enum reachability. Move setup of
5312 edge flags to set_all_edges_as_executable and only do it when
5313 reachability is REACHABLE_BLOCKS.
5314 * domwalk.h (enum dom_walker::reachability): New enum.
5315 (dom_walker::dom_walker): Convert bool param
5316 "skip_unreachable_blocks" to enum reachability.
5317 (set_all_edges_as_executable): New decl.
5318 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5319 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5321 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5322 but converting true to REACHABLE_BLOCKS.
5323 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5325 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5326 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5327 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5329 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5330 if check_all_array_refs will be called.
5332 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5334 * tree.c (selftest::test_location_wrappers): Add more test
5337 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5339 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5340 (selftest::test_bit_in_range): Likewise.
5342 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5345 * doc/sourcebuild.texi (vect_float): Say that the selector
5346 only describes the situation when -funsafe-math-optimizations is on.
5347 (vect_float_strict): Document.
5349 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5351 PR tree-optimization/83965
5352 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5353 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5354 instead of checking only for a reduction.
5355 (vect_recog_widen_sum_pattern): Likewise.
5357 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5359 * predict.c (probably_never_executed): Only use precise profile info.
5360 (compute_function_frequency): Skip after inlining hack since we now
5361 have quality checking.
5363 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5365 * profile-count.h (profile_probability::very_unlikely,
5366 profile_probability::unlikely, profile_probability::even): Set
5367 precision to guessed.
5369 2018-01-23 Richard Biener <rguenther@suse.de>
5371 PR tree-optimization/83963
5372 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5373 Properly terminate dominator walk when crossing the exit edge not
5374 when visiting its source block.
5376 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5379 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5380 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5382 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5384 PR tree-optimization/83957
5385 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5386 semicolon after for body surrounded by braces.
5388 PR tree-optimization/83081
5389 * profile-count.h (profile_probability::split): New method.
5390 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5391 Use profile_probability::split.
5392 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5393 when splitting a single conditional jump into 2.
5395 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5397 PR tree-optimization/69452
5398 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5401 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5403 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5404 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5405 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5407 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5409 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5410 * config/rl78/rl78.md (movdi): New define_expand.
5411 * config/rl78/rl78.c (rl78_split_movdi): New function.
5413 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5416 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5418 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5419 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5420 128-bit to produce an UNSPEC move to get the double word with the
5421 signbit and then a shift directly to do signbit.
5422 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5423 implementation with a new version that just does either a direct
5424 move or a regular move. Move memory interface to separate insns.
5425 Move insns so they are next to the expander.
5426 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5427 with signbit move. Split big and little endian case.
5428 (signbit<mode>2_dm_mem_le): Likewise.
5429 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5430 (signbit<mode>2_dm2): Likewise.
5432 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5434 * config/rl78/rl78.md (anddi3): New define_expand.
5436 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5438 * config/rl78/rl78.md (umindi3): New define_expand.
5440 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5442 * config/rl78/rl78.md (smindi3): New define_expand.
5444 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5446 * config/rl78/rl78.md (smaxdi3): New define_expand.
5448 2018-01-22 Carl Love <cel@us.ibm.com>
5450 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5451 LVX_V1TI): Add macro expansion.
5452 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5453 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5454 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5455 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5456 Change check to determine if the instruction is a byte reversing
5457 entry. Fix typo in comment.
5458 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5459 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5460 Add def_builtin calls for new builtins.
5461 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5462 Add define_insn expansion.
5464 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5466 * config/rl78/rl78.md (umaxdi3): New define_expand.
5468 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5470 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5471 for non-QImode registers.
5473 2018-01-22 Richard Biener <rguenther@suse.de>
5475 PR tree-optimization/83963
5476 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5477 including the loop exit block.
5478 (scop_detection::merge_sese): Likewise.
5479 (scop_detection::add_scop): Do it here instead.
5481 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5483 * doc/sourcebuild.texi (arm_softfloat): Document.
5485 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5488 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5489 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5490 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5492 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5493 David Edelsohn <dje.gcc@gmail.com>
5496 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5497 Change "crset eq" to "crset 2".
5498 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5499 (*call_indirect_aix<mode>_nospec): Likewise.
5500 (*call_value_indirect_aix<mode>_nospec): Likewise.
5501 (*call_indirect_elfv2<mode>_nospec): Likewise.
5502 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5503 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5504 change assembly output from . to $.
5505 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5506 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5507 (*tablejump<mode>_internal1_nospec): Likewise.
5509 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5512 * config/sh/sh_optimize_sett_clrt.cc:
5513 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5515 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5517 PR tree-optimization/83940
5518 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5519 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5520 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5521 use it to pass back the definition type.
5522 (vect_check_store_rhs): Likewise rhs_dt_out.
5523 (vect_build_gather_load_calls): Add a mask_dt argument and use
5524 it instead of a call to vect_is_simple_use.
5525 (vectorizable_store): Update calls to vect_check_load_store_mask
5526 and vect_check_store_rhs. Use the dt returned by the latter instead
5527 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5528 instead of calls to vect_is_simple_use. Pass the scalar rather
5529 than the vector operand to vect_is_simple_use when handling
5530 second and subsequent copies of an rhs value.
5531 (vectorizable_load): Update calls to vect_check_load_store_mask
5532 and vect_build_gather_load_calls. Use the cached mask_dt and
5533 gs_info.offset_dt instead of calls to vect_is_simple_use.
5535 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5538 * tree-emutls.c: Include gimplify.h.
5539 (lower_emutls_2): New function.
5540 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5541 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5542 it before further processing.
5545 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5546 UINTVAL (trueop1) instead of INTVAL (op1).
5548 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5552 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5553 INCOMING_FRAME_SP_OFFSET if not defined.
5554 (scan_trace): Add ENTRY argument. If true and
5555 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5556 emit a note to adjust the CFA offset.
5557 (create_cfi_notes): Adjust scan_trace callers.
5558 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5559 INCOMING_FRAME_SP_OFFSET in the CIE.
5560 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5561 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5563 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5564 * doc/tm.texi: Regenerated.
5566 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5568 PR rtl-optimization/83147
5569 * lra-constraints.c (remove_inheritance_pseudos): Use
5570 lra_substitute_pseudo_within_insn.
5572 2018-01-19 Tom de Vries <tom@codesourcery.com>
5573 Cesar Philippidis <cesar@codesourcery.com>
5576 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5578 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5581 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5582 spaces for function labels.
5584 2018-01-19 Martin Liska <mliska@suse.cz>
5586 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5587 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5588 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5589 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5590 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5591 (PRED_CONST_RETURN): Change from 69 to 65.
5592 (PRED_NULL_RETURN): Change from 91 to 71.
5593 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5594 (PRED_LOOP_GUARD): Change from 66 to 73.
5596 2018-01-19 Martin Liska <mliska@suse.cz>
5598 * predict.c (predict_insn_def): Add new assert.
5599 (struct branch_predictor): Change type to signed integer.
5600 (test_prediction_value_range): Amend test to cover
5602 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5603 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5604 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5605 (PRED_LOOP_IV_COMPARE): Likewise.
5606 * predict.h (PROB_UNINITIALIZED): Define new constant.
5608 2018-01-19 Martin Liska <mliska@suse.cz>
5610 * predict.c (dump_prediction): Add new format for
5611 analyze_brprob.py script which is enabled with -details
5613 * profile-count.h (precise_p): New function.
5615 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5617 PR tree-optimization/83922
5618 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5619 there are no statements that need masking.
5620 (vect_active_double_reduction_p): New function.
5621 (vect_analyze_loop_operations): Use it when handling phis that
5622 are not in the loop header.
5624 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5626 PR tree-optimization/83914
5627 * tree-vect-loop.c (vectorizable_induction): Don't convert
5628 init_expr or apply the peeling adjustment for inductions
5629 that are nested within the vectorized loop.
5631 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5633 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5636 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5640 * function.h (gimplify_parameters): Add gimple_seq * argument.
5641 * function.c: Include gimple.h and options.h.
5642 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5643 for the added local temporaries if needed.
5644 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5645 if there are any parameter cleanups, wrap whole body into a
5646 try/finally with the cleanups.
5648 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5651 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5652 Use GET_MODE_CLASS for scalar floating point.
5654 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5658 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5659 Fix call of call_cgraph_insertion_hooks.
5661 2018-01-18 Martin Sebor <msebor@redhat.com>
5663 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5665 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5668 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5671 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5674 * common.opt: (-ffile-prefix-map): New option.
5675 * opts.c (common_handle_option): Defer it.
5676 * opts-global.c (handle_common_deferred_options): Handle it.
5677 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5678 * file-prefix-map.h: New file.
5679 (remap_debug_filename, add_debug_prefix_map): ...here.
5680 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5681 * final.c (debug_prefix_map, add_debug_prefix_map
5682 remap_debug_filename): Move to...
5683 * file-prefix-map.c: New file.
5684 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5685 generalize, get rid of alloca(), use strrchr() instead of strchr().
5686 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5687 Implement in terms of add_prefix_map().
5688 (remap_macro_filename, remap_debug_filename): Implement in term of
5690 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5691 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5692 * dbxout.c: Include file-prefix-map.h.
5693 * varasm.c: Likewise.
5694 * vmsdbgout.c: Likewise.
5695 * xcoffout.c: Likewise.
5696 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5697 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5698 * doc/invoke.texi (-ffile-prefix-map): Document.
5699 (-fdebug-prefix-map): Update description.
5701 2018-01-18 Martin Liska <mliska@suse.cz>
5703 * config/i386/i386.c (indirect_thunk_name): Document that also
5705 (output_indirect_thunk): Document why both instructions
5706 (pause and lfence) are generated.
5708 2018-01-18 Richard Biener <rguenther@suse.de>
5710 PR tree-optimization/83887
5711 * graphite-scop-detection.c
5712 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5713 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5714 (scop_detection::merge_sese): Re-implement with a flood-fill
5715 algorithm that properly finds a SESE region if it exists.
5717 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5720 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5721 pointer_diff optimizations use view_convert instead of convert.
5723 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5725 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5726 Generate different code for -mno-speculate-indirect-jumps.
5727 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5728 (*call_indirect_aix<mode>): Disable for
5729 -mno-speculate-indirect-jumps.
5730 (*call_indirect_aix<mode>_nospec): New define_insn.
5731 (*call_value_indirect_aix<mode>): Disable for
5732 -mno-speculate-indirect-jumps.
5733 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5734 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5735 -mno-speculate-indirect-jumps.
5736 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5738 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5740 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5741 long double type, set the flags for noting the default long double
5742 type, even if we don't pass or return a long double type.
5744 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5747 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5750 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5752 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5753 support for merge[hl].
5754 (fold_mergehl_helper): New helper function.
5755 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5756 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5757 (altivec_vmrglw_direct): Add xxmrglw insn.
5759 2018-01-17 Andrew Waterman <andrew@sifive.com>
5761 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5762 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5764 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5767 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5768 call the lto_location_cache before reading the
5769 DECL_SOURCE_LOCATION of the types.
5771 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5772 Richard Sandiford <richard.sandiford@linaro.org>
5774 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5775 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5776 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5777 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5778 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5780 * config/aarch64/constraints.md (aarch64_movti_operand):
5782 * config/aarch64/predicates.md (Uti): Add new constraint.
5784 2018-01-17 Carl Love <cel@us.ibm.com>
5786 * config/rs6000/vsx.md (define_expand xl_len_r,
5787 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5788 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5790 (define_expand, define_insn): Move the shift left from the
5791 define_insn to the define_expand for lxvl and stxvl instructions.
5792 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5793 and XL_LEN_R definitions to PURE.
5795 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5797 * config/i386/i386.c (indirect_thunk_name): Declare regno
5798 as unsigned int. Compare regno with INVALID_REGNUM.
5799 (output_indirect_thunk): Ditto.
5800 (output_indirect_thunk_function): Ditto.
5801 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5802 in the call to output_indirect_thunk_function.
5804 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5807 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5808 rather than the size of inner_type to determine the stack slot size
5809 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5811 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5814 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5817 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5819 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5820 endian Linux systems to optionally enable multilibs for selecting
5821 the long double type if the user configured an explicit type.
5822 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5823 have no long double multilibs if not defined.
5824 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5825 warn if the user used -mabi={ieee,ibm}longdouble and we built
5826 multilibs for long double.
5827 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5828 appropriate multilib option.
5829 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5831 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5832 for building long double multilibs.
5833 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5835 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5837 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5840 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5842 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5845 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5848 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5851 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5853 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5854 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5856 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5858 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5859 different rtl trees depending on TARGET_64BIT.
5860 (rs6000_gen_lvx): Likewise.
5862 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5864 * config/visium/visium.md (nop): Tweak comment.
5865 (hazard_nop): Likewise.
5867 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5869 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5870 -mspeculate-indirect-jumps.
5871 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5872 for -mno-speculate-indirect-jumps.
5873 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5874 (*call_value_indirect_elfv2<mode>): Disable for
5875 -mno-speculate-indirect-jumps.
5876 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5877 (indirect_jump): Emit different RTL for
5878 -mno-speculate-indirect-jumps.
5879 (*indirect_jump<mode>): Disable for
5880 -mno-speculate-indirect-jumps.
5881 (*indirect_jump<mode>_nospec): New define_insn.
5882 (tablejump): Emit different RTL for
5883 -mno-speculate-indirect-jumps.
5884 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5885 (tablejumpsi_nospec): New define_expand.
5886 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5887 (tablejumpdi_nospec): New define_expand.
5888 (*tablejump<mode>_internal1): Disable for
5889 -mno-speculate-indirect-jumps.
5890 (*tablejump<mode>_internal1_nospec): New define_insn.
5891 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5894 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5896 * caller-save.c (insert_save): Drop unnecessary parameter. All
5899 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5900 Richard Biener <rguenth@suse.de>
5903 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5904 return early, inline manually is_gimple_sizepos. Make sure if we
5905 call gimplify_expr we don't end up with a gimple constant.
5906 * tree.c (variably_modified_type_p): Don't return true for
5907 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5908 * gimplify.h (is_gimple_sizepos): Remove.
5910 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5912 PR tree-optimization/83857
5913 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5914 vectorizable_live_operation for pure SLP statements.
5915 (vectorizable_live_operation): Handle PHIs.
5917 2018-01-16 Richard Biener <rguenther@suse.de>
5919 PR tree-optimization/83867
5920 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5921 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5923 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5926 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5927 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5928 If off is not INTEGER_CST, issue a may not be aligned warning
5929 rather than isn't aligned. Use isn%'t rather than isn't.
5930 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5932 <case MULT_EXPR>: Improve the case when bottom and one of the
5933 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5934 operand, in that case check if the other operand is multiple of
5935 bottom divided by the INTEGER_CST operand.
5937 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5940 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5941 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5942 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5943 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5944 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5945 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5946 (pa_function_arg_size): New function.
5948 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5950 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5951 in a separate statement.
5953 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5955 PR tree-optimization/83847
5956 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5957 group gathers and scatters.
5959 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5961 PR rtl-optimization/86620
5962 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5964 PR rtl-optimization/83213
5965 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5966 to last if both are JUMP_INSNs.
5968 PR tree-optimization/83843
5969 * gimple-ssa-store-merging.c
5970 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5971 store_immediate_info for bswap/nop orig_stores.
5973 2018-01-15 Andrew Waterman <andrew@sifive.com>
5975 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5977 <UDIV>: Increase cost if !TARGET_DIV.
5979 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5981 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5982 (define_attr "cr_logical_3op"): New.
5983 (cceq_ior_compare): Adjust.
5984 (cceq_ior_compare_complement): Adjust.
5985 (*cceq_rev_compare): Adjust.
5986 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5987 (is_cracked_insn): Adjust.
5988 (insn_must_be_first_in_group): Adjust.
5989 * config/rs6000/40x.md: Adjust.
5990 * config/rs6000/440.md: Adjust.
5991 * config/rs6000/476.md: Adjust.
5992 * config/rs6000/601.md: Adjust.
5993 * config/rs6000/603.md: Adjust.
5994 * config/rs6000/6xx.md: Adjust.
5995 * config/rs6000/7450.md: Adjust.
5996 * config/rs6000/7xx.md: Adjust.
5997 * config/rs6000/8540.md: Adjust.
5998 * config/rs6000/cell.md: Adjust.
5999 * config/rs6000/e300c2c3.md: Adjust.
6000 * config/rs6000/e500mc.md: Adjust.
6001 * config/rs6000/e500mc64.md: Adjust.
6002 * config/rs6000/e5500.md: Adjust.
6003 * config/rs6000/e6500.md: Adjust.
6004 * config/rs6000/mpc.md: Adjust.
6005 * config/rs6000/power4.md: Adjust.
6006 * config/rs6000/power5.md: Adjust.
6007 * config/rs6000/power6.md: Adjust.
6008 * config/rs6000/power7.md: Adjust.
6009 * config/rs6000/power8.md: Adjust.
6010 * config/rs6000/power9.md: Adjust.
6011 * config/rs6000/rs64.md: Adjust.
6012 * config/rs6000/titan.md: Adjust.
6014 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6016 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6017 ix86_indirect_branch_register logic.
6019 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6021 * config/i386/constraints.md (Bs): Update
6022 ix86_indirect_branch_register check. Don't check
6023 ix86_indirect_branch_register with GOT_memory_operand.
6025 * config/i386/predicates.md (GOT_memory_operand): Don't check
6026 ix86_indirect_branch_register here.
6027 (GOT32_symbol_operand): Likewise.
6029 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6031 * config/i386/predicates.md (constant_call_address_operand):
6032 Rewrite ix86_indirect_branch_register logic.
6033 (sibcall_insn_operand): Likewise.
6035 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6037 * config/i386/constraints.md (Bs): Replace
6038 ix86_indirect_branch_thunk_register with
6039 ix86_indirect_branch_register.
6041 * config/i386/i386.md (indirect_jump): Likewise.
6042 (tablejump): Likewise.
6043 (*sibcall_memory): Likewise.
6044 (*sibcall_value_memory): Likewise.
6045 Peepholes of indirect call and jump via memory: Likewise.
6046 * config/i386/i386.opt: Likewise.
6047 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6048 (GOT_memory_operand): Likewise.
6049 (call_insn_operand): Likewise.
6050 (sibcall_insn_operand): Likewise.
6051 (GOT32_symbol_operand): Likewise.
6053 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6056 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6057 type rather than type addr's type points to.
6058 (expand_omp_atomic_mutex): Likewise.
6059 (expand_omp_atomic): Likewise.
6061 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6064 * config/i386/i386.c (output_indirect_thunk_function): Use
6065 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6066 for __x86_return_thunk.
6068 2018-01-15 Richard Biener <rguenther@suse.de>
6071 * expmed.c (extract_bit_field_1): Fix typo.
6073 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6076 * config/arm/iterators.md (VF): New mode iterator.
6077 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6078 Remove integer-related logic from pattern.
