1 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
10 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
11 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
12 and -mstring-compare-inline-limit.
14 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
18 for stack clash protection in a register whenever we need it to be in
21 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
23 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
24 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
26 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
29 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
31 (mlong-double-): Ditto.
32 * config/rs6000/sysv4.opt (msdata=): Ditto.
35 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
37 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
39 "vector int vec_ldl (int, long int *)", and
40 "vector unsigned int vec_ldl (int, unsigned long int *)".
41 Add comments and entries for
42 "vector bool char vec_ldl (int, bool char *)",
43 "vector bool short vec_ldl (int, bool short *)",
44 "vector bool int vec_ldl (int, bool int *)",
45 "vector bool long long vec_ldl (int, bool long long *)",
46 "vector pixel vec_ldl (int, pixel *)",
47 "vector long long vec_ldl (int, long long *)",
48 "vector unsigned long long vec_ldl (int, unsigned long long *)".
49 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
50 type tree bool_long_long_type_node and correct definition of
51 bool_V2DI_type_node to make reference to this new type tree.
52 (rs6000_mangle_type): Replace erroneous reference to
53 bool_long_type_node with bool_long_long_type_node.
54 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
55 comments to emphasize sign distinctions for char and int types and
56 replace RS6000_BTI_bool_long constant with
57 RS6000_BTI_bool_long_long constant. Also add comment to restrict
58 use of RS6000_BTI_pixel.
59 (bool_long_type_node): Remove this macro definition.
60 (bool_long_long_type_node): New macro definition
62 2018-04-10 Jakub Jelinek <jakub@redhat.com>
64 PR rtl-optimization/85300
65 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
66 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
67 simplify_unary_operation fails.
69 2018-04-10 Martin Liska <mliska@suse.cz>
71 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
72 cgraph_edge and ipa_ref.
74 2018-04-10 Jakub Jelinek <jakub@redhat.com>
79 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
80 computation of the VEC_MERGE selector from mask.
81 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
82 Fix decoding of the VEC_MERGE selector into mask.
84 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
86 PR tree-optimization/85286
87 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
89 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
91 * final.c (final_1): Set insn_last_address as well as
94 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
97 * explow.c (emit_stack_probe): Call validize_mem on memory location
98 before passing it to gen_probe_stack. Create address operand and
99 legitimize it for the probe_stack_address case.
101 2018-04-09 Jan Hubicka <jh@suse.cz>
104 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
105 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
106 * tree.c (free_lang_data_in_type): Fix handling of binfos;
108 (free_lang_data): Rebuild type inheritance graph.
110 2018-04-09 Martin Sebor <msebor@redhat.com>
112 * invoke.texi (-finline-small-functions): Mention other optimization
114 (-findirect-inlining, -fpartial-inlining): Same.
115 (-finline-functions-called-once): Same.
116 (-freorder-blocks-and-partition): Same.
118 2018-04-09 Jan Hubicka <jh@suse.cz>
121 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
122 jumps; choose last target that matches the criteria (i.e.
123 no partition changes for non-crossing jumps).
124 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
125 support for redirecting crossing jumps to non-crossing.
127 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
129 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
130 also for naked functions.
132 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
134 * config/arc/arc.md (add_shift): New pattern.
135 (add_shift2): Likewise.
136 (sub_shift): Likewise.
137 (sub_shift_cmp0_noout): Likewise.
138 (compare_si_ashiftsi): Likewise.
139 (xbfu_cmp0_noout): New combine pattern.
140 (xbfu_cmp0"): Likewise.
141 (movsi_set_cc_insn): Place the predicable variant first.
142 (commutative_binary_cmp0_noout): Remove clobber.
143 (commutative_binary_cmp0): New pattern.
144 (noncommutative_binary_cmp0): Likewise.
145 (noncommutative_binary_cmp0_noout): Likewise.
146 (noncommutative_binary_comparison_result_used): Removed.
147 (rsub_cmp0): New pattern.
148 (rsub_cmp0_noout): Likewise.
149 (extzvsi): Changed, keep only meaningful variants.
150 (SQH, SEZ): New iterators.
151 (SQH_postfix): New mode attribute.
152 (SEZ_prefix): New code attribute.
153 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
154 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
155 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
157 (noncommutative_operator): Check the availability of barrel
160 2018-04-09 Richard Biener <rguenther@suse.de>
162 PR tree-optimization/85284
163 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
164 Only use the niter constraining form of simple_iv when the exit
167 2018-04-09 Tom de Vries <tom@codesourcery.com>
170 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
171 (define_expand "*memory_barrier"): New define_expand.