6079 (neon_vabd<mode>_3): Likewise.
6081 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6084 * common.opt (fstrict-overflow): No longer an alias.
6085 (fwrapv-pointer): New option.
6086 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6087 also for pointer types based on flag_wrapv_pointer.
6088 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6089 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6090 opts->x_flag_wrapv got set.
6091 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6092 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6093 POINTER_TYPE_OVERFLOW_UNDEFINED.
6094 * match.pd: Likewise in address comparison pattern.
6095 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6097 2018-01-15 Richard Biener <rguenther@suse.de>
6100 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6101 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6102 Reset type names to their identifier if their TYPE_DECL doesn't
6103 have linkage (and thus is used for ODR and devirt).
6104 (save_debug_info_for_decl): Remove.
6105 (save_debug_info_for_type): Likewise.
6106 (add_tree_to_fld_list): Adjust.
6107 * tree-pretty-print.c (dump_generic_node): Make dumping of
6108 type names more robust.
6110 2018-01-15 Richard Biener <rguenther@suse.de>
6112 * BASE-VER: Bump to 8.0.1.
6114 2018-01-14 Martin Sebor <msebor@redhat.com>
6117 * builtins.c (check_access): Avoid warning when the no-warning bit
6120 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6122 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6123 * ira-color (allocno_hard_regs_compare): Likewise.
6125 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6128 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6129 Use .pushsection/.popsection.
6131 2018-01-14 Martin Sebor <msebor@redhat.com>
6134 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6136 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6138 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6139 entry from extra_headers.
6140 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6141 extra_headers, make the list bitwise identical to the i?86-*-* one.
6143 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6145 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6146 -mcmodel=large with -mindirect-branch=thunk,
6147 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6148 -mfunction-return=thunk-extern.
6149 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6150 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6151 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6153 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6155 * config/i386/i386.c (print_reg): Print the name of the full
6156 integer register without '%'.
6157 (ix86_print_operand): Handle 'V'.
6158 * doc/extend.texi: Document 'V' modifier.
6160 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6162 * config/i386/constraints.md (Bs): Disallow memory operand for
6163 -mindirect-branch-register.
6165 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6166 (GOT_memory_operand): Likewise.
6167 (call_insn_operand): Likewise.
6168 (sibcall_insn_operand): Likewise.
6169 (GOT32_symbol_operand): Likewise.
6170 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6171 for -mindirect-branch-register.
6172 (tablejump): Likewise.
6173 (*sibcall_memory): Likewise.
6174 (*sibcall_value_memory): Likewise.
6175 Disallow peepholes of indirect call and jump via memory for
6176 -mindirect-branch-register.
6177 (*call_pop): Replace m with Bw.
6178 (*call_value_pop): Likewise.
6179 (*sibcall_pop_memory): Replace m with Bs.
6180 * config/i386/i386.opt (mindirect-branch-register): New option.
6181 * doc/invoke.texi: Document -mindirect-branch-register option.
6183 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6185 * config/i386/i386-protos.h (ix86_output_function_return): New.
6186 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6187 set function_return_type.
6188 (indirect_thunk_name): Add ret_p to indicate thunk for function
6190 (output_indirect_thunk_function): Pass false to
6191 indirect_thunk_name.
6192 (ix86_output_indirect_branch_via_reg): Likewise.
6193 (ix86_output_indirect_branch_via_push): Likewise.
6194 (output_indirect_thunk_function): Create alias for function
6195 return thunk if regno < 0.
6196 (ix86_output_function_return): New function.
6197 (ix86_handle_fndecl_attribute): Handle function_return.
6198 (ix86_attribute_table): Add function_return.
6199 * config/i386/i386.h (machine_function): Add
6200 function_return_type.
6201 * config/i386/i386.md (simple_return_internal): Use
6202 ix86_output_function_return.
6203 (simple_return_internal_long): Likewise.
6204 * config/i386/i386.opt (mfunction-return=): New option.
6205 (indirect_branch): Mention -mfunction-return=.
6206 * doc/extend.texi: Document function_return function attribute.
6207 * doc/invoke.texi: Document -mfunction-return= option.
6209 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6211 * config/i386/i386-opts.h (indirect_branch): New.
6212 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6213 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6214 with local indirect jump when converting indirect call and jump.
6215 (ix86_set_indirect_branch_type): New.
6216 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6217 (indirectlabelno): New.
6218 (indirect_thunk_needed): Likewise.
6219 (indirect_thunk_bnd_needed): Likewise.
6220 (indirect_thunks_used): Likewise.
6221 (indirect_thunks_bnd_used): Likewise.
6222 (INDIRECT_LABEL): Likewise.
6223 (indirect_thunk_name): Likewise.
6224 (output_indirect_thunk): Likewise.
6225 (output_indirect_thunk_function): Likewise.
6226 (ix86_output_indirect_branch_via_reg): Likewise.
6227 (ix86_output_indirect_branch_via_push): Likewise.
6228 (ix86_output_indirect_branch): Likewise.
6229 (ix86_output_indirect_jmp): Likewise.
6230 (ix86_code_end): Call output_indirect_thunk_function if needed.
6231 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6233 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6234 (ix86_attribute_table): Add indirect_branch.
6235 * config/i386/i386.h (machine_function): Add indirect_branch_type
6236 and has_local_indirect_jump.
6237 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6239 (tablejump): Likewise.
6240 (*indirect_jump): Use ix86_output_indirect_jmp.
6241 (*tablejump_1): Likewise.
6242 (simple_return_indirect_internal): Likewise.
6243 * config/i386/i386.opt (mindirect-branch=): New option.
6244 (indirect_branch): New.
6247 (thunk-inline): Likewise.
6248 (thunk-extern): Likewise.
6249 * doc/extend.texi: Document indirect_branch function attribute.
6250 * doc/invoke.texi: Document -mindirect-branch= option.
6252 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6255 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6257 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6259 * ipa-inline.c (want_inline_small_function_p): Return false if
6260 inlining has already failed with CIF_FINAL_ERROR.
6261 (update_caller_keys): Call want_inline_small_function_p before
6263 (update_callee_keys): Likewise.
6265 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6267 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6269 (rs6000_quadword_masked_address_p): Likewise.
6270 (quad_aligned_load_p): Likewise.
6271 (quad_aligned_store_p): Likewise.
6272 (const_load_sequence_p): Add comment to describe the outer-most loop.
6273 (mimic_memory_attributes_and_flags): New function.
6274 (rs6000_gen_stvx): Likewise.
6275 (replace_swapped_aligned_store): Likewise.
6276 (rs6000_gen_lvx): Likewise.
6277 (replace_swapped_aligned_load): Likewise.
6278 (replace_swapped_load_constant): Capitalize argument name in
6279 comment describing this function.
6280 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6281 and stores that access quad-word aligned addresses and replace
6282 with stvx or lvx instructions when appropriate.
6283 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6284 New function prototype.
6285 (rs6000_quadword_masked_address_p): Likewise.
6286 (rs6000_gen_lvx): Likewise.
6287 (rs6000_gen_stvx): Likewise.
6288 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6289 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6290 when memory address is aligned.
6291 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6292 this split to select lvx instruction when memory address is aligned.
6293 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6294 instruction when memory address is aligned.
6295 (*vsx_le_perm_load_v16qi): Likewise.
6296 (four unnamed splitters): Modify to select the stvx instruction
6297 when memory is aligned.
6299 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6301 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6302 which appears in the queue multiple times.
6304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6305 Alan Hayward <alan.hayward@arm.com>
6306 David Sherwood <david.sherwood@arm.com>
6308 * tree-vectorizer.h (vec_lower_bound): New structure.
6309 (_loop_vec_info): Add check_nonzero and lower_bounds.
6310 (LOOP_VINFO_CHECK_NONZERO): New macro.
6311 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6312 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6313 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6314 fields. Make seg_len the distance travelled, not including the
6316 (dr_direction_indicator): Declare.
6317 (dr_zero_step_indicator): Likewise.
6318 (dr_known_forward_stride_p): Likewise.
6319 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6321 (runtime_alias_check_p): Allow runtime alias checks with
6323 (operator ==): Compare access_size and align.
6324 (prune_runtime_alias_test_list): Rework for new distinction between
6325 the access_size and seg_len.
6326 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6328 (get_segment_min_max): New function.
6329 (create_intersect_range_checks): Use it.
6330 (dr_step_indicator): New function.
6331 (dr_direction_indicator): Likewise.
6332 (dr_zero_step_indicator): Likewise.
6333 (dr_known_forward_stride_p): Likewise.
6334 * tree-loop-distribution.c (data_ref_segment_size): Return
6335 DR_STEP * (niters - 1).
6336 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6338 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6339 (vect_preserves_scalar_order_p): New function, split out from...
6340 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6341 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6342 (vect_vfa_access_size): New function.
6343 (vect_vfa_align): Likewise.
6344 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6345 (dump_lower_bound): New function.
6346 (vect_check_lower_bound): Likewise.
6347 (vect_small_gap_p): Likewise.
6348 (vectorizable_with_step_bound_p): Likewise.
6349 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6350 depencies if the vectorization factor is 1. Convert the checks
6351 for nonzero steps into checks on the bounds of DR_STEP. Try using
6352 a bunds check for variable steps if the minimum required step is
6353 relatively small. Update calls to the dr_with_seg_len
6354 constructor and to vect_compile_time_alias.
6355 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6357 (vect_loop_versioning): Call it.
6358 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6360 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6362 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6363 Alan Hayward <alan.hayward@arm.com>
6364 David Sherwood <david.sherwood@arm.com>
6366 * doc/sourcebuild.texi (vect_scatter_store): Document.
6367 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6369 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6371 * genopinit.c (main): Add supports_vec_scatter_store and
6372 supports_vec_scatter_store_cached to target_optabs.
6373 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6374 IFN_MASK_SCATTER_STORE.
6375 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6377 * internal-fn.h (internal_store_fn_p): Declare.
6378 (internal_fn_stored_value_index): Likewise.
6379 * internal-fn.c (scatter_store_direct): New macro.
6380 (expand_scatter_store_optab_fn): New function.
6381 (direct_scatter_store_optab_supported_p): New macro.
6382 (internal_store_fn_p): New function.
6383 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6384 IFN_MASK_SCATTER_STORE.
6385 (internal_fn_mask_index): Likewise.
6386 (internal_fn_stored_value_index): New function.
6387 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6389 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6390 * optabs-query.c (supports_vec_scatter_store_p): New function.
6391 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6392 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6393 true for scatter stores.
6394 (vect_gather_scatter_fn_p): Handle scatter stores too.
6395 (vect_check_gather_scatter): Consider using scatter stores if
6396 supports_vec_scatter_store_p.
6397 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6399 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6400 internal_fn_stored_value_index.
6401 (check_load_store_masking): Handle scatter stores too.
6402 (vect_get_store_rhs): Make public.
6403 (vectorizable_call): Use internal_store_fn_p.
6404 (vectorizable_store): Handle scatter store internal functions.
6405 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6406 when deciding whether the end of the group has been reached.
6407 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6408 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6409 (mask_scatter_store<mode>): New insns.
6411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6412 Alan Hayward <alan.hayward@arm.com>
6413 David Sherwood <david.sherwood@arm.com>
6415 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6416 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6417 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6419 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6420 Use vect_truncate_gather_scatter_offset if we can't treat the
6421 operation as a normal gather load or scatter store.
6422 (get_group_load_store_type): Take the gather_scatter_info
6423 as argument. Try using a gather load or scatter store for
6424 single-element groups.
6425 (get_load_store_type): Update calls to get_group_load_store_type
6426 and vect_use_strided_gather_scatters_p.
6428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6429 Alan Hayward <alan.hayward@arm.com>
6430 David Sherwood <david.sherwood@arm.com>
6432 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6433 optional tree argument.
6434 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6436 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6437 but continue to use the current value as a fallback.
6438 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6439 to compare the updates.
6440 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6441 (get_load_store_type): Use it when handling a strided access.
6442 (vect_get_strided_load_store_ops): New function.
6443 (vect_get_data_ptr_increment): Likewise.
6444 (vectorizable_load): Handle strided gather loads. Always pass
6445 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6447 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6448 Alan Hayward <alan.hayward@arm.com>
6449 David Sherwood <david.sherwood@arm.com>
6451 * doc/md.texi (gather_load@var{m}): Document.
6452 (mask_gather_load@var{m}): Likewise.
6453 * genopinit.c (main): Add supports_vec_gather_load and
6454 supports_vec_gather_load_cached to target_optabs.
6455 * optabs-tree.c (init_tree_optimization_optabs): Use
6456 ggc_cleared_alloc to allocate target_optabs.
6457 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6458 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6460 * internal-fn.h (internal_load_fn_p): Declare.
6461 (internal_gather_scatter_fn_p): Likewise.
6462 (internal_fn_mask_index): Likewise.
6463 (internal_gather_scatter_fn_supported_p): Likewise.
6464 * internal-fn.c (gather_load_direct): New macro.
6465 (expand_gather_load_optab_fn): New function.
6466 (direct_gather_load_optab_supported_p): New macro.
6467 (direct_internal_fn_optab): New function.
6468 (internal_load_fn_p): Likewise.
6469 (internal_gather_scatter_fn_p): Likewise.
6470 (internal_fn_mask_index): Likewise.
6471 (internal_gather_scatter_fn_supported_p): Likewise.
6472 * optabs-query.c (supports_at_least_one_mode_p): New function.
6473 (supports_vec_gather_load_p): Likewise.
6474 * optabs-query.h (supports_vec_gather_load_p): Declare.
6475 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6476 and memory_type field.
6477 (NUM_PATTERNS): Bump to 15.
6478 * tree-vect-data-refs.c: Include internal-fn.h.
6479 (vect_gather_scatter_fn_p): New function.
6480 (vect_describe_gather_scatter_call): Likewise.
6481 (vect_check_gather_scatter): Try using internal functions for
6482 gather loads. Recognize existing calls to a gather load function.
6483 (vect_analyze_data_refs): Consider using gather loads if
6484 supports_vec_gather_load_p.
6485 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6486 (vect_get_gather_scatter_offset_type): Likewise.
6487 (vect_convert_mask_for_vectype): Likewise.
6488 (vect_add_conversion_to_patterm): Likewise.
6489 (vect_try_gather_scatter_pattern): Likewise.
6490 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6491 (vect_vect_recog_func_ptrs): Add it.
6492 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6493 internal_fn_mask_index and internal_gather_scatter_fn_p.
6494 (check_load_store_masking): Take the gather_scatter_info as an
6495 argument and handle gather loads.
6496 (vect_get_gather_scatter_ops): New function.
6497 (vectorizable_call): Check internal_load_fn_p.
6498 (vectorizable_load): Likewise. Handle gather load internal
6500 (vectorizable_store): Update call to check_load_store_masking.
6501 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6502 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6503 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6504 (aarch64_gather_scale_operand_d): New predicates.
6505 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6506 (mask_gather_load<mode>): New insns.
6508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6509 Alan Hayward <alan.hayward@arm.com>
6510 David Sherwood <david.sherwood@arm.com>
6512 * optabs.def (fold_left_plus_optab): New optab.
6513 * doc/md.texi (fold_left_plus_@var{m}): Document.
6514 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6515 * internal-fn.c (fold_left_direct): Define.
6516 (expand_fold_left_optab_fn): Likewise.
6517 (direct_fold_left_optab_supported_p): Likewise.
6518 * fold-const-call.c (fold_const_fold_left): New function.
6519 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6520 * tree-parloops.c (valid_reduction_p): New function.
6521 (gather_scalar_reductions): Use it.
6522 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6523 (vect_finish_replace_stmt): Declare.
6524 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6525 (needs_fold_left_reduction_p): New function, split out from...
6526 (vect_is_simple_reduction): ...here. Accept reductions that
6527 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6528 (vect_force_simple_reduction): Also store the reduction type in
6529 the assignment's STMT_VINFO_REDUC_TYPE.
6530 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6531 (merge_with_identity): New function.
6532 (vect_expand_fold_left): Likewise.
6533 (vectorize_fold_left_reduction): Likewise.
6534 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6535 scalar phi in place for it. Check for target support and reject
6536 cases that would reassociate the operation. Defer the transform
6537 phase to vectorize_fold_left_reduction.
6538 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6539 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6540 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6544 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6545 call to ifc_temp_var.
6547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6548 Alan Hayward <alan.hayward@arm.com>
6549 David Sherwood <david.sherwood@arm.com>
6551 * target.def (legitimize_address_displacement): Take the original
6552 offset as a poly_int.
6553 * targhooks.h (default_legitimize_address_displacement): Update
6555 * targhooks.c (default_legitimize_address_displacement): Likewise.
6556 * doc/tm.texi: Regenerate.
6557 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6558 as an argument, moving assert of ad->disp == ad->disp_term to...
6559 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6560 Try calling targetm.legitimize_address_displacement before expanding
6561 the address rather than afterwards, and adjust for the new interface.
6562 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6563 Match the new hook interface. Handle SVE addresses.
6564 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6569 * Makefile.in (OBJS): Add early-remat.o.
6570 * target.def (select_early_remat_modes): New hook.
6571 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6572 * doc/tm.texi: Regenerate.
6573 * targhooks.h (default_select_early_remat_modes): Declare.
6574 * targhooks.c (default_select_early_remat_modes): New function.
6575 * timevar.def (TV_EARLY_REMAT): New timevar.
6576 * passes.def (pass_early_remat): New pass.
6577 * tree-pass.h (make_pass_early_remat): Declare.
6578 * early-remat.c: New file.
6579 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6581 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6584 Alan Hayward <alan.hayward@arm.com>
6585 David Sherwood <david.sherwood@arm.com>
6587 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6588 vfm1 with a bound_epilog parameter.
6589 (vect_do_peeling): Update calls accordingly, and move the prologue
6590 call earlier in the function. Treat the base bound_epilog as 0 for
6591 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6592 this base when peeling for gaps.
6593 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6594 with fully-masked loops.
6595 (vect_estimate_min_profitable_iters): Handle the single peeled
6596 iteration in that case.
6598 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6599 Alan Hayward <alan.hayward@arm.com>
6600 David Sherwood <david.sherwood@arm.com>
6602 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6603 single-element interleaving even if the size is not a power of 2.
6604 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6605 accesses for single-element interleaving if the group size is
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/md.texi (fold_extract_last_@var{m}): Document.
6613 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6614 * optabs.def (fold_extract_last_optab): New optab.
6615 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6616 * internal-fn.c (fold_extract_direct): New macro.
6617 (expand_fold_extract_optab_fn): Likewise.
6618 (direct_fold_extract_optab_supported_p): Likewise.
6619 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6620 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6621 EXTRACT_LAST_REDUCTION.
6622 (get_initial_def_for_reduction): Do not create an initial vector
6623 for EXTRACT_LAST_REDUCTION reductions.
6624 (vectorizable_reduction): Leave the scalar phi in place for
6625 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6626 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6627 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6628 transform phase to vectorizable_condition.
6629 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6631 (vect_finish_stmt_generation): ...here.
6632 (vect_finish_replace_stmt): New function.
6633 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6634 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6636 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6639 Alan Hayward <alan.hayward@arm.com>
6640 David Sherwood <david.sherwood@arm.com>
6642 * doc/md.texi (extract_last_@var{m}): Document.
6643 * optabs.def (extract_last_optab): New optab.
6644 * internal-fn.def (EXTRACT_LAST): New internal function.
6645 * internal-fn.c (cond_unary_direct): New macro.
6646 (expand_cond_unary_optab_fn): Likewise.
6647 (direct_cond_unary_optab_supported_p): Likewise.
6648 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6649 loops using EXTRACT_LAST.
6650 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6651 (extract_last_<mode>): ...this optab.
6652 (vec_extract<mode><Vel>): Update accordingly.
6654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6655 Alan Hayward <alan.hayward@arm.com>
6656 David Sherwood <david.sherwood@arm.com>
6658 * target.def (empty_mask_is_expensive): New hook.
6659 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6660 * doc/tm.texi: Regenerate.
6661 * targhooks.h (default_empty_mask_is_expensive): Declare.
6662 * targhooks.c (default_empty_mask_is_expensive): New function.
6663 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6664 if the target says that empty masks are expensive.
6665 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6667 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6669 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6670 Alan Hayward <alan.hayward@arm.com>
6671 David Sherwood <david.sherwood@arm.com>
6673 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6674 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6675 (vect_use_loop_mask_for_alignment_p): New function.
6676 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6677 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6678 niters_skip argument. Make sure that the first niters_skip elements
6679 of the first iteration are inactive.
6680 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6681 Update call to vect_set_loop_masks_directly.
6682 (get_misalign_in_elems): New function, split out from...
6683 (vect_gen_prolog_loop_niters): ...here.
6684 (vect_update_init_of_dr): Take a code argument that specifies whether
6685 the adjustment should be added or subtracted.
6686 (vect_update_init_of_drs): Likewise.
6687 (vect_prepare_for_masked_peels): New function.
6688 (vect_do_peeling): Skip prologue peeling if we're using a mask
6689 instead. Update call to vect_update_inits_of_drs.
6690 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6692 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6693 alignment. Do not include the number of peeled iterations in
6694 the minimum threshold in that case.
6695 (vectorizable_induction): Adjust the start value down by
6696 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6697 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6698 Take the number of skipped iterations into account when calculating
6700 * tree-vect-stmts.c (vect_gen_while_not): New function.
6702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6703 Alan Hayward <alan.hayward@arm.com>
6704 David Sherwood <david.sherwood@arm.com>
6706 * doc/sourcebuild.texi (vect_fully_masked): Document.
6707 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6709 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6711 (vect_analyze_loop_2): ...here. Don't check the vectorization
6712 factor against the number of loop iterations if the loop is
6715 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6716 Alan Hayward <alan.hayward@arm.com>
6717 David Sherwood <david.sherwood@arm.com>
6719 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6720 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6721 (dump_groups): Update accordingly.
6722 (iv_use::mem_type): New member variable.
6723 (address_p): New function.
6724 (record_use): Add a mem_type argument and initialize the new
6726 (record_group_use): Add a mem_type argument. Use address_p.
6727 Remove obsolete null checks of base_object. Update call to record_use.
6728 (find_interesting_uses_op): Update call to record_group_use.
6729 (find_interesting_uses_cond): Likewise.
6730 (find_interesting_uses_address): Likewise.
6731 (get_mem_type_for_internal_fn): New function.
6732 (find_address_like_use): Likewise.
6733 (find_interesting_uses_stmt): Try find_address_like_use before
6734 calling find_interesting_uses_op.
6735 (addr_offset_valid_p): Use the iv mem_type field as the type
6736 of the addressed memory.
6737 (add_autoinc_candidates): Likewise.
6738 (get_address_cost): Likewise.
6739 (split_small_address_groups_p): Use address_p.
6740 (split_address_groups): Likewise.
6741 (add_iv_candidate_for_use): Likewise.
6742 (autoinc_possible_for_pair): Likewise.
6743 (rewrite_groups): Likewise.
6744 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6745 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6746 (get_alias_ptr_type_for_ptr_address): New function.
6747 (rewrite_use_address): Rewrite address uses in calls that were
6748 identified by find_address_like_use.
6750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6751 Alan Hayward <alan.hayward@arm.com>
6752 David Sherwood <david.sherwood@arm.com>
6754 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6756 * gimple-expr.h (is_gimple_addressable: Likewise.
6757 * gimple-expr.c (is_gimple_address): Likewise.
6758 * internal-fn.c (expand_call_mem_ref): New function.
6759 (expand_mask_load_optab_fn): Use it.
6760 (expand_mask_store_optab_fn): Likewise.
6762 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6763 Alan Hayward <alan.hayward@arm.com>
6764 David Sherwood <david.sherwood@arm.com>
6766 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6767 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6768 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6769 (cond_umax@var{mode}): Document.
6770 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6771 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6772 (cond_umin_optab, cond_umax_optab): New optabs.
6773 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6774 (COND_IOR, COND_XOR): New internal functions.
6775 * internal-fn.h (get_conditional_internal_fn): Declare.
6776 * internal-fn.c (cond_binary_direct): New macro.
6777 (expand_cond_binary_optab_fn): Likewise.
6778 (direct_cond_binary_optab_supported_p): Likewise.
6779 (get_conditional_internal_fn): New function.
6780 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6781 Cope with reduction statements that are vectorized as calls rather
6783 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6784 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6785 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6786 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6787 (UNSPEC_COND_EOR): New unspecs.
6788 (optab): Add mappings for them.
6789 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6790 (sve_int_op, sve_fp_op): New int attributes.
6792 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6793 Alan Hayward <alan.hayward@arm.com>
6794 David Sherwood <david.sherwood@arm.com>
6796 * optabs.def (while_ult_optab): New optab.
6797 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6798 * internal-fn.def (WHILE_ULT): New internal function.
6799 * internal-fn.h (direct_internal_fn_supported_p): New override
6800 that takes two types as argument.
6801 * internal-fn.c (while_direct): New macro.
6802 (expand_while_optab_fn): New function.
6803 (convert_optab_supported_p): Likewise.
6804 (direct_while_optab_supported_p): New macro.
6805 * wide-int.h (wi::udiv_ceil): New function.
6806 * tree-vectorizer.h (rgroup_masks): New structure.
6807 (vec_loop_masks): New typedef.
6808 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6810 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6811 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6812 (vect_max_vf): New function.
6813 (slpeel_make_loop_iterate_ntimes): Delete.
6814 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6815 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6816 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6817 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6818 internal-fn.h, stor-layout.h and optabs-query.h.
6819 (vect_set_loop_mask): New function.
6820 (add_preheader_seq): Likewise.
6821 (add_header_seq): Likewise.
6822 (interleave_supported_p): Likewise.
6823 (vect_maybe_permute_loop_masks): Likewise.
6824 (vect_set_loop_masks_directly): Likewise.
6825 (vect_set_loop_condition_masked): Likewise.
6826 (vect_set_loop_condition_unmasked): New function, split out from
6827 slpeel_make_loop_iterate_ntimes.
6828 (slpeel_make_loop_iterate_ntimes): Rename to..
6829 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6830 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6831 (vect_do_peeling): Update call accordingly.
6832 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6834 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6835 mask_compare_type, can_fully_mask_p and fully_masked_p.
6836 (release_vec_loop_masks): New function.
6837 (_loop_vec_info): Use it to free the loop masks.
6838 (can_produce_all_loop_masks_p): New function.
6839 (vect_get_max_nscalars_per_iter): Likewise.
6840 (vect_verify_full_masking): Likewise.
6841 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6842 retries, and free the mask rgroups before retrying. Check loop-wide
6843 reasons for disallowing fully-masked loops. Make the final decision
6844 about whether use a fully-masked loop or not.
6845 (vect_estimate_min_profitable_iters): Do not assume that peeling
6846 for the number of iterations will be needed for fully-masked loops.
6847 (vectorizable_reduction): Disable fully-masked loops.
6848 (vectorizable_live_operation): Likewise.
6849 (vect_halve_mask_nunits): New function.
6850 (vect_double_mask_nunits): Likewise.
6851 (vect_record_loop_mask): Likewise.
6852 (vect_get_loop_mask): Likewise.
6853 (vect_transform_loop): Handle the case in which the final loop
6854 iteration might handle a partial vector. Call vect_set_loop_condition
6855 instead of slpeel_make_loop_iterate_ntimes.
6856 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6857 (check_load_store_masking): New function.
6858 (prepare_load_store_mask): Likewise.
6859 (vectorizable_store): Handle fully-masked loops.
6860 (vectorizable_load): Likewise.
6861 (supportable_widening_operation): Use vect_halve_mask_nunits for
6863 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6864 (vect_gen_while): New function.
6865 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6866 (aarch64_uqdec<mode>): New insn.
6868 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6869 Alan Hayward <alan.hayward@arm.com>
6870 David Sherwood <david.sherwood@arm.com>
6872 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6873 (reduc_xor_scal_optab): New optabs.
6874 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6875 (reduc_xor_scal_@var{m}): Document.
6876 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6877 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6879 * fold-const-call.c (fold_const_call): Handle them.
6880 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6881 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6882 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6883 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6884 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6885 (UNSPEC_XORV): New unspecs.
6886 (optab): Add entries for them.
6887 (BITWISEV): New int iterator.
6888 (bit_reduc_op): New int attributes.
6890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6891 Alan Hayward <alan.hayward@arm.com>
6892 David Sherwood <david.sherwood@arm.com>
6894 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6895 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6896 * optabs.def (vec_shl_insert_optab): New optab.
6897 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6898 (duplicate_and_interleave): Likewise.
6899 * tree-vect-loop.c: Include internal-fn.h.
6900 (neutral_op_for_slp_reduction): New function, split out from
6901 get_initial_defs_for_reduction.
6902 (get_initial_def_for_reduction): Handle option 2 for variable-length
6903 vectors by loading the neutral value into a vector and then shifting
6904 the initial value into element 0.
6905 (get_initial_defs_for_reduction): Replace the code argument with
6906 the neutral value calculated by neutral_op_for_slp_reduction.
6907 Use gimple_build_vector for constant-length vectors.
6908 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6909 but the first group_size elements have a neutral value.
6910 Use duplicate_and_interleave otherwise.
6911 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6912 Update call to get_initial_defs_for_reduction. Handle SLP
6913 reductions for variable-length vectors by creating one vector
6914 result for each scalar result, with the elements associated
6915 with other scalar results stubbed out with the neutral value.
6916 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6917 Require IFN_VEC_SHL_INSERT for double reductions on
6918 variable-length vectors, or SLP reductions that have
6919 a neutral value. Require can_duplicate_and_interleave_p
6920 support for variable-length unchained SLP reductions if there
6921 is no neutral value, such as for MIN/MAX reductions. Also require
6922 the number of vector elements to be a multiple of the number of
6923 SLP statements when doing variable-length unchained SLP reductions.
6924 Update call to vect_create_epilog_for_reduction.
6925 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6926 and remove initial values.
6927 (duplicate_and_interleave): Make public.
6928 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6929 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6931 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6932 Alan Hayward <alan.hayward@arm.com>
6933 David Sherwood <david.sherwood@arm.com>
6935 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6936 (can_duplicate_and_interleave_p): New function.
6937 (vect_get_and_check_slp_defs): Take the vector of statements
6938 rather than just the current one. Remove excess parentheses.
6939 Restriction rejectinon of vect_constant_def and vect_external_def
6940 for variable-length vectors to boolean types, or types for which
6941 can_duplicate_and_interleave_p is false.
6942 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6943 (duplicate_and_interleave): New function.
6944 (vect_get_constant_vectors): Use gimple_build_vector for
6945 constant-length vectors and suitable variable-length constant
6946 vectors. Use duplicate_and_interleave for other variable-length
6947 vectors. Don't defer the update when inserting new statements.
6949 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6950 Alan Hayward <alan.hayward@arm.com>
6951 David Sherwood <david.sherwood@arm.com>
6953 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6954 min_profitable_iters doesn't go negative.
6956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6957 Alan Hayward <alan.hayward@arm.com>
6958 David Sherwood <david.sherwood@arm.com>
6960 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6961 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6962 * optabs.def (vec_mask_load_lanes_optab): New optab.
6963 (vec_mask_store_lanes_optab): Likewise.
6964 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6965 (MASK_STORE_LANES): Likewise.
6966 * internal-fn.c (mask_load_lanes_direct): New macro.
6967 (mask_store_lanes_direct): Likewise.
6968 (expand_mask_load_optab_fn): Handle masked operations.
6969 (expand_mask_load_lanes_optab_fn): New macro.
6970 (expand_mask_store_optab_fn): Handle masked operations.
6971 (expand_mask_store_lanes_optab_fn): New macro.
6972 (direct_mask_load_lanes_optab_supported_p): Likewise.
6973 (direct_mask_store_lanes_optab_supported_p): Likewise.
6974 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6976 (vect_load_lanes_supported): Likewise.
6977 * tree-vect-data-refs.c (strip_conversion): New function.
6978 (can_group_stmts_p): Likewise.
6979 (vect_analyze_data_ref_accesses): Use it instead of checking
6980 for a pair of assignments.
6981 (vect_store_lanes_supported): Take a masked_p parameter.
6982 (vect_load_lanes_supported): Likewise.
6983 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6984 vect_store_lanes_supported and vect_load_lanes_supported.
6985 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6986 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6987 parameter. Don't allow gaps for masked accesses.
6988 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6989 and vect_load_lanes_supported.
6990 (get_load_store_type): Take a masked_p parameter and update
6991 call to get_group_load_store_type.
6992 (vectorizable_store): Update call to get_load_store_type.
6993 Handle IFN_MASK_STORE_LANES.
6994 (vectorizable_load): Update call to get_load_store_type.
6995 Handle IFN_MASK_LOAD_LANES.
6997 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6998 Alan Hayward <alan.hayward@arm.com>
6999 David Sherwood <david.sherwood@arm.com>
7001 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7003 * config/aarch64/aarch64-protos.h
7004 (aarch64_sve_struct_memory_operand_p): Declare.
7005 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7006 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7007 (VPRED, vpred): Handle SVE structure modes.
7008 * config/aarch64/constraints.md (Utx): New constraint.
7009 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7010 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7011 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7012 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7013 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7014 structure modes. Split into pieces after RA.
7015 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7016 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7018 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7019 SVE structure modes.
7020 (aarch64_classify_address): Likewise.
7021 (sizetochar): Move earlier in file.
7022 (aarch64_print_operand): Handle SVE register lists.
7023 (aarch64_array_mode): New function.
7024 (aarch64_sve_struct_memory_operand_p): Likewise.
7025 (TARGET_ARRAY_MODE): Redefine.
7027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7028 Alan Hayward <alan.hayward@arm.com>
7029 David Sherwood <david.sherwood@arm.com>
7031 * target.def (array_mode): New target hook.
7032 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7033 * doc/tm.texi: Regenerate.
7034 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7035 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7036 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7038 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7042 Alan Hayward <alan.hayward@arm.com>
7043 David Sherwood <david.sherwood@arm.com>
7045 * fold-const.c (fold_binary_loc): Check the argument types
7046 rather than the result type when testing for a vector operation.
7048 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7050 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7051 * doc/tm.texi: Regenerate.
7053 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7054 Alan Hayward <alan.hayward@arm.com>
7055 David Sherwood <david.sherwood@arm.com>
7057 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7058 (sve): Document new AArch64 extension.
7059 * doc/md.texi (w): Extend the description of the AArch64
7060 constraint to include SVE vectors.
7061 (Upl, Upa): Document new AArch64 predicate constraints.
7062 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7064 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7065 (msve-vector-bits=): New option.
7066 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7067 SVE when these are disabled.
7068 (sve): New extension.
7069 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7070 modes. Adjust their number of units based on aarch64_sve_vg.
7071 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7072 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7073 aarch64_addr_query_type.
7074 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7075 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7076 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7077 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7078 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7079 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7080 (aarch64_simd_imm_zero_p): Delete.
7081 (aarch64_check_zero_based_sve_index_immediate): Declare.
7082 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7083 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7084 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7085 (aarch64_sve_float_mul_immediate_p): Likewise.
7086 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7088 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7089 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7090 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7091 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7092 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7093 (aarch64_regmode_natural_size): Likewise.
7094 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7095 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7097 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7098 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7099 for VG and the SVE predicate registers.
7100 (V_ALIASES): Add a "z"-prefixed alias.
7101 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7102 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7103 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7104 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7105 (REG_CLASS_NAMES): Add entries for them.
7106 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7107 and the predicate registers.
7108 (aarch64_sve_vg): Declare.
7109 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7110 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7111 (REGMODE_NATURAL_SIZE): Define.
7112 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7114 * config/aarch64/aarch64.c: Include cfgrtl.h.
7115 (simd_immediate_info): Add a constructor for series vectors,
7116 and an associated step field.
7117 (aarch64_sve_vg): New variable.
7118 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7119 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7120 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7121 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7122 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7123 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7124 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7125 (aarch64_get_mask_mode): New functions.
7126 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7127 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7128 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7129 predicate modes and predicate registers. Explicitly restrict
7130 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7131 to store a vector mode if it is recognized by
7132 aarch64_classify_vector_mode.
7133 (aarch64_regmode_natural_size): New function.
7134 (aarch64_hard_regno_caller_save_mode): Return the original mode
7136 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7137 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7138 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7139 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7141 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7142 does not overlap dest if the function is frame-related. Handle
7144 (aarch64_split_add_offset): New function.
7145 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7146 them aarch64_add_offset.
7147 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7148 and update call to aarch64_sub_sp.
7149 (aarch64_add_cfa_expression): New function.
7150 (aarch64_expand_prologue): Pass extra temporary registers to the
7151 functions above. Handle the case in which we need to emit new
7152 DW_CFA_expressions for registers that were originally saved
7153 relative to the stack pointer, but now have to be expressed
7154 relative to the frame pointer.
7155 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7157 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7158 IP0 and IP1 values for SVE frames.
7159 (aarch64_expand_vec_series): New function.
7160 (aarch64_expand_sve_widened_duplicate): Likewise.
7161 (aarch64_expand_sve_const_vector): Likewise.
7162 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7163 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7164 into the register, rather than emitting a SET directly.
7165 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7166 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7167 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7168 (offset_9bit_signed_scaled_p): New functions.
7169 (aarch64_replicate_bitmask_imm): New function.
7170 (aarch64_bitmask_imm): Use it.
7171 (aarch64_cannot_force_const_mem): Reject expressions involving
7172 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7173 (aarch64_classify_index): Handle SVE indices, by requiring
7174 a plain register index with a scale that matches the element size.
7175 (aarch64_classify_address): Handle SVE addresses. Assert that
7176 the mode of the address is VOIDmode or an integer mode.
7177 Update call to aarch64_classify_symbol.
7178 (aarch64_classify_symbolic_expression): Update call to
7179 aarch64_classify_symbol.
7180 (aarch64_const_vec_all_in_range_p): New function.
7181 (aarch64_print_vector_float_operand): Likewise.
7182 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7183 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7184 and the FP immediates 1.0 and 0.5.
7185 (aarch64_print_address_internal): Handle SVE addresses.
7186 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7187 (aarch64_regno_regclass): Handle predicate registers.
7188 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7190 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7191 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7192 (aarch64_convert_sve_vector_bits): New function.
7193 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7194 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7196 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7197 Handle SVE vector and predicate modes. Accept VL-based constants
7198 that need only one temporary register, and VL offsets that require
7199 no temporary registers.
7200 (aarch64_conditional_register_usage): Mark the predicate registers
7201 as fixed if SVE isn't available.
7202 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7203 Return true for SVE vector and predicate modes.
7204 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7205 rather than an unsigned int. Handle SVE modes.
7206 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7208 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7210 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7211 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7212 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7213 (aarch64_sve_float_mul_immediate_p): New functions.
7214 (aarch64_sve_valid_immediate): New function.
7215 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7216 Explicitly reject structure modes. Check for INDEX constants.
7217 Handle PTRUE and PFALSE constants.
7218 (aarch64_check_zero_based_sve_index_immediate): New function.
7219 (aarch64_simd_imm_zero_p): Delete.
7220 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7221 vector modes. Accept constants in the range of CNT[BHWD].
7222 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7223 ask for an Advanced SIMD mode.
7224 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7225 (aarch64_simd_vector_alignment): Handle SVE predicates.
7226 (aarch64_vectorize_preferred_vector_alignment): New function.
7227 (aarch64_simd_vector_alignment_reachable): Use it instead of
7229 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7230 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7232 (MAX_VECT_LEN): Delete.
7233 (expand_vec_perm_d): Add a vec_flags field.
7234 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7235 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7236 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7238 (aarch64_evpc_rev): Rename to...
7239 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7240 (aarch64_evpc_rev_global): New function.
7241 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7242 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7244 (aarch64_evpc_sve_tbl): New function.
7245 (aarch64_expand_vec_perm_const_1): Update after rename of
7246 aarch64_evpc_rev. Handle SVE permutes too, trying
7247 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7248 than aarch64_evpc_tbl.
7249 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7250 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7251 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7252 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7253 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7254 (aarch64_expand_sve_vcond): New functions.
7255 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7256 of aarch64_vector_mode_p.
7257 (aarch64_dwarf_poly_indeterminate_value): New function.
7258 (aarch64_compute_pressure_classes): Likewise.
7259 (aarch64_can_change_mode_class): Likewise.
7260 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7261 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7262 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7263 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7264 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7265 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7266 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7267 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7269 (Dn, Dl, Dr): Accept const as well as const_vector.
7270 (Dz): Likewise. Compare against CONST0_RTX.
7271 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7272 of "vector" where appropriate.
7273 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7274 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7275 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7276 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7277 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7278 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7279 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7280 (v_int_equiv): Extend to SVE modes.
7281 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7283 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7284 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7285 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7286 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7287 (SVE_COND_FP_CMP): New int iterators.
7288 (perm_hilo): Handle the new unpack unspecs.
7289 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7291 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7292 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7293 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7294 (aarch64_equality_operator, aarch64_constant_vector_operand)
7295 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7296 (aarch64_sve_nonimmediate_operand): Likewise.
7297 (aarch64_sve_general_operand): Likewise.
7298 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7299 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7300 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7301 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7302 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7303 (aarch64_sve_float_arith_immediate): Likewise.
7304 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7305 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7306 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7307 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7308 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7309 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7310 (aarch64_sve_float_arith_operand): Likewise.
7311 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7312 (aarch64_sve_float_mul_operand): Likewise.
7313 (aarch64_sve_vec_perm_operand): Likewise.
7314 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7315 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7316 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7317 as well as const_vector.
7318 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7319 in file. Use CONST0_RTX and CONSTM1_RTX.
7320 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7321 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7322 Use aarch64_simd_imm_zero.
7323 * config/aarch64/aarch64-sve.md: New file.
7324 * config/aarch64/aarch64.md: Include it.
7325 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7326 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7327 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7328 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7329 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7330 (sve): New attribute.
7331 (enabled): Disable instructions with the sve attribute unless
7333 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7334 aarch64_expand_mov_immediate.
7335 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7336 CNT[BHSD] immediates.
7337 (movti): Split CONST_POLY_INT moves into two halves.
7338 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7339 Split additions that need a temporary here if the destination
7340 is the stack pointer.
7341 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7342 (*add<mode>3_poly_1): New instruction.
7343 (set_clobber_cc): New expander.
7345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7347 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7348 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7349 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7350 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7351 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7352 Change innermode from fixed_mode_size to machine_mode.
7353 (simplify_subreg): Update call accordingly. Handle a constant-sized
7354 subreg of a variable-length CONST_VECTOR.
7356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7357 Alan Hayward <alan.hayward@arm.com>
7358 David Sherwood <david.sherwood@arm.com>
7360 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7361 (add_offset_to_base): New function, split out from...
7362 (create_mem_ref): ...here. When handling a scale other than 1,
7363 check first whether the address is valid without the offset.
7364 Add it into the base if so, leaving the index and scale as-is.
7366 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7369 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7370 fold_for_warn before checking if arg2 is INTEGER_CST.
7372 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7374 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7375 (store_multiple_operation): Delete.
7376 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7377 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7378 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7379 guarded by TARGET_STRING.
7380 (rs6000_output_load_multiple): Delete.
7381 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7382 OPTION_MASK_STRING / TARGET_STRING handling.
7383 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7384 (const rs6000_opt_masks) <"string">: Change mask to 0.
7385 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7386 (MASK_STRING): Delete.
7387 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7389 (load_multiple): Delete.
7396 (store_multiple): Delete.
7403 (movmemsi_8reg): Delete.
7404 (corresponding unnamed define_insn): Delete.
7405 (movmemsi_6reg): Delete.
7406 (corresponding unnamed define_insn): Delete.
7407 (movmemsi_4reg): Delete.
7408 (corresponding unnamed define_insn): Delete.
7409 (movmemsi_2reg): Delete.
7410 (corresponding unnamed define_insn): Delete.
7411 (movmemsi_1reg): Delete.
7412 (corresponding unnamed define_insn): Delete.
7413 * config/rs6000/rs6000.opt (mno-string): New.
7414 (mstring): Replace by deprecation warning stub.
7415 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7417 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7419 * regrename.c (regrename_do_replace): If replacing the same
7420 reg multiple times, try to reuse last created gen_raw_REG.
7423 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7424 main to workaround a bug in GDB.
7426 2018-01-12 Tom de Vries <tom@codesourcery.com>
7429 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7431 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7433 PR rtl-optimization/80481
7434 * ira-color.c (get_cap_member): New function.
7435 (allocnos_conflict_by_live_ranges_p): Use it.
7436 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7437 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7439 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7442 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7443 (*saddl_se_1): Ditto.
7445 (*ssubl_se_1): Ditto.
7447 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7449 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7450 rather than wi::to_widest for DR_INITs.
7451 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7452 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7453 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7455 (vect_analyze_group_access_1): Note that here.
7457 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7459 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7460 polynomial type sizes.
7462 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7464 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7465 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7466 (gimple_add_tmp_var): Likewise.
7468 2018-01-12 Martin Liska <mliska@suse.cz>
7470 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7471 (gimple_alloc_sizes): Likewise.
7472 (dump_gimple_statistics): Use PRIu64 in printf format.
7473 * gimple.h: Change uint64_t to int.
7475 2018-01-12 Martin Liska <mliska@suse.cz>
7477 * tree-core.h: Use uint64_t instead of int.
7478 * tree.c (tree_node_counts): Likewise.
7479 (tree_node_sizes): Likewise.
7480 (dump_tree_statistics): Use PRIu64 in printf format.
7482 2018-01-12 Martin Liska <mliska@suse.cz>
7484 * Makefile.in: As qsort_chk is implemented in vec.c, add
7485 vec.o to linkage of gencfn-macros.
7486 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7487 passing the info to record_node_allocation_statistics.
7488 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7490 * ggc-common.c (struct ggc_usage): Add operator== and use
7491 it in operator< and compare function.
7492 * mem-stats.h (struct mem_usage): Likewise.
7493 * vec.c (struct vec_usage): Remove operator< and compare
7494 function. Can be simply inherited.
7496 2018-01-12 Martin Jambor <mjambor@suse.cz>
7499 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7500 * tree-ssa-math-opts.c: Include domwalk.h.
7501 (convert_mult_to_fma_1): New function.
7502 (fma_transformation_info): New type.
7503 (fma_deferring_state): Likewise.
7504 (cancel_fma_deferring): New function.
7505 (result_of_phi): Likewise.
7506 (last_fma_candidate_feeds_initial_phi): Likewise.
7507 (convert_mult_to_fma): Added deferring logic, split actual
7508 transformation to convert_mult_to_fma_1.
7509 (math_opts_dom_walker): New type.
7510 (math_opts_dom_walker::after_dom_children): New method, body moved
7511 here from pass_optimize_widening_mul::execute, added deferring logic
7513 (pass_optimize_widening_mul::execute): Moved most of code to
7514 math_opts_dom_walker::after_dom_children.
7515 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7516 * config/i386/i386.c (ix86_option_override_internal): Added
7517 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7519 2018-01-12 Richard Biener <rguenther@suse.de>
7522 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7523 inline instance vars.
7525 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7528 * config/rx/rx.c (rx_is_restricted_memory_address):
7531 2018-01-12 Richard Biener <rguenther@suse.de>
7533 PR tree-optimization/80846
7534 * target.def (split_reduction): New target hook.
7535 * targhooks.c (default_split_reduction): New function.
7536 * targhooks.h (default_split_reduction): Declare.
7537 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7538 target requests first reduce vectors by combining low and high
7540 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7541 (get_vectype_for_scalar_type_and_size): Export.
7542 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7543 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7544 * doc/tm.texi: Regenerate.
7545 * config/i386/i386.c (ix86_split_reduction): Implement
7546 TARGET_VECTORIZE_SPLIT_REDUCTION.
7548 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7551 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7552 in PIC mode except for TARGET_VXWORKS_RTP.
7553 * config/sparc/sparc.c: Include cfgrtl.h.
7554 (TARGET_INIT_PIC_REG): Define.
7555 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7556 (sparc_pic_register_p): New predicate.
7557 (sparc_legitimate_address_p): Use it.
7558 (sparc_legitimize_pic_address): Likewise.
7559 (sparc_delegitimize_address): Likewise.
7560 (sparc_mode_dependent_address_p): Likewise.
7561 (gen_load_pcrel_sym): Remove 4th parameter.
7562 (load_got_register): Adjust call to above. Remove obsolete stuff.
7563 (sparc_expand_prologue): Do not call load_got_register here.
7564 (sparc_flat_expand_prologue): Likewise.
7565 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7566 (sparc_use_pseudo_pic_reg): New function.
7567 (sparc_init_pic_reg): Likewise.
7568 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7569 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7571 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7573 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7574 Add item for branch_cost.
7576 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7578 PR rtl-optimization/83565
7579 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7580 not extend the result to a larger mode for rotate operations.
7581 (num_sign_bit_copies1): Likewise.
7583 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7586 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7588 Use values-Xc.o for -pedantic.
7589 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7591 2018-01-12 Martin Liska <mliska@suse.cz>
7594 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7596 (possible_polymorphic_call_targets): Use it.
7597 (ipa_devirt): Likewise.
7599 2018-01-12 Martin Liska <mliska@suse.cz>
7601 * profile-count.h (enum profile_quality): Use 0 as invalid
7602 enum value of profile_quality.
7604 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7606 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7607 -mext-string options.
7609 2018-01-12 Richard Biener <rguenther@suse.de>
7611 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7612 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7613 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7615 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7617 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7619 * configure.ac (--with-long-double-format): Add support for the
7620 configuration option to change the default long double format on
7622 * config.gcc (powerpc*-linux*-*): Likewise.
7623 * configure: Regenerate.
7624 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7625 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7626 used without modification.
7628 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7630 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7631 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7632 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7633 MISC_BUILTIN_SPEC_BARRIER.
7634 (rs6000_init_builtins): Likewise.
7635 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7637 (speculation_barrier): New define_insn.
7638 * doc/extend.texi: Document __builtin_speculation_barrier.
7640 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7643 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7644 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7645 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7647 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7648 integral modes instead of "ss" and "sd".
7649 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7650 vectors with 32-bit and 64-bit elements.
7651 (vecdupssescalarmodesuffix): New mode attribute.
7652 (vec_dup<mode>): Use it.
7654 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7657 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7658 frame if argument is passed on stack.
7660 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7663 * ree.c (combine_reaching_defs): Optimize also
7664 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7665 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7667 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7670 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7672 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7675 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7676 after they are computed.
7678 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7680 PR tree-optimization/83695
7681 * gimple-loop-linterchange.cc
7682 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7683 reset cached scev information after interchange.
7684 (pass_linterchange::execute): Remove call to scev_reset_htab.
7686 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7688 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7689 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7690 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7691 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7692 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7693 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7694 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7695 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7696 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7697 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7698 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7699 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7700 (V_lane_reg): Likewise.
7701 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7703 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7704 (vfmal_lane_low<mode>_intrinsic,
7705 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7706 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7707 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7708 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7709 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7710 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7712 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7714 * config/arm/arm-cpus.in (fp16fml): New feature.
7715 (ALL_SIMD): Add fp16fml.
7716 (armv8.2-a): Add fp16fml as an option.
7717 (armv8.3-a): Likewise.
7718 (armv8.4-a): Add fp16fml as part of fp16.
7719 * config/arm/arm.h (TARGET_FP16FML): Define.
7720 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7722 * config/arm/arm-modes.def (V2HF): Define.
7723 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7724 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7725 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7726 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7727 vfmsl_low, vfmsl_high): New set of builtins.
7728 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7729 (vfml_op): New code attribute.
7730 (VFMLHALVES): New int iterator.
7731 (VFML, VFMLSEL): New mode attributes.
7732 (V_reg): Define mapping for V2HF.
7733 (V_hi, V_lo): New mode attributes.
7734 (VF_constraint): Likewise.
7735 (vfml_half, vfml_half_selector): New int attributes.
7736 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7738 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7739 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7741 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7742 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7743 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7744 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7746 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7747 Document new effective target and option set.
7749 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7751 * config/arm/arm-cpus.in (armv8_4): New feature.
7752 (ARMv8_4a): New fgroup.
7753 (armv8.4-a): New arch.
7754 * config/arm/arm-tables.opt: Regenerate.
7755 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7756 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7757 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7758 Add matching rules for -march=armv8.4-a and extensions.
7759 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7761 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7764 * config/rx/rx.md (BW): New mode attribute.
7765 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7767 2018-01-11 Richard Biener <rguenther@suse.de>
7769 PR tree-optimization/83435
7770 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7771 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7772 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7774 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7775 Alan Hayward <alan.hayward@arm.com>
7776 David Sherwood <david.sherwood@arm.com>
7778 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7780 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7781 (aarch64_print_address_internal): Use it to check for a zero offset.
7783 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7784 Alan Hayward <alan.hayward@arm.com>
7785 David Sherwood <david.sherwood@arm.com>
7787 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7788 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7789 Return a poly_int64 rather than a HOST_WIDE_INT.
7790 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7791 rather than a HOST_WIDE_INT.
7792 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7793 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7794 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7795 final_offset from HOST_WIDE_INT to poly_int64.
7796 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7797 to_constant when getting the number of units in an Advanced SIMD
7799 (aarch64_builtin_vectorized_function): Check for a constant number
7801 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7803 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7804 attribute instead of GET_MODE_NUNITS.
7805 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7806 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7807 GET_MODE_SIZE for fixed-size registers.
7808 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7809 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7810 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7811 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7812 (aarch64_print_operand, aarch64_print_address_internal)
7813 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7814 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7815 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7816 Handle polynomial GET_MODE_SIZE.
7817 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7818 wider than SImode without modification.
7819 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7820 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7821 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7822 passing and returning SVE modes.
7823 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7824 rather than GEN_INT.
7825 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7826 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7827 (aarch64_allocate_and_probe_stack_space): Likewise.
7828 (aarch64_layout_frame): Cope with polynomial offsets.
7829 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7830 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7832 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7833 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7834 poly_int64 rather than a HOST_WIDE_INT.
7835 (aarch64_get_separate_components, aarch64_process_components)
7836 (aarch64_expand_prologue, aarch64_expand_epilogue)
7837 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7838 (aarch64_anchor_offset): New function, split out from...
7839 (aarch64_legitimize_address): ...here.
7840 (aarch64_builtin_vectorization_cost): Handle polynomial
7841 TYPE_VECTOR_SUBPARTS.
7842 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7844 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7845 number of elements from the PARALLEL rather than the mode.
7846 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7847 rather than GET_MODE_BITSIZE.
7848 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7849 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7850 (aarch64_expand_vec_perm_const_1): Handle polynomial
7851 d->perm.length () and d->perm elements.
7852 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7853 Apply to_constant to d->perm elements.
7854 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7855 polynomial CONST_VECTOR_NUNITS.
7856 (aarch64_move_pointer): Take amount as a poly_int64 rather
7858 (aarch64_progress_pointer): Avoid temporary variable.
7859 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7860 the mode attribute instead of GET_MODE.
7862 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7863 Alan Hayward <alan.hayward@arm.com>
7864 David Sherwood <david.sherwood@arm.com>
7866 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7867 x exists before using it.
7868 (aarch64_add_constant_internal): Rename to...
7869 (aarch64_add_offset_1): ...this. Replace regnum with separate
7870 src and dest rtxes. Handle the case in which they're different,
7871 including when the offset is zero. Replace scratchreg with an rtx.
7872 Use 2 additions if there is no spare register into which we can
7873 move a 16-bit constant.
7874 (aarch64_add_constant): Delete.
7875 (aarch64_add_offset): Replace reg with separate src and dest
7876 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7877 Use aarch64_add_offset_1.
7878 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7879 an rtx rather than an int. Take the delta as a poly_int64
7880 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7881 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7882 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7883 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7884 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7886 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7887 aarch64_add_constant.
7889 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7891 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7892 Use scalar_float_mode.
7894 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7896 * config/aarch64/aarch64-simd.md
7897 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7898 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7899 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7900 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7901 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7902 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7903 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7904 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7905 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7906 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7908 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7911 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7912 targ_options->x_arm_arch_string is non NULL.
7914 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7916 * config/aarch64/aarch64.h
7917 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7919 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7922 * expmed.c (emit_store_flag_force): Swap if const op0
7923 and change VOIDmode to mode of op0.
7925 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7927 PR rtl-optimization/83761
7928 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7929 than bytes to mode_for_size.
7931 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7934 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7935 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7938 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7941 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7942 when in layout mode.
7943 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7944 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7947 2018-01-10 Michael Collison <michael.collison@arm.com>
7949 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7950 * config/aarch64/aarch64-option-extension.def: Add
7951 AARCH64_OPT_EXTENSION of 'fp16fml'.
7952 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7953 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7954 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7955 * config/aarch64/constraints.md (Ui7): New constraint.
7956 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7957 (VFMLA_SEL_W): Ditto.
7960 (VFMLA16_LOW): New int iterator.
7961 (VFMLA16_HIGH): Ditto.
7962 (UNSPEC_FMLAL): New unspec.
7963 (UNSPEC_FMLSL): Ditto.
7964 (UNSPEC_FMLAL2): Ditto.
7965 (UNSPEC_FMLSL2): Ditto.
7966 (f16mac): New code attribute.
7967 * config/aarch64/aarch64-simd-builtins.def
7968 (aarch64_fmlal_lowv2sf): Ditto.
7969 (aarch64_fmlsl_lowv2sf): Ditto.
7970 (aarch64_fmlalq_lowv4sf): Ditto.
7971 (aarch64_fmlslq_lowv4sf): Ditto.
7972 (aarch64_fmlal_highv2sf): Ditto.
7973 (aarch64_fmlsl_highv2sf): Ditto.
7974 (aarch64_fmlalq_highv4sf): Ditto.
7975 (aarch64_fmlslq_highv4sf): Ditto.
7976 (aarch64_fmlal_lane_lowv2sf): Ditto.
7977 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7978 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7979 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7980 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7981 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7982 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7983 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7984 (aarch64_fmlal_lane_highv2sf): Ditto.
7985 (aarch64_fmlsl_lane_highv2sf): Ditto.
7986 (aarch64_fmlal_laneq_highv2sf): Ditto.
7987 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7988 (aarch64_fmlalq_lane_highv4sf): Ditto.
7989 (aarch64_fmlsl_lane_highv4sf): Ditto.
7990 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7991 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7992 * config/aarch64/aarch64-simd.md:
7993 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7994 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7995 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7996 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7997 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7998 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7999 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8000 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8001 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8002 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8003 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8004 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8005 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8006 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8007 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8008 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8009 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8010 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8011 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8012 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8013 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8014 (vfmlsl_low_u32): Ditto.
8015 (vfmlalq_low_u32): Ditto.
8016 (vfmlslq_low_u32): Ditto.
8017 (vfmlal_high_u32): Ditto.
8018 (vfmlsl_high_u32): Ditto.
8019 (vfmlalq_high_u32): Ditto.
8020 (vfmlslq_high_u32): Ditto.
8021 (vfmlal_lane_low_u32): Ditto.
8022 (vfmlsl_lane_low_u32): Ditto.
8023 (vfmlal_laneq_low_u32): Ditto.
8024 (vfmlsl_laneq_low_u32): Ditto.
8025 (vfmlalq_lane_low_u32): Ditto.
8026 (vfmlslq_lane_low_u32): Ditto.
8027 (vfmlalq_laneq_low_u32): Ditto.
8028 (vfmlslq_laneq_low_u32): Ditto.
8029 (vfmlal_lane_high_u32): Ditto.
8030 (vfmlsl_lane_high_u32): Ditto.
8031 (vfmlal_laneq_high_u32): Ditto.
8032 (vfmlsl_laneq_high_u32): Ditto.
8033 (vfmlalq_lane_high_u32): Ditto.
8034 (vfmlslq_lane_high_u32): Ditto.
8035 (vfmlalq_laneq_high_u32): Ditto.
8036 (vfmlslq_laneq_high_u32): Ditto.
8037 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8038 (AARCH64_FL_FOR_ARCH8_4): New.
8039 (AARCH64_ISA_F16FML): New ISA flag.
8040 (TARGET_F16FML): New feature flag for fp16fml.
8041 (doc/invoke.texi): Document new fp16fml option.
8043 2018-01-10 Michael Collison <michael.collison@arm.com>
8045 * config/aarch64/aarch64-builtins.c:
8046 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8047 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8048 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8049 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8050 (AARCH64_ISA_SHA3): New ISA flag.
8051 (TARGET_SHA3): New feature flag for sha3.
8052 * config/aarch64/iterators.md (sha512_op): New int attribute.
8053 (CRYPTO_SHA512): New int iterator.
8054 (UNSPEC_SHA512H): New unspec.
8055 (UNSPEC_SHA512H2): Ditto.
8056 (UNSPEC_SHA512SU0): Ditto.
8057 (UNSPEC_SHA512SU1): Ditto.
8058 * config/aarch64/aarch64-simd-builtins.def
8059 (aarch64_crypto_sha512hqv2di): New builtin.
8060 (aarch64_crypto_sha512h2qv2di): Ditto.
8061 (aarch64_crypto_sha512su0qv2di): Ditto.
8062 (aarch64_crypto_sha512su1qv2di): Ditto.
8063 (aarch64_eor3qv8hi): Ditto.
8064 (aarch64_rax1qv2di): Ditto.
8065 (aarch64_xarqv2di): Ditto.
8066 (aarch64_bcaxqv8hi): Ditto.
8067 * config/aarch64/aarch64-simd.md:
8068 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8069 (aarch64_crypto_sha512su0qv2di): Ditto.
8070 (aarch64_crypto_sha512su1qv2di): Ditto.
8071 (aarch64_eor3qv8hi): Ditto.
8072 (aarch64_rax1qv2di): Ditto.
8073 (aarch64_xarqv2di): Ditto.
8074 (aarch64_bcaxqv8hi): Ditto.
8075 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8076 (vsha512h2q_u64): Ditto.
8077 (vsha512su0q_u64): Ditto.
8078 (vsha512su1q_u64): Ditto.
8079 (veor3q_u16): Ditto.
8080 (vrax1q_u64): Ditto.
8082 (vbcaxq_u16): Ditto.
8083 * config/arm/types.md (crypto_sha512): New type attribute.
8084 (crypto_sha3): Ditto.
8085 (doc/invoke.texi): Document new sha3 option.
8087 2018-01-10 Michael Collison <michael.collison@arm.com>
8089 * config/aarch64/aarch64-builtins.c:
8090 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8091 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8092 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8093 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8094 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8095 (AARCH64_ISA_SM4): New ISA flag.
8096 (TARGET_SM4): New feature flag for sm4.
8097 * config/aarch64/aarch64-simd-builtins.def
8098 (aarch64_sm3ss1qv4si): Ditto.
8099 (aarch64_sm3tt1aq4si): Ditto.
8100 (aarch64_sm3tt1bq4si): Ditto.
8101 (aarch64_sm3tt2aq4si): Ditto.
8102 (aarch64_sm3tt2bq4si): Ditto.
8103 (aarch64_sm3partw1qv4si): Ditto.
8104 (aarch64_sm3partw2qv4si): Ditto.
8105 (aarch64_sm4eqv4si): Ditto.
8106 (aarch64_sm4ekeyqv4si): Ditto.
8107 * config/aarch64/aarch64-simd.md:
8108 (aarch64_sm3ss1qv4si): Ditto.
8109 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8110 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8111 (aarch64_sm4eqv4si): Ditto.
8112 (aarch64_sm4ekeyqv4si): Ditto.
8113 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8114 (sm3part_op): Ditto.
8115 (CRYPTO_SM3TT): Ditto.
8116 (CRYPTO_SM3PART): Ditto.
8117 (UNSPEC_SM3SS1): New unspec.
8118 (UNSPEC_SM3TT1A): Ditto.
8119 (UNSPEC_SM3TT1B): Ditto.
8120 (UNSPEC_SM3TT2A): Ditto.
8121 (UNSPEC_SM3TT2B): Ditto.
8122 (UNSPEC_SM3PARTW1): Ditto.
8123 (UNSPEC_SM3PARTW2): Ditto.
8124 (UNSPEC_SM4E): Ditto.
8125 (UNSPEC_SM4EKEY): Ditto.
8126 * config/aarch64/constraints.md (Ui2): New constraint.
8127 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8128 * config/arm/types.md (crypto_sm3): New type attribute.
8129 (crypto_sm4): Ditto.
8130 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8131 (vsm3tt1aq_u32): Ditto.
8132 (vsm3tt1bq_u32): Ditto.
8133 (vsm3tt2aq_u32): Ditto.
8134 (vsm3tt2bq_u32): Ditto.
8135 (vsm3partw1q_u32): Ditto.
8136 (vsm3partw2q_u32): Ditto.
8137 (vsm4eq_u32): Ditto.
8138 (vsm4ekeyq_u32): Ditto.
8139 (doc/invoke.texi): Document new sm4 option.
8141 2018-01-10 Michael Collison <michael.collison@arm.com>
8143 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8144 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8145 (AARCH64_FL_FOR_ARCH8_4): New.
8146 (AARCH64_FL_V8_4): New flag.
8147 (doc/invoke.texi): Document new armv8.4-a option.
8149 2018-01-10 Michael Collison <michael.collison@arm.com>
8151 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8152 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8153 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8154 * config/aarch64/aarch64-option-extension.def: Add
8155 AARCH64_OPT_EXTENSION of 'sha2'.
8156 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8157 (crypto): Disable sha2 and aes if crypto disabled.
8158 (crypto): Enable aes and sha2 if enabled.
8159 (simd): Disable sha2 and aes if simd disabled.
8160 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8162 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8163 (TARGET_SHA2): New feature flag for sha2.
8164 (TARGET_AES): New feature flag for aes.
8165 * config/aarch64/aarch64-simd.md:
8166 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8167 conditional on TARGET_AES.
8168 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8169 (aarch64_crypto_sha1hsi): Make pattern conditional
8171 (aarch64_crypto_sha1hv4si): Ditto.
8172 (aarch64_be_crypto_sha1hv4si): Ditto.
8173 (aarch64_crypto_sha1su1v4si): Ditto.
8174 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8175 (aarch64_crypto_sha1su0v4si): Ditto.
8176 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8177 (aarch64_crypto_sha256su0v4si): Ditto.
8178 (aarch64_crypto_sha256su1v4si): Ditto.
8179 (doc/invoke.texi): Document new aes and sha2 options.
8181 2018-01-10 Martin Sebor <msebor@redhat.com>
8183 PR tree-optimization/83781
8184 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8187 2018-01-11 Martin Sebor <msebor@gmail.com>
8188 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8190 PR tree-optimization/83501
8191 PR tree-optimization/81703
8193 * tree-ssa-strlen.c (get_string_cst): Rename...
8194 (get_string_len): ...to this. Handle global constants.
8195 (handle_char_store): Adjust.
8197 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8198 Jim Wilson <jimw@sifive.com>
8200 * config/riscv/riscv-protos.h (riscv_output_return): New.
8201 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8202 (riscv_attribute_table, riscv_output_return),
8203 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8204 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8205 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8207 (riscv_expand_prologue): Add early return for naked function.
8208 (riscv_expand_epilogue): Likewise.
8209 (riscv_function_ok_for_sibcall): Return false for naked function.
8210 (riscv_set_current_function): New.
8211 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8212 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8213 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8214 * doc/extend.texi (RISC-V Function Attributes): New.
8216 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8218 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8219 check for 128-bit long double before checking TCmode.
8220 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8221 128-bit long doubles before checking TFmode or TCmode.
8222 (FLOAT128_IBM_P): Likewise.
8224 2018-01-10 Martin Sebor <msebor@redhat.com>
8226 PR tree-optimization/83671
8227 * builtins.c (c_strlen): Unconditionally return zero for the empty
8229 Use -Warray-bounds for warnings.
8230 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8231 for non-constant array indices with COMPONENT_REF, arrays of
8232 arrays, and pointers to arrays.
8233 (gimple_fold_builtin_strlen): Determine and set length range for
8234 non-constant character arrays.
8236 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8239 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8242 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8244 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8246 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8249 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8250 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8251 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8252 indexed_or_indirect_operand predicate.
8253 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8254 (*vsx_le_perm_load_v8hi): Likewise.
8255 (*vsx_le_perm_load_v16qi): Likewise.
8256 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8257 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8258 (*vsx_le_perm_store_v8hi): Likewise.
8259 (*vsx_le_perm_store_v16qi): Likewise.
8260 (eight unnamed splitters): Likewise.
8262 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8264 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8265 * config/rs6000/emmintrin.h: Likewise.
8266 * config/rs6000/mmintrin.h: Likewise.
8267 * config/rs6000/xmmintrin.h: Likewise.
8269 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8272 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8274 * tree.c (tree_nop_conversion): Return true for location wrapper
8276 (maybe_wrap_with_location): New function.
8277 (selftest::check_strip_nops): New function.
8278 (selftest::test_location_wrappers): New function.
8279 (selftest::tree_c_tests): Call it.
8280 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8281 (maybe_wrap_with_location): New decl.
8282 (EXPR_LOCATION_WRAPPER_P): New macro.
8283 (location_wrapper_p): New inline function.
8284 (tree_strip_any_location_wrapper): New inline function.
8286 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8289 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8290 stack_realign_offset for the largest alignment of stack slot
8292 (ix86_find_max_used_stack_alignment): New function.
8293 (ix86_finalize_stack_frame_flags): Use it. Set
8294 max_used_stack_alignment if we don't realign stack.
8295 * config/i386/i386.h (machine_function): Add
8296 max_used_stack_alignment.
8298 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8300 * config/arm/arm.opt (-mbranch-cost): New option.
8301 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8304 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8307 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8308 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8310 2018-01-10 Richard Biener <rguenther@suse.de>
8313 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8314 early out so it also covers the case where we have a non-NULL
8317 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8319 PR tree-optimization/83753
8320 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8321 for non-strided grouped accesses if the number of elements is 1.
8323 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8326 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8327 * i386.h (TARGET_USE_GATHER): Define.
8328 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8330 2018-01-10 Martin Liska <mliska@suse.cz>
8333 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8334 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8336 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8337 CLEANUP_NO_PARTITIONING is not set.
8339 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8341 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8342 for vectors, as a partial revert of r254296.
8343 * rtl.h (const_vec_p): Delete.
8344 (const_vec_duplicate_p): Don't test for vector CONSTs.
8345 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8346 * expmed.c (make_tree): Likewise.
8349 * common.md (E, F): Use CONSTANT_P instead of checking for
8351 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8352 checking for CONST_VECTOR.
8354 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8357 * predict.c (force_edge_cold): Handle in more sane way edges
8360 2018-01-09 Carl Love <cel@us.ibm.com>
8362 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8364 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8365 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8366 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8367 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8368 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8369 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8370 * config/rs6000/rs6000-protos.h: Add extern defition for
8371 rs6000_generate_float2_double_code.
8372 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8374 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8375 (float2_v2df): Add define_expand.
8377 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8380 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8381 op_mode in the force_to_mode call.
8383 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8385 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8386 instead of checking each element individually.
8387 (aarch64_evpc_uzp): Likewise.
8388 (aarch64_evpc_zip): Likewise.
8389 (aarch64_evpc_ext): Likewise.
8390 (aarch64_evpc_rev): Likewise.
8391 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8392 instead of checking each element individually. Return true without
8394 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8395 whether all selected elements come from the same input, instead of
8396 checking each element individually. Remove calls to gen_rtx_REG,
8397 start_sequence and end_sequence and instead assert that no rtl is
8400 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8402 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8403 order of HIGH and CONST checks.
8405 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8407 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8408 if the destination isn't an SSA_NAME.
8410 2018-01-09 Richard Biener <rguenther@suse.de>
8412 PR tree-optimization/83668
8413 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8415 (canonicalize_loop_form): ... here, renamed from ...
8416 (canonicalize_loop_closed_ssa_form): ... this and amended to
8417 swap successor edges for loop exit blocks to make us use
8418 the RPO order we need for initial schedule generation.
8420 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8422 PR tree-optimization/64811
8423 * match.pd: When optimizing comparisons with Inf, avoid
8424 introducing or losing exceptions from comparisons with NaN.
8426 2018-01-09 Martin Liska <mliska@suse.cz>
8429 * asan.c (shadow_mem_size): Add gcc_assert.
8431 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8433 Don't save registers in main().
8436 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8437 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8438 * config/avr/avr.c (avr_set_current_function): Don't error if
8439 naked, OS_task or OS_main are specified at the same time.
8440 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8442 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8444 * common/config/avr/avr-common.c (avr_option_optimization_table):
8445 Switch on -mmain-is-OS_task for optimizing compilations.
8447 2018-01-09 Richard Biener <rguenther@suse.de>
8449 PR tree-optimization/83572
8450 * graphite.c: Include cfganal.h.
8451 (graphite_transform_loops): Connect infinite loops to exit
8452 and remove fake edges at the end.
8454 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8456 * ipa-inline.c (edge_badness): Revert accidental checkin.
8458 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8461 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8462 symbols; not inline clones.
8464 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8467 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8468 hard registers. Formatting fixes.
8470 PR preprocessor/83722
8471 * gcc.c (try_generate_repro): Pass
8472 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8473 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8476 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8477 Kito Cheng <kito.cheng@gmail.com>
8479 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8480 (riscv_leaf_function_p): Delete.
8481 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8483 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8485 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8487 (do_ifelse): New function.
8488 (do_isel): New function.
8489 (do_sub3): New function.
8490 (do_add3): New function.
8491 (do_load_mask_compare): New function.
8492 (do_overlap_load_compare): New function.
8493 (expand_compare_loop): New function.
8494 (expand_block_compare): Call expand_compare_loop() when appropriate.
8495 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8497 (-mblock-compare-inline-loop-limit): New option.
8499 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8502 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8503 Reverse order of second and third operands in first alternative.
8504 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8505 of first and second elements in UNSPEC_VPERMR vector.
8506 (altivec_expand_vec_perm_le): Likewise.
8508 2018-01-08 Jeff Law <law@redhat.com>
8510 PR rtl-optimizatin/81308
8511 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8512 (process_switch): If group_case_labels makes a change, then set
8514 (pass_convert_switch::execute): If a switch is converted, then
8515 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8517 PR rtl-optimization/81308
8518 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8521 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8523 PR target/83663 - Revert r255946
8524 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8525 generation for cases where splatting a value is not useful.
8526 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8527 across a vec_duplicate and a paradoxical subreg forming a vector
8528 mode to a vec_concat.
8530 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8532 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8533 -march=armv8.3-a variants.
8534 * config/arm/t-multilib: Likewise.
8535 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8537 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8539 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8541 (cceq_ior_compare_complement): Give it a name so I can use it, and
8542 change boolean_or_operator predicate to boolean_operator so it can
8543 be used to generate a crand.
8544 (eqne): New code iterator.
8545 (bd/bd_neg): New code_attrs.
8546 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8547 a single define_insn.
8548 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8549 decrement (bdnzt/bdnzf/bdzt/bdzf).
8550 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8551 with the new names of the branch decrement patterns, and added the
8552 names of the branch decrement conditional patterns.
8554 2018-01-08 Richard Biener <rguenther@suse.de>
8556 PR tree-optimization/83563
8557 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8560 2018-01-08 Richard Biener <rguenther@suse.de>
8563 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8565 2018-01-08 Richard Biener <rguenther@suse.de>
8567 PR tree-optimization/83685
8568 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8569 references to abnormals.
8571 2018-01-08 Richard Biener <rguenther@suse.de>
8574 * dwarf2out.c (output_indirect_strings): Handle empty
8575 skeleton_debug_str_hash.
8576 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8578 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8580 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8581 (emit_store_direct): Likewise.
8582 (arc_trampoline_adjust_address): Likewise.
8583 (arc_asm_trampoline_template): New function.
8584 (arc_initialize_trampoline): Use asm_trampoline_template.
8585 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8586 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8587 * config/arc/arc.md (flush_icache): Delete pattern.
8589 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8591 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8592 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8595 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8598 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8599 by not USED_FOR_TARGET.
8600 (make_pass_resolve_sw_modes): Likewise.
8602 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8604 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8607 2018-01-08 Richard Biener <rguenther@suse.de>
8610 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8612 2018-01-08 Richard Biener <rguenther@suse.de>
8615 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8617 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8620 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8621 basic blocks with a small number of successors.
8622 (convert_control_dep_chain_into_preds): Improve handling of
8624 (dump_predicates): Split apart into...
8625 (dump_pred_chain): ...here...
8626 (dump_pred_info): ...and here.
8627 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8628 (can_chain_union_be_invalidated_p): Improve check for invalidation
8630 (uninit_uses_cannot_happen): Avoid unnecessary if
8631 convert_control_dep_chain_into_preds yielded nothing.
8633 2018-01-06 Martin Sebor <msebor@redhat.com>
8635 PR tree-optimization/83640
8636 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8637 subtracting negative offset from size.
8638 (builtin_access::overlap): Adjust offset bounds of the access to fall
8639 within the size of the object if possible.
8641 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8643 PR rtl-optimization/83699
8644 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8645 extract_bit_field_as_subreg to cases in which the extracted
8646 value is also a vector.
8648 * lra-constraints.c (process_alt_operands): Test for the equivalence
8649 substitutions when detecting a possible reload cycle.
8651 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8654 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8655 by default if flag_selective_schedling{,2}. Formatting fixes.
8657 PR rtl-optimization/83682
8658 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8659 if it has non-VECTOR_MODE element mode.
8660 (vec_duplicate_p): Likewise.
8663 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8664 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8666 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8669 * config/i386/i386-builtin.def
8670 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8671 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8672 Require also OPTION_MASK_ISA_AVX512F in addition to
8673 OPTION_MASK_ISA_GFNI.
8674 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8675 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8676 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8677 to OPTION_MASK_ISA_GFNI.
8678 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8679 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8680 OPTION_MASK_ISA_AVX512BW.
8681 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8682 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8683 addition to OPTION_MASK_ISA_GFNI.
8684 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8685 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8686 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8687 to OPTION_MASK_ISA_GFNI.
8688 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8689 a requirement for all ISAs rather than any of them with a few
8691 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8693 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8694 bitmasks to be enabled with 3 exceptions, instead of requiring any
8695 enabled ISA with lots of exceptions.
8696 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8697 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8698 Change avx512bw in isa attribute to avx512f.
8699 * config/i386/sgxintrin.h: Add license boilerplate.
8700 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8701 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8702 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8703 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8705 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8706 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8707 temporarily sse2 rather than sse if not enabled already.
8710 * config/i386/sse.md (VI248_VLBW): Rename to ...
8711 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8712 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8713 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8714 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8715 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8716 mode iterator instead of VI248_VLBW.
8718 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8720 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8721 (record_modified): Skip clobbers; add debug output.
8722 (param_change_prob): Use sreal frequencies.
8724 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8726 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8727 punt for user-aligned variables.
8729 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8731 * tree-chrec.c (chrec_contains_symbols): Return true for
8734 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8737 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8738 of (x|y) == x for BICS pattern.
8740 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8742 PR tree-optimization/83605
8743 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8744 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8747 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8749 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8750 * config/epiphany/rtems.h: New file.
8752 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8753 Uros Bizjak <ubizjak@gmail.com>
8756 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8757 QIreg_operand instead of register_operand predicate.
8758 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8759 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8760 comments instead of -fmitigate[-_]rop.
8762 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8765 * cgraphunit.c (symbol_table::compile): Switch to text_section
8766 before calling assembly_start debug hook.
8767 * run-rtl-passes.c (run_rtl_passes): Likewise.
8770 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8772 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8773 range_int_cst_p rather than !symbolic_range_p before calling
8774 extract_range_from_multiplicative_op_1.
8776 2018-01-04 Jeff Law <law@redhat.com>
8778 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8779 redundant test in assertion.
8781 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8783 * doc/rtl.texi: Document machine_mode wrapper classes.
8785 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8787 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8790 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8792 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8793 the VEC_PERM_EXPR fold to fail.
8795 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8798 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8799 to switched_sections.
8801 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8804 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8807 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8810 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8811 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8813 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8816 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8817 is BLKmode and bitpos not zero or mode change is needed.
8819 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8822 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8825 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8828 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8829 instead of MULT rtx. Update all corresponding splitters.
8831 (*ssub<modesuffix>): Ditto.
8833 (*cmp_sadd_di): Update split patterns.
8834 (*cmp_sadd_si): Ditto.
8835 (*cmp_sadd_sidi): Ditto.
8836 (*cmp_ssub_di): Ditto.
8837 (*cmp_ssub_si): Ditto.
8838 (*cmp_ssub_sidi): Ditto.
8839 * config/alpha/predicates.md (const23_operand): New predicate.
8840 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8841 Look for ASHIFT, not MULT inner operand.
8842 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8844 2018-01-04 Martin Liska <mliska@suse.cz>
8846 PR gcov-profile/83669
8847 * gcov.c (output_intermediate_file): Add version to intermediate
8849 * doc/gcov.texi: Document new field 'version' in intermediate
8850 file format. Fix location of '-k' option of gcov command.
8852 2018-01-04 Martin Liska <mliska@suse.cz>
8855 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8857 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8859 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8861 2018-01-03 Martin Sebor <msebor@redhat.com>
8863 PR tree-optimization/83655
8864 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8865 checking calls with invalid arguments.
8867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8869 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8870 (vectorizable_mask_load_store): Delete.
8871 (vectorizable_call): Return false for masked loads and stores.
8872 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8873 instead of gimple_assign_rhs1.
8874 (vectorizable_load): Handle IFN_MASK_LOAD.
8875 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8879 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8881 (vectorizable_mask_load_store): ...here.
8882 (vectorizable_load): ...and here.
8884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8886 * tree-vect-stmts.c (vect_build_all_ones_mask)
8887 (vect_build_zero_merge_argument): New functions, split out from...
8888 (vectorizable_load): ...here.
8890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8892 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8894 (vectorizable_mask_load_store): ...here.
8895 (vectorizable_store): ...and here.
8897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8899 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8901 (vectorizable_mask_load_store): ...here.
8903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8905 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8906 (vect_model_store_cost): Take a vec_load_store_type instead of a
8908 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8909 (vect_model_store_cost): Take a vec_load_store_type instead of a
8911 (vectorizable_mask_load_store): Update accordingly.
8912 (vectorizable_store): Likewise.
8913 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8917 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8918 IFN_MASK_LOAD calls here rather than...
8919 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8922 Alan Hayward <alan.hayward@arm.com>
8923 David Sherwood <david.sherwood@arm.com>
8925 * expmed.c (extract_bit_field_1): For vector extracts,
8926 fall back to extract_bit_field_as_subreg if vec_extract
8929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8930 Alan Hayward <alan.hayward@arm.com>
8931 David Sherwood <david.sherwood@arm.com>
8933 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8934 they are variable or constant sized.
8935 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8936 slots for constant-sized data.
8938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8939 Alan Hayward <alan.hayward@arm.com>
8940 David Sherwood <david.sherwood@arm.com>
8942 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8943 handling COND_EXPRs with boolean comparisons, try to find a better
8944 basis for the mask type than the boolean itself.
8946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8948 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8949 is calculated and how it can be overridden.
8950 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8951 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8953 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8957 Alan Hayward <alan.hayward@arm.com>
8958 David Sherwood <david.sherwood@arm.com>
8960 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8961 Remove the mode argument.
8962 (aarch64_simd_valid_immediate): Remove the mode and inverse
8964 * config/aarch64/iterators.md (bitsize): New iterator.
8965 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8966 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8967 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8968 aarch64_simd_valid_immediate.
8969 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8970 (aarch64_reg_or_bic_imm): Likewise.
8971 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8972 with an insn_type enum and msl with a modifier_type enum.
8973 Replace element_width with a scalar_mode. Change the shift
8974 to unsigned int. Add constructors for scalar_float_mode and
8975 scalar_int_mode elements.
8976 (aarch64_vect_float_const_representable_p): Delete.
8977 (aarch64_can_const_movi_rtx_p)
8978 (aarch64_simd_scalar_immediate_valid_for_move)
8979 (aarch64_simd_make_constant): Update call to
8980 aarch64_simd_valid_immediate.
8981 (aarch64_advsimd_valid_immediate_hs): New function.
8982 (aarch64_advsimd_valid_immediate): Likewise.
8983 (aarch64_simd_valid_immediate): Remove mode and inverse
8984 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8985 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8986 and aarch64_float_const_representable_p on the result.
8987 (aarch64_output_simd_mov_immediate): Remove mode argument.
8988 Update call to aarch64_simd_valid_immediate and use of
8989 simd_immediate_info.
8990 (aarch64_output_scalar_simd_mov_immediate): Update call
8993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8994 Alan Hayward <alan.hayward@arm.com>
8995 David Sherwood <david.sherwood@arm.com>
8997 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8998 (mode_nunits): Likewise CONST_MODE_NUNITS.
8999 * machmode.def (ADJUST_NUNITS): Document.
9000 * genmodes.c (mode_data::need_nunits_adj): New field.
9001 (blank_mode): Update accordingly.
9002 (adj_nunits): New variable.
9003 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9005 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9006 listed in adj_nunits.
9007 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9008 listed in adj_nunits. Don't emit case statements for such modes.
9009 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9010 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9011 nothing if adj_nunits is nonnull.
9012 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9013 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9014 (emit_mode_fbit): Update use of print_maybe_const_decl.
9015 (emit_move_size): Likewise. Treat the array as non-const
9017 (emit_mode_adjustments): Handle adj_nunits.
9019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9021 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9022 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9023 (VECTOR_MODES): Use it.
9024 (make_vector_modes): Take the prefix as an argument.
9026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9027 Alan Hayward <alan.hayward@arm.com>
9028 David Sherwood <david.sherwood@arm.com>
9030 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9031 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9032 for MODE_VECTOR_BOOL.
9033 * machmode.def (VECTOR_BOOL_MODE): Document.
9034 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9035 (make_vector_bool_mode): New function.
9036 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9038 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9039 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9041 * stor-layout.c (int_mode_for_mode): Likewise.
9042 * tree.c (build_vector_type_for_mode): Likewise.
9043 * varasm.c (output_constant_pool_2): Likewise.
9044 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9045 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9046 for MODE_VECTOR_BOOL.
9047 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9048 of mode class checks.
9049 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9050 instead of a list of mode class checks.
9051 (expand_vector_scalar_condition): Likewise.
9052 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9055 Alan Hayward <alan.hayward@arm.com>
9056 David Sherwood <david.sherwood@arm.com>
9058 * machmode.h (mode_size): Change from unsigned short to
9060 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9061 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9062 or if measurement_type is not polynomial.
9063 (fixed_size_mode::includes_p): Check for constant-sized modes.
9064 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9065 return a poly_uint16 rather than an unsigned short.
9066 (emit_mode_size): Change the type of mode_size from unsigned short
9067 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9068 (emit_mode_adjustments): Cope with polynomial vector sizes.
9069 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9071 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9073 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9074 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9075 * caller-save.c (setup_save_areas): Likewise.
9076 (replace_reg_with_saved_mem): Likewise.
9077 * calls.c (emit_library_call_value_1): Likewise.
9078 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9079 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9080 (gen_lowpart_for_combine): Likewise.
9081 * convert.c (convert_to_integer_1): Likewise.
9082 * cse.c (equiv_constant, cse_insn): Likewise.
9083 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9084 (cselib_subst_to_values): Likewise.
9085 * dce.c (word_dce_process_block): Likewise.
9086 * df-problems.c (df_word_lr_mark_ref): Likewise.
9087 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9088 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9089 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9090 (rtl_for_decl_location): Likewise.
9091 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9092 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9093 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9094 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9095 (expand_expr_real_1): Likewise.
9096 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9097 (pad_below): Likewise.
9098 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9099 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9100 * ira.c (get_subreg_tracking_sizes): Likewise.
9101 * ira-build.c (ira_create_allocno_objects): Likewise.
9102 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9103 (ira_sort_regnos_for_alter_reg): Likewise.
9104 * ira-costs.c (record_operand_costs): Likewise.
9105 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9106 (resolve_simple_move): Likewise.
9107 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9108 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9109 (lra_constraints): Likewise.
9110 (CONST_POOL_OK_P): Reject variable-sized modes.
9111 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9112 (add_pseudo_to_slot, lra_spill): Likewise.
9113 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9114 * optabs-query.c (get_best_extraction_insn): Likewise.
9115 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9116 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9117 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9118 * recog.c (offsettable_address_addr_space_p): Likewise.
9119 * regcprop.c (maybe_mode_change): Likewise.
9120 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9121 * regrename.c (build_def_use): Likewise.
9122 * regstat.c (dump_reg_info): Likewise.
9123 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9124 (find_reloads, find_reloads_subreg_address): Likewise.
9125 * reload1.c (eliminate_regs_1): Likewise.
9126 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9127 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9128 (simplify_binary_operation_1, simplify_subreg): Likewise.
9129 * targhooks.c (default_function_arg_padding): Likewise.
9130 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9131 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9132 (verify_gimple_assign_ternary): Likewise.
9133 * tree-inline.c (estimate_move_cost): Likewise.
9134 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9135 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9136 (get_address_cost_ainc): Likewise.
9137 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9138 (vect_supportable_dr_alignment): Likewise.
9139 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9140 (vectorizable_reduction): Likewise.
9141 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9142 (vectorizable_operation, vectorizable_load): Likewise.
9143 * tree.c (build_same_sized_truth_vector_type): Likewise.
9144 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9145 * var-tracking.c (emit_note_insn_var_location): Likewise.
9146 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9147 (ADDR_VEC_ALIGN): Likewise.
9149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9150 Alan Hayward <alan.hayward@arm.com>
9151 David Sherwood <david.sherwood@arm.com>
9153 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9155 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9156 or if measurement_type is polynomial.
9157 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9158 * combine.c (make_extraction): Likewise.
9159 * dse.c (find_shift_sequence): Likewise.
9160 * dwarf2out.c (mem_loc_descriptor): Likewise.
9161 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9162 (extract_bit_field, extract_low_bits): Likewise.
9163 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9164 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9165 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9166 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9167 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9168 * reload.c (find_reloads): Likewise.
9169 * reload1.c (alter_reg): Likewise.
9170 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9171 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9172 * tree-if-conv.c (predicate_mem_writes): Likewise.
9173 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9174 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9175 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9176 * valtrack.c (dead_debug_insert_temp): Likewise.
9177 * varasm.c (mergeable_constant_section): Likewise.
9178 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9181 Alan Hayward <alan.hayward@arm.com>
9182 David Sherwood <david.sherwood@arm.com>
9184 * expr.c (expand_assignment): Cope with polynomial mode sizes
9185 when assigning to a CONCAT.
9187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9188 Alan Hayward <alan.hayward@arm.com>
9189 David Sherwood <david.sherwood@arm.com>
9191 * machmode.h (mode_precision): Change from unsigned short to
9193 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9195 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9196 or if measurement_type is not polynomial.
9197 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9198 in which the mode is already known to be a scalar_int_mode.
9199 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9200 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9202 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9203 for GET_MODE_PRECISION.
9204 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9205 for GET_MODE_PRECISION.
9206 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9208 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9209 (expand_field_assignment, make_extraction): Likewise.
9210 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9211 (get_last_value): Likewise.
9212 * convert.c (convert_to_integer_1): Likewise.
9213 * cse.c (cse_insn): Likewise.
9214 * expr.c (expand_expr_real_1): Likewise.
9215 * lra-constraints.c (simplify_operand_subreg): Likewise.
9216 * optabs-query.c (can_atomic_load_p): Likewise.
9217 * optabs.c (expand_atomic_load): Likewise.
9218 (expand_atomic_store): Likewise.
9219 * ree.c (combine_reaching_defs): Likewise.
9220 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9221 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9222 * tree.h (type_has_mode_precision_p): Likewise.
9223 * ubsan.c (instrument_si_overflow): Likewise.
9225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9226 Alan Hayward <alan.hayward@arm.com>
9227 David Sherwood <david.sherwood@arm.com>
9229 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9230 polynomial numbers of units.
9231 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9232 (valid_vector_subparts_p): New function.
9233 (build_vector_type): Remove temporary shim and take the number
9234 of units as a poly_uint64 rather than an int.
9235 (build_opaque_vector_type): Take the number of units as a
9236 poly_uint64 rather than an int.
9237 * tree.c (build_vector_from_ctor): Handle polynomial
9238 TYPE_VECTOR_SUBPARTS.
9239 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9240 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9241 (build_vector_from_val): If the number of units is variable,
9242 use build_vec_duplicate_cst for constant operands and
9243 VEC_DUPLICATE_EXPR otherwise.
9244 (make_vector_type): Remove temporary is_constant ().
9245 (build_vector_type, build_opaque_vector_type): Take the number of
9246 units as a poly_uint64 rather than an int.
9247 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9249 * cfgexpand.c (expand_debug_expr): Likewise.
9250 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9251 (store_constructor, expand_expr_real_1): Likewise.
9252 (const_scalar_mask_from_tree): Likewise.
9253 * fold-const-call.c (fold_const_reduction): Likewise.
9254 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9255 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9256 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9257 (fold_relational_const): Likewise.
9258 (native_interpret_vector): Likewise. Change the size from an
9259 int to an unsigned int.
9260 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9261 TYPE_VECTOR_SUBPARTS.
9262 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9263 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9264 duplicating a non-constant operand into a variable-length vector.
9265 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9266 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9267 * ipa-icf.c (sem_variable::equals): Likewise.
9268 * match.pd: Likewise.
9269 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9270 * print-tree.c (print_node): Likewise.
9271 * stor-layout.c (layout_type): Likewise.
9272 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9273 * tree-cfg.c (verify_gimple_comparison): Likewise.
9274 (verify_gimple_assign_binary): Likewise.
9275 (verify_gimple_assign_ternary): Likewise.
9276 (verify_gimple_assign_single): Likewise.
9277 * tree-pretty-print.c (dump_generic_node): Likewise.
9278 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9279 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9280 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9281 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9282 (vect_shift_permute_load_chain): Likewise.
9283 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9284 (expand_vector_condition, optimize_vector_constructor): Likewise.
9285 (lower_vec_perm, get_compute_type): Likewise.
9286 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9287 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9288 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9289 (vect_recog_mask_conversion_pattern): Likewise.
9290 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9291 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9292 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9293 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9294 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9295 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9296 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9297 (supportable_widening_operation): Likewise.
9298 (supportable_narrowing_operation): Likewise.
9299 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9301 * varasm.c (output_constant): Likewise.
9303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9304 Alan Hayward <alan.hayward@arm.com>
9305 David Sherwood <david.sherwood@arm.com>
9307 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9308 so that both the length == 3 and length != 3 cases set up their
9309 own permute vectors. Add comments explaining why we know the
9310 number of elements is constant.
9311 (vect_permute_load_chain): Likewise.
9313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9314 Alan Hayward <alan.hayward@arm.com>
9315 David Sherwood <david.sherwood@arm.com>
9317 * machmode.h (mode_nunits): Change from unsigned char to
9319 (ONLY_FIXED_SIZE_MODES): New macro.
9320 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9321 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9322 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9324 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9325 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9326 or if measurement_type is not polynomial.
9327 * genmodes.c (ZERO_COEFFS): New macro.
9328 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9330 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9331 Use ZERO_COEFFS when emitting initializers.
9332 * data-streamer.h (bp_pack_poly_value): New function.
9333 (bp_unpack_poly_value): Likewise.
9334 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9335 for GET_MODE_NUNITS.
9336 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9337 for GET_MODE_NUNITS.
9338 * tree.c (make_vector_type): Remove temporary shim and make
9339 the real function take the number of units as a poly_uint64
9341 (build_vector_type_for_mode): Handle polynomial nunits.
9342 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9343 * emit-rtl.c (const_vec_series_p_1): Likewise.
9344 (gen_rtx_CONST_VECTOR): Likewise.
9345 * fold-const.c (test_vec_duplicate_folding): Likewise.
9346 * genrecog.c (validate_pattern): Likewise.
9347 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9348 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9349 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9350 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9351 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9352 * rtlanal.c (subreg_get_info): Likewise.
9353 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9354 (vect_grouped_load_supported): Likewise.
9355 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9356 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9357 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9358 (simplify_const_unary_operation, simplify_binary_operation_1)
9359 (simplify_const_binary_operation, simplify_ternary_operation)
9360 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9361 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9362 instead of CONST_VECTOR_NUNITS.
9363 * varasm.c (output_constant_pool_2): Likewise.
9364 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9365 explicit-encoded elements in the XVEC for variable-length vectors.
9367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9369 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9372 Alan Hayward <alan.hayward@arm.com>
9373 David Sherwood <david.sherwood@arm.com>
9375 * coretypes.h (fixed_size_mode): Declare.
9376 (fixed_size_mode_pod): New typedef.
9377 * builtins.h (target_builtins::x_apply_args_mode)
9378 (target_builtins::x_apply_result_mode): Change type to
9379 fixed_size_mode_pod.
9380 * builtins.c (apply_args_size, apply_result_size, result_vector)
9381 (expand_builtin_apply_args_1, expand_builtin_apply)
9382 (expand_builtin_return): Update accordingly.
9384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9386 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9387 * cselib.c (cselib_hash_rtx): Likewise.
9388 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9389 CONST_VECTOR encoding.
9391 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9392 Jeff Law <law@redhat.com>
9395 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9396 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9397 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9398 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9401 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9402 explicitly probe *sp in a noreturn function if there were any callee
9403 register saves or frame pointer is needed.
9405 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9408 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9409 BLKmode for ternary, binary or unary expressions.
9412 * var-tracking.c (delete_vta_debug_insn): New inline function.
9413 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9414 insns from get_insns () to NULL instead of each bb separately.
9415 Use delete_vta_debug_insn. No longer static.
9416 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9417 delete_vta_debug_insns callers.
9418 * rtl.h (delete_vta_debug_insns): Declare.
9419 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9420 instead of variable_tracking_main.
9422 2018-01-03 Martin Sebor <msebor@redhat.com>
9424 PR tree-optimization/83603
9425 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9426 arguments past the endof the argument list in functions declared
9427 without a prototype.
9428 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9429 Avoid checking when arguments are null.
9431 2018-01-03 Martin Sebor <msebor@redhat.com>
9434 * doc/extend.texi (attribute const): Fix a typo.
9435 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9436 issuing -Wsuggest-attribute for void functions.
9438 2018-01-03 Martin Sebor <msebor@redhat.com>
9440 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9441 offset_int::from instead of wide_int::to_shwi.
9442 (maybe_diag_overlap): Remove assertion.
9443 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9444 * gimple-ssa-sprintf.c (format_directive): Same.
9445 (parse_directive): Same.
9446 (sprintf_dom_walker::compute_format_length): Same.
9447 (try_substitute_return_value): Same.
9449 2018-01-03 Jeff Law <law@redhat.com>
9452 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9453 non-constant residual for zero at runtime and avoid probing in
9454 that case. Reorganize code for trailing problem to mirror handling
9457 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9459 PR tree-optimization/83501
9460 * tree-ssa-strlen.c (get_string_cst): New.
9461 (handle_char_store): Call get_string_cst.
9463 2018-01-03 Martin Liska <mliska@suse.cz>
9465 PR tree-optimization/83593
9466 * tree-ssa-strlen.c: Include tree-cfg.h.
9467 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9468 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9469 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9471 (strlen_dom_walker::before_dom_children): Call
9472 gimple_purge_dead_eh_edges. Dump tranformation with details
9474 (strlen_dom_walker::before_dom_children): Update call by adding
9475 new argument cleanup_eh.
9476 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9478 2018-01-03 Martin Liska <mliska@suse.cz>
9481 * cif-code.def (VARIADIC_THUNK): New enum value.
9482 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9485 2018-01-03 Jan Beulich <jbeulich@suse.com>
9487 * sse.md (mov<mode>_internal): Tighten condition for when to use
9488 vmovdqu<ssescalarsize> for TI and OI modes.
9490 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9492 Update copyright years.
9494 2018-01-03 Martin Liska <mliska@suse.cz>
9497 * ipa-visibility.c (function_and_variable_visibility): Skip
9498 functions with noipa attribure.
9500 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9502 * gcc.c (process_command): Update copyright notice dates.
9503 * gcov-dump.c (print_version): Ditto.
9504 * gcov.c (print_version): Ditto.
9505 * gcov-tool.c (print_version): Ditto.
9506 * gengtype.c (create_file): Ditto.
9507 * doc/cpp.texi: Bump @copying's copyright year.
9508 * doc/cppinternals.texi: Ditto.
9509 * doc/gcc.texi: Ditto.
9510 * doc/gccint.texi: Ditto.
9511 * doc/gcov.texi: Ditto.
9512 * doc/install.texi: Ditto.
9513 * doc/invoke.texi: Ditto.
9515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9517 * vector-builder.h (vector_builder::m_full_nelts): Change from
9518 unsigned int to poly_uint64.
9519 (vector_builder::full_nelts): Update prototype accordingly.
9520 (vector_builder::new_vector): Likewise.
9521 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9522 (vector_builder::operator ==): Likewise.
9523 (vector_builder::finalize): Likewise.
9524 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9525 Take the number of elements as a poly_uint64 rather than an
9527 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9528 from unsigned int to poly_uint64.
9529 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9530 (vec_perm_indices::new_vector): Likewise.
9531 (vec_perm_indices::length): Likewise.
9532 (vec_perm_indices::nelts_per_input): Likewise.
9533 (vec_perm_indices::input_nelts): Likewise.
9534 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9535 number of elements per input as a poly_uint64 rather than an
9536 unsigned int. Use the original encoding for variable-length
9537 vectors, rather than clamping each individual element.
9538 For the second and subsequent elements in each pattern,
9539 clamp the step and base before clamping their sum.
9540 (vec_perm_indices::series_p): Handle polynomial element counts.
9541 (vec_perm_indices::all_in_range_p): Likewise.
9542 (vec_perm_indices_to_tree): Likewise.
9543 (vec_perm_indices_to_rtx): Likewise.
9544 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9545 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9546 (tree_vector_builder::new_binary_operation): Handle polynomial
9547 element counts. Return false if we need to know the number
9548 of elements at compile time.
9549 * fold-const.c (fold_vec_perm): Punt if the number of elements
9550 isn't known at compile time.
9552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9554 * vec-perm-indices.h (vec_perm_builder): Change element type
9555 from HOST_WIDE_INT to poly_int64.
9556 (vec_perm_indices::element_type): Update accordingly.
9557 (vec_perm_indices::clamp): Handle polynomial element_types.
9558 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9559 (vec_perm_indices::all_in_range_p): Likewise.
9560 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9562 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9563 polynomial vec_perm_indices element types.
9564 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9565 * fold-const.c (fold_vec_perm): Likewise.
9566 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9567 * tree-vect-generic.c (lower_vec_perm): Likewise.
9568 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9569 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9570 element type to HOST_WIDE_INT.
9572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9573 Alan Hayward <alan.hayward@arm.com>
9574 David Sherwood <david.sherwood@arm.com>
9576 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9577 rather than an int. Use plus_constant.
9578 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9579 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9582 Alan Hayward <alan.hayward@arm.com>
9583 David Sherwood <david.sherwood@arm.com>
9585 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9586 a HOST_WIDE_INT to a poly_int64.
9588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9589 Alan Hayward <alan.hayward@arm.com>
9590 David Sherwood <david.sherwood@arm.com>
9592 * calls.c (load_register_parameters): Cope with polynomial
9593 mode sizes. Require a constant size for BLKmode parameters
9594 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9595 forces a parameter to be padded at the lsb end in order to
9596 fill a complete number of words, require the parameter size
9597 to be ordered wrt UNITS_PER_WORD.
9599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9600 Alan Hayward <alan.hayward@arm.com>
9601 David Sherwood <david.sherwood@arm.com>
9603 * reload1.c (spill_stack_slot_width): Change element type
9604 from unsigned int to poly_uint64_pod.
9605 (alter_reg): Treat mode sizes as polynomial.
9607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9608 Alan Hayward <alan.hayward@arm.com>
9609 David Sherwood <david.sherwood@arm.com>
9611 * reload.c (complex_word_subreg_p): New function.
9612 (reload_inner_reg_of_subreg, push_reload): Use it.
9614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9615 Alan Hayward <alan.hayward@arm.com>
9616 David Sherwood <david.sherwood@arm.com>
9618 * lra-constraints.c (process_alt_operands): Reject matched
9619 operands whose sizes aren't ordered.
9620 (match_reload): Refer to this check here.
9622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9623 Alan Hayward <alan.hayward@arm.com>
9624 David Sherwood <david.sherwood@arm.com>
9626 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9627 that the mode size is in the set {1, 2, 4, 8, 16}.
9629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9630 Alan Hayward <alan.hayward@arm.com>
9631 David Sherwood <david.sherwood@arm.com>
9633 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9634 Use plus_constant instead of gen_rtx_PLUS.
9636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9637 Alan Hayward <alan.hayward@arm.com>
9638 David Sherwood <david.sherwood@arm.com>
9640 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9641 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9642 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9643 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9644 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9645 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9646 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9647 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9648 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9649 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9651 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9652 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9653 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9654 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9655 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9656 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9657 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9658 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9659 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9660 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9662 * expr.c (emit_move_resolve_push): Treat the input and result
9663 of PUSH_ROUNDING as a poly_int64.
9664 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9665 (emit_push_insn): Likewise.
9666 * lra-eliminations.c (mark_not_eliminable): Likewise.
9667 * recog.c (push_operand): Likewise.
9668 * reload1.c (elimination_effects): Likewise.
9669 * rtlanal.c (nonzero_bits1): Likewise.
9670 * calls.c (store_one_arg): Likewise. Require the padding to be
9671 known at compile time.
9673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9674 Alan Hayward <alan.hayward@arm.com>
9675 David Sherwood <david.sherwood@arm.com>
9677 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9678 Use plus_constant instead of gen_rtx_PLUS.
9680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9681 Alan Hayward <alan.hayward@arm.com>
9682 David Sherwood <david.sherwood@arm.com>
9684 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9688 Alan Hayward <alan.hayward@arm.com>
9689 David Sherwood <david.sherwood@arm.com>
9691 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9692 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9693 via stack temporaries. Treat the mode size as polynomial too.
9695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9696 Alan Hayward <alan.hayward@arm.com>
9697 David Sherwood <david.sherwood@arm.com>
9699 * expr.c (expand_expr_real_2): When handling conversions involving
9700 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9701 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9702 as a poly_uint64 too.
9704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9705 Alan Hayward <alan.hayward@arm.com>
9706 David Sherwood <david.sherwood@arm.com>
9708 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9711 Alan Hayward <alan.hayward@arm.com>
9712 David Sherwood <david.sherwood@arm.com>
9714 * combine.c (can_change_dest_mode): Handle polynomial
9715 REGMODE_NATURAL_SIZE.
9716 * expmed.c (store_bit_field_1): Likewise.
9717 * expr.c (store_constructor): Likewise.
9718 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9719 and polynomial REGMODE_NATURAL_SIZE.
9720 (gen_lowpart_common): Likewise.
9721 * reginfo.c (record_subregs_of_mode): Likewise.
9722 * rtlanal.c (read_modify_subreg_p): Likewise.
9724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9725 Alan Hayward <alan.hayward@arm.com>
9726 David Sherwood <david.sherwood@arm.com>
9728 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9729 numbers of elements.
9731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9732 Alan Hayward <alan.hayward@arm.com>
9733 David Sherwood <david.sherwood@arm.com>
9735 * match.pd: Cope with polynomial numbers of vector elements.
9737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9738 Alan Hayward <alan.hayward@arm.com>
9739 David Sherwood <david.sherwood@arm.com>
9741 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9742 in a POINTER_PLUS_EXPR.
9744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9745 Alan Hayward <alan.hayward@arm.com>
9746 David Sherwood <david.sherwood@arm.com>
9748 * omp-simd-clone.c (simd_clone_subparts): New function.
9749 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9750 (ipa_simd_modify_function_body): Likewise.
9752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9753 Alan Hayward <alan.hayward@arm.com>
9754 David Sherwood <david.sherwood@arm.com>
9756 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9757 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9758 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9759 (expand_vector_condition, vector_element): Likewise.
9760 (subparts_gt): New function.
9761 (get_compute_type): Use subparts_gt.
9762 (count_type_subparts): Delete.
9763 (expand_vector_operations_1): Use subparts_gt instead of
9764 count_type_subparts.
9766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9767 Alan Hayward <alan.hayward@arm.com>
9768 David Sherwood <david.sherwood@arm.com>
9770 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9771 (vect_compile_time_alias): ...this new function. Do the calculation
9772 on poly_ints rather than trees.
9773 (vect_prune_runtime_alias_test_list): Update call accordingly.
9775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9776 Alan Hayward <alan.hayward@arm.com>
9777 David Sherwood <david.sherwood@arm.com>
9779 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9781 (vect_schedule_slp_instance): Likewise.
9783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9784 Alan Hayward <alan.hayward@arm.com>
9785 David Sherwood <david.sherwood@arm.com>
9787 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9788 constant and extern definitions for variable-length vectors.
9789 (vect_get_constant_vectors): Note that the number of units
9790 is known to be constant.
9792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9793 Alan Hayward <alan.hayward@arm.com>
9794 David Sherwood <david.sherwood@arm.com>
9796 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9797 of units as polynomial. Choose between WIDE and NARROW based
9800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9801 Alan Hayward <alan.hayward@arm.com>
9802 David Sherwood <david.sherwood@arm.com>
9804 * tree-vect-stmts.c (simd_clone_subparts): New function.
9805 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9808 Alan Hayward <alan.hayward@arm.com>
9809 David Sherwood <david.sherwood@arm.com>
9811 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9812 vectors as polynomial. Use build_index_vector for
9815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9816 Alan Hayward <alan.hayward@arm.com>
9817 David Sherwood <david.sherwood@arm.com>
9819 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9820 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9821 for variable-length vectors.
9822 (vectorizable_mask_load_store): Treat the number of units as
9823 polynomial, asserting that it is constant if the condition has
9824 already been enforced.
9825 (vectorizable_store, vectorizable_load): Likewise.
9827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9828 Alan Hayward <alan.hayward@arm.com>
9829 David Sherwood <david.sherwood@arm.com>
9831 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9832 of units as polynomial. Punt if we can't tell at compile time
9833 which vector contains the final result.
9835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9836 Alan Hayward <alan.hayward@arm.com>
9837 David Sherwood <david.sherwood@arm.com>
9839 * tree-vect-loop.c (vectorizable_induction): Treat the number
9840 of units as polynomial. Punt on SLP inductions. Use an integer
9841 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9842 cast of such a series for variable-length floating-point
9845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9846 Alan Hayward <alan.hayward@arm.com>
9847 David Sherwood <david.sherwood@arm.com>
9849 * tree.h (build_index_vector): Declare.
9850 * tree.c (build_index_vector): New function.
9851 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9852 of units as polynomial, forcibly converting it to a constant if
9853 vectorizable_reduction has already enforced the condition.
9854 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9855 to create a {1,2,3,...} vector.
9856 (vectorizable_reduction): Treat the number of units as polynomial.
9857 Choose vectype_in based on the largest scalar element size rather
9858 than the smallest number of units. Enforce the restrictions
9861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9862 Alan Hayward <alan.hayward@arm.com>
9863 David Sherwood <david.sherwood@arm.com>
9865 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9866 number of units as polynomial.
9868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9869 Alan Hayward <alan.hayward@arm.com>
9870 David Sherwood <david.sherwood@arm.com>
9872 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9873 * target.def (autovectorize_vector_sizes): Return the vector sizes
9874 by pointer, using vector_sizes rather than a bitmask.
9875 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9876 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9877 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9879 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9880 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9881 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9882 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9883 * omp-general.c (omp_max_vf): Likewise.
9884 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9885 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9886 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9887 * tree-vect-slp.c (vect_slp_bb): Likewise.
9888 * doc/tm.texi: Regenerate.
9889 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9891 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9892 the vector size as a poly_uint64 rather than an unsigned int.
9893 (current_vector_size): Change from an unsigned int to a poly_uint64.
9894 (get_vectype_for_scalar_type): Update accordingly.
9895 * tree.h (build_truth_vector_type): Take the size and number of
9896 units as a poly_uint64 rather than an unsigned int.
9897 (build_vector_type): Add a temporary overload that takes
9898 the number of units as a poly_uint64 rather than an unsigned int.
9899 * tree.c (make_vector_type): Likewise.
9900 (build_truth_vector_type): Take the number of units as a poly_uint64
9901 rather than an unsigned int.
9903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9904 Alan Hayward <alan.hayward@arm.com>
9905 David Sherwood <david.sherwood@arm.com>
9907 * target.def (get_mask_mode): Take the number of units and length
9908 as poly_uint64s rather than unsigned ints.
9909 * targhooks.h (default_get_mask_mode): Update accordingly.
9910 * targhooks.c (default_get_mask_mode): Likewise.
9911 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9912 * doc/tm.texi: Regenerate.
9914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9915 Alan Hayward <alan.hayward@arm.com>
9916 David Sherwood <david.sherwood@arm.com>
9918 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9919 * omp-general.c (omp_max_vf): Likewise.
9920 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9921 (expand_omp_simd): Handle polynomial safelen.
9922 * omp-low.c (omplow_simd_context): Add a default constructor.
9923 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9924 (lower_rec_simd_input_clauses): Update accordingly.
9925 (lower_rec_input_clauses): Likewise.
9927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9928 Alan Hayward <alan.hayward@arm.com>
9929 David Sherwood <david.sherwood@arm.com>
9931 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9932 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9933 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9934 (vect_analyze_slp_cost): Likewise.
9935 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9936 (vect_model_load_cost): Likewise.
9938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9939 Alan Hayward <alan.hayward@arm.com>
9940 David Sherwood <david.sherwood@arm.com>
9942 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9943 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9944 from an unsigned int * to a poly_uint64_pod *.
9945 (calculate_unrolling_factor): New function.
9946 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9949 Alan Hayward <alan.hayward@arm.com>
9950 David Sherwood <david.sherwood@arm.com>
9952 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9953 from an unsigned int to a poly_uint64.
9954 (_loop_vec_info::slp_unrolling_factor): Likewise.
9955 (_loop_vec_info::vectorization_factor): Change from an int
9957 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9958 (vect_get_num_vectors): New function.
9959 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9960 (vect_get_num_copies): Use vect_get_num_vectors.
9961 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9962 to an unsigned int *.
9963 (vect_analyze_data_refs): Change min_vf from an int * to a
9965 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9966 than an unsigned HOST_WIDE_INT.
9967 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9968 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9969 to an unsigned int *.
9970 (vect_analyze_data_ref_dependences): Likewise.
9971 (vect_compute_data_ref_alignment): Handle polynomial vf.
9972 (vect_enhance_data_refs_alignment): Likewise.
9973 (vect_prune_runtime_alias_test_list): Likewise.
9974 (vect_shift_permute_load_chain): Likewise.
9975 (vect_supportable_dr_alignment): Likewise.
9976 (dependence_distance_ge_vf): Take the vectorization factor as a
9977 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9978 (vect_analyze_data_refs): Change min_vf from an int * to a
9980 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9981 vfm1 as a poly_uint64 rather than an int. Make the same change
9982 for the returned bound_scalar.
9983 (vect_gen_vector_loop_niters): Handle polynomial vf.
9984 (vect_do_peeling): Likewise. Update call to
9985 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9986 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9988 * tree-vect-loop.c (vect_determine_vectorization_factor)
9989 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9990 (vect_get_known_peeling_cost): Likewise.
9991 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9992 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9993 (vect_transform_loop): Likewise. Use the lowest possible VF when
9994 updating the upper bounds of the loop.
9995 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9997 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9998 polynomial unroll factors.
9999 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10000 (vect_make_slp_decision): Likewise.
10001 (vect_supported_load_permutation_p): Likewise, and polynomial
10003 (vect_analyze_slp_cost): Handle polynomial vf.
10004 (vect_slp_analyze_node_operations): Likewise.
10005 (vect_slp_analyze_bb_1): Likewise.
10006 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10007 than an unsigned HOST_WIDE_INT.
10008 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10009 (vectorizable_load): Handle polynomial vf.
10010 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10012 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10015 Alan Hayward <alan.hayward@arm.com>
10016 David Sherwood <david.sherwood@arm.com>
10018 * match.pd: Handle bit operations involving three constants
10019 and try to fold one pair.
10021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10023 * tree-vect-loop-manip.c: Include gimple-fold.h.
10024 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10025 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10026 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10027 Add a path that uses a step of VF instead of 1, but disable it
10029 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10030 and niters_no_overflow parameters. Update calls to
10031 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10032 Create a new SSA name if the latter choses to use a ste other
10033 than zero, and return it via niters_vector_mult_vf_var.
10034 * tree-vect-loop.c (vect_transform_loop): Update calls to
10035 vect_do_peeling, vect_gen_vector_loop_niters and
10036 slpeel_make_loop_iterate_ntimes.
10037 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10038 (vect_gen_vector_loop_niters): Update declarations after above changes.
10040 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10042 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10043 128-bit round to integer instructions.
10044 (ceil<mode>2): Likewise.
10045 (btrunc<mode>2): Likewise.
10046 (round<mode>2): Likewise.
10048 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10050 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10051 unaligned VSX load/store on P8/P9.
10052 (expand_block_clear): Allow the use of unaligned VSX
10053 load/store on P8/P9.
10055 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10057 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10059 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10060 swap associated with both a load and a store.
10062 2018-01-02 Andrew Waterman <andrew@sifive.com>
10064 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10065 * config/riscv/riscv.md (clear_cache): Use it.
10067 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10069 * web.c: Remove out-of-date comment.
10071 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10073 * expr.c (fixup_args_size_notes): Check that any existing
10074 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10075 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10076 (emit_single_push_insn): ...here.
10078 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10080 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10081 (const_vector_encoded_nelts): New function.
10082 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10083 (const_vector_int_elt, const_vector_elt): Declare.
10084 * emit-rtl.c (const_vector_int_elt_1): New function.
10085 (const_vector_elt): Likewise.
10086 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10087 of CONST_VECTOR_ELT.
10089 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10091 * expr.c: Include rtx-vector-builder.h.
10092 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10093 directly on the tree encoding.
10094 (const_vector_from_tree): Likewise.
10095 * optabs.c: Include rtx-vector-builder.h.
10096 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10097 sequence of "u" values.
10098 * vec-perm-indices.c: Include rtx-vector-builder.h.
10099 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10100 directly on the vec_perm_indices encoding.
10102 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10104 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10105 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10106 * rtx-vector-builder.h: New file.
10107 * rtx-vector-builder.c: Likewise.
10108 * rtl.h (rtx_def::u2): Add a const_vector field.
10109 (CONST_VECTOR_NPATTERNS): New macro.
10110 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10111 (CONST_VECTOR_DUPLICATE_P): Likewise.
10112 (CONST_VECTOR_STEPPED_P): Likewise.
10113 (CONST_VECTOR_ENCODED_ELT): Likewise.
10114 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10115 (unwrap_const_vec_duplicate): Likewise.
10116 (const_vec_series_p): Check for a non-duplicated vector encoding.
10117 Say that the function only returns true for integer vectors.
10118 * emit-rtl.c: Include rtx-vector-builder.h.
10119 (gen_const_vec_duplicate_1): Delete.
10120 (gen_const_vector): Call gen_const_vec_duplicate instead of
10121 gen_const_vec_duplicate_1.
10122 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10123 (gen_const_vec_duplicate): Use rtx_vector_builder.
10124 (gen_const_vec_series): Likewise.
10125 (gen_rtx_CONST_VECTOR): Likewise.
10126 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10127 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10128 Build a new vector rather than modifying a CONST_VECTOR in-place.
10129 (handle_special_swappables): Update call accordingly.
10130 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10131 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10132 Build a new vector rather than modifying a CONST_VECTOR in-place.
10133 (handle_special_swappables): Update call accordingly.
10135 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10137 * simplify-rtx.c (simplify_const_binary_operation): Use
10138 CONST_VECTOR_ELT instead of XVECEXP.
10140 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10142 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10143 the selector elements to be different from the data elements
10144 if the selector is a VECTOR_CST.
10145 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10146 ssizetype for the selector.
10148 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10150 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10151 before testing each element individually.
10152 * tree-vect-generic.c (lower_vec_perm): Likewise.
10154 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10156 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10157 * selftest-run-tests.c (selftest::run_tests): Call it.
10158 * vector-builder.h (vector_builder::operator ==): New function.
10159 (vector_builder::operator !=): Likewise.
10160 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10161 (vec_perm_indices::all_from_input_p): New function.
10162 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10163 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10164 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10165 instead of reading the VECTOR_CST directly. Detect whether both
10166 vector inputs are the same before constructing the vec_perm_indices,
10167 and update the number of inputs argument accordingly. Use the
10168 utility functions added above. Only construct sel2 if we need to.
10170 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10172 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10173 the broadcast of the low byte.
10174 (expand_mult_highpart): Use an explicit encoding for the permutes.
10175 * optabs-query.c (can_mult_highpart_p): Likewise.
10176 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10177 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10178 (vectorizable_bswap): Likewise.
10179 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10180 explicit encoding for the power-of-2 permutes.
10181 (vect_permute_store_chain): Likewise.
10182 (vect_grouped_load_supported): Likewise.
10183 (vect_permute_load_chain): Likewise.
10185 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10187 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10188 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10189 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10190 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10191 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10192 (vect_gen_perm_mask_any): Likewise.
10194 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10196 * int-vector-builder.h: New file.
10197 * vec-perm-indices.h: Include int-vector-builder.h.
10198 (vec_perm_indices): Redefine as an int_vector_builder.
10199 (auto_vec_perm_indices): Delete.
10200 (vec_perm_builder): Redefine as a stand-alone class.
10201 (vec_perm_indices::vec_perm_indices): New function.
10202 (vec_perm_indices::clamp): Likewise.
10203 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10204 (vec_perm_indices::new_vector): New function.
10205 (vec_perm_indices::new_expanded_vector): Update for new
10206 vec_perm_indices class.
10207 (vec_perm_indices::rotate_inputs): New function.
10208 (vec_perm_indices::all_in_range_p): Operate directly on the
10209 encoded form, without computing elided elements.
10210 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10211 encoding. Update for new vec_perm_indices class.
10212 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10213 the given vec_perm_builder.
10214 (expand_vec_perm_var): Update vec_perm_builder constructor.
10215 (expand_mult_highpart): Use vec_perm_builder instead of
10216 auto_vec_perm_indices.
10217 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10218 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10219 or double series encoding as appropriate.
10220 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10221 vec_perm_indices instead of auto_vec_perm_indices.
10222 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10223 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10224 (vect_permute_store_chain): Likewise.
10225 (vect_grouped_load_supported): Likewise.
10226 (vect_permute_load_chain): Likewise.
10227 (vect_shift_permute_load_chain): Likewise.
10228 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10229 (vect_transform_slp_perm_load): Likewise.
10230 (vect_schedule_slp_instance): Likewise.
10231 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10232 (vectorizable_mask_load_store): Likewise.
10233 (vectorizable_bswap): Likewise.
10234 (vectorizable_store): Likewise.
10235 (vectorizable_load): Likewise.
10236 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10237 vec_perm_indices instead of auto_vec_perm_indices. Use
10238 tree_to_vec_perm_builder to read the vector from a tree.
10239 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10240 vec_perm_builder instead of a vec_perm_indices.
10241 (have_whole_vector_shift): Use vec_perm_builder and
10242 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10243 truncation to calc_vec_perm_mask_for_shift.
10244 (vect_create_epilog_for_reduction): Likewise.
10245 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10246 from auto_vec_perm_indices to vec_perm_indices.
10247 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10248 instead of changing individual elements.
10249 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10250 the vector in d.perm.
10251 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10252 from auto_vec_perm_indices to vec_perm_indices.
10253 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10254 instead of changing individual elements.
10255 (arm_vectorize_vec_perm_const): Use new_vector to install
10256 the vector in d.perm.
10257 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10258 Update vec_perm_builder constructor.
10259 (rs6000_expand_interleave): Likewise.
10260 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10261 (rs6000_expand_interleave): Likewise.
10263 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10265 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10266 to qimode could truncate the indices.
10267 * optabs.c (expand_vec_perm_var): Likewise.
10269 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10271 * Makefile.in (OBJS): Add vec-perm-indices.o.
10272 * vec-perm-indices.h: New file.
10273 * vec-perm-indices.c: Likewise.
10274 * target.h (vec_perm_indices): Replace with a forward class
10276 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10277 * optabs.h: Include vec-perm-indices.h.
10278 (expand_vec_perm): Delete.
10279 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10280 (expand_vec_perm_const): Declare.
10281 * target.def (vec_perm_const_ok): Replace with...
10282 (vec_perm_const): ...this new hook.
10283 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10284 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10285 * doc/tm.texi: Regenerate.
10286 * optabs.def (vec_perm_const): Delete.
10287 * doc/md.texi (vec_perm_const): Likewise.
10288 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10289 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10290 expand_vec_perm for constant permutation vectors. Assert that
10291 the mode of variable permutation vectors is the integer equivalent
10292 of the mode that is being permuted.
10293 * optabs-query.h (selector_fits_mode_p): Declare.
10294 * optabs-query.c: Include vec-perm-indices.h.
10295 (selector_fits_mode_p): New function.
10296 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10297 is defined, instead of checking whether the vec_perm_const_optab
10298 exists. Use targetm.vectorize.vec_perm_const instead of
10299 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10300 fit in the vector mode before using a variable permute.
10301 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10302 vec_perm_indices instead of an rtx.
10303 (expand_vec_perm): Replace with...
10304 (expand_vec_perm_const): ...this new function. Take the selector
10305 as a vec_perm_indices rather than an rtx. Also take the mode of
10306 the selector. Update call to shift_amt_for_vec_perm_mask.
10307 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10308 Use vec_perm_indices::new_expanded_vector to expand the original
10309 selector into bytes. Check whether the indices fit in the vector
10310 mode before using a variable permute.
10311 (expand_vec_perm_var): Make global.
10312 (expand_mult_highpart): Use expand_vec_perm_const.
10313 * fold-const.c: Includes vec-perm-indices.h.
10314 * tree-ssa-forwprop.c: Likewise.
10315 * tree-vect-data-refs.c: Likewise.
10316 * tree-vect-generic.c: Likewise.
10317 * tree-vect-loop.c: Likewise.
10318 * tree-vect-slp.c: Likewise.
10319 * tree-vect-stmts.c: Likewise.
10320 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10322 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10323 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10324 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10325 (aarch64_vectorize_vec_perm_const): ...this new function.
10326 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10327 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10328 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10329 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10330 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10331 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10332 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10334 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10335 check for NEON modes.
10336 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10337 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10338 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10339 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10341 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10342 the old VEC_PERM_CONST conditions.
10343 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10344 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10345 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10346 (ia64_vectorize_vec_perm_const_ok): Merge into...
10347 (ia64_vectorize_vec_perm_const): ...this new function.
10348 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10349 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10350 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10351 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10352 * config/mips/mips.c (mips_expand_vec_perm_const)
10353 (mips_vectorize_vec_perm_const_ok): Merge into...
10354 (mips_vectorize_vec_perm_const): ...this new function.
10355 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10356 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10357 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10358 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10359 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10360 (rs6000_expand_vec_perm_const): Delete.
10361 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10363 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10364 (altivec_expand_vec_perm_const_le): Take each operand individually.
10365 Operate on constant selectors rather than rtxes.
10366 (altivec_expand_vec_perm_const): Likewise. Update call to
10367 altivec_expand_vec_perm_const_le.
10368 (rs6000_expand_vec_perm_const): Delete.
10369 (rs6000_vectorize_vec_perm_const_ok): Delete.
10370 (rs6000_vectorize_vec_perm_const): New function.
10371 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10372 an element count and rtx array.
10373 (rs6000_expand_extract_even): Update call accordingly.
10374 (rs6000_expand_interleave): Likewise.
10375 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10376 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10377 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10378 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10379 (rs6000_expand_vec_perm_const): Delete.
10380 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10381 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10382 (altivec_expand_vec_perm_const_le): Take each operand individually.
10383 Operate on constant selectors rather than rtxes.
10384 (altivec_expand_vec_perm_const): Likewise. Update call to
10385 altivec_expand_vec_perm_const_le.
10386 (rs6000_expand_vec_perm_const): Delete.
10387 (rs6000_vectorize_vec_perm_const_ok): Delete.
10388 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10389 reference to the SPE evmerge intructions.
10390 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10391 an element count and rtx array.
10392 (rs6000_expand_extract_even): Update call accordingly.
10393 (rs6000_expand_interleave): Likewise.
10394 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10395 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10397 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10399 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10401 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10402 vector mode and that that mode matches the mode of the data
10404 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10405 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10406 directly using expand_vec_perm_1 when forcing selectors into
10408 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10410 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10412 * optabs-query.h (can_vec_perm_p): Delete.
10413 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10414 * optabs-query.c (can_vec_perm_p): Split into...
10415 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10416 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10417 particular selector is valid.
10418 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10419 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10420 (vect_grouped_load_supported): Likewise.
10421 (vect_shift_permute_load_chain): Likewise.
10422 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10423 (vect_transform_slp_perm_load): Likewise.
10424 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10425 (vectorizable_bswap): Likewise.
10426 (vect_gen_perm_mask_checked): Likewise.
10427 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10428 implementations of variable permutation vectors into account
10429 when deciding which selector to use.
10430 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10431 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10432 with a false third argument.
10433 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10434 to test whether the constant selector is valid and can_vec_perm_var_p
10435 to test whether a variable selector is valid.
10437 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10439 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10440 * optabs-query.c (can_vec_perm_p): Likewise.
10441 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10442 instead of vec_perm_indices.
10443 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10444 (vect_gen_perm_mask_checked): Likewise,
10445 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10446 (vect_gen_perm_mask_checked): Likewise,
10448 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10450 * optabs-query.h (qimode_for_vec_perm): Declare.
10451 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10452 (qimode_for_vec_perm): ...this new function.
10453 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10455 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10457 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10458 does not have a conditional at the top.
10460 2018-01-02 Richard Biener <rguenther@suse.de>
10462 * ipa-inline.c (big_speedup_p): Fix expression.
10464 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10467 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10470 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10474 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10475 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10476 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10477 cond_taken_branch_cost 3->4.
10479 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10481 PR tree-optimization/83581
10482 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10483 TODO_cleanup_cfg if any changes have been made.
10485 PR middle-end/83608
10486 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10487 convert_modes if target mode has the right side, but different mode
10490 PR middle-end/83609
10491 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10492 last argument when extracting from CONCAT. If either from_real or
10493 from_imag is NULL, use expansion through memory. If result is not
10494 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10495 the parts directly to inner mode, if even that fails, use expansion
10498 PR middle-end/83623
10499 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10500 check for bswap in mode rather than HImode and use that in expand_unop
10503 Copyright (C) 2018 Free Software Foundation, Inc.
10505 Copying and distribution of this file, with or without modification,
10506 are permitted in any medium without royalty provided the copyright
10507 notice and this notice are preserved.