172 (define_insn "memory_barrier"): New insn.
174 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
176 PR rtl-optimization/80463
177 PR rtl-optimization/83972
178 PR rtl-optimization/83480
180 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
181 correct producer for the insn.
182 (tidy_control_flow): Fixup seqnos in case of debug insns.
184 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
186 PR rtl-optimization/83913
188 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
189 different sched-times when merging exprs.
191 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
193 PR rtl-optimization/83962
195 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
196 tidy_fallthru_edge and tidy_control_flow.
198 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
200 PR rtl-optimization/83530
202 * sel-sched.c (force_next_insn): New global variable.
203 (remove_insn_for_debug): When force_next_insn is true, also leave only
204 next insn in the ready list.
205 (sel_sched_region): When the region wasn't scheduled, make another pass
206 over it with force_next_insn set to 1.
208 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
210 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
212 * config/nds32/constants.md (unspec_volatile_element): Add enum values
213 for interrupt control.
214 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
215 functions for interrupt control.
216 * config/nds32/nds32-intrinsic.md: Likewise.
217 * config/nds32/nds32_intrinsic.h: Likewise.
218 * config/nds32/nds32.h (nds32_builtins): Likewise.
220 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
222 * config/nds32/nds32.c (nds32_init_machine_status,
223 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
224 strict_aligned_p field.
225 (nds32_expand_to_rtl_hook): New function.
226 (TARGET_EXPAND_TO_RTL_HOOK): Define.
227 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
229 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
230 Chung-Ju Wu <jasonwucj@gmail.com>
232 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
233 * config/nds32/nds32-n7.md: New file.
234 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
235 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
237 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
238 * config/nds32/nds32.md (pipeline_model): Add n7.
239 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
240 * config/nds32/pipelines.md: Include n7 settings.
242 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
243 Chung-Ju Wu <jasonwucj@gmail.com>
245 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
246 * config/nds32/nds32-e8.md: New file.
247 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
248 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
250 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
251 * config/nds32/nds32.md (pipeline_model): Add e8.
252 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
253 * config/nds32/pipelines.md: Include e8 settings.
255 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
256 Chung-Ju Wu <jasonwucj@gmail.com>
258 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
259 * config/nds32/nds32-n8.md: New file.
260 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
261 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
263 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
264 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
265 * config/nds32/nds32.md (pipeline_model): Add n8.
266 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
267 * config/nds32/pipelines.md: Include n8 settings.
269 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
270 Chung-Ju Wu <jasonwucj@gmail.com>
272 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
273 * config/nds32/nds32-n9-2r1w.md: New file.
274 * config/nds32/nds32-n9-3r2w.md: New file.
275 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
276 nds32_register_ports): New or modify for cpu n9.
277 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
279 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
280 * config/nds32/nds32-utils.c: New file.
281 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
282 TARGET_MUL_SLOW): Define.
283 * config/nds32/nds32.md (pipeline_model): New attribute.
284 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
285 New options that support cpu n9.
286 * config/nds32/pipelines.md: Include n9 settings.
287 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
289 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
291 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
292 information if necessary.
293 (output_cond_branch_compare_zero): Likewise.
294 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
295 (nds32_target_alignment): Refine for alignment.
296 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
297 (FUNCTION_BOUNDARY): Modify.
298 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
300 * config/nds32/nds32.opt (malways-align, malign-functions): New.
302 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
304 * config/nds32/constants.md (unspec_volatile_element): Add values for
305 TLB operation and data prefetch.
306 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
307 functions for TLB operation and data prefetch.
308 * config/nds32/nds32-intrinsic.md: Likewise.
309 * config/nds32/nds32_intrinsic.h: Likewise.
310 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
311 (nds32_print_operand): Likewise.
312 * config/nds32/nds32.h (nds32_builtins): Likewise.
314 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
315 Andrew Pinski <pinsika@gcc.gnu.org>
318 * match.pd: Use constant_boolean_node of correct type instead of
319 boolean_true_node or boolean_false_node for simplifying
320 pointer comparisons to zero.
322 2018-04-07 Jakub Jelinek <jakub@redhat.com>
324 PR tree-optimization/80021
325 * tree.c (verify_type_variant): Make error call in verify_variant_match
326 translatable and remove final full stop.
328 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
330 * config/nds32/constants.md (unspec_volatile_element): Add
331 UNSPEC_VOLATILE_EH_RETURN.
332 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
333 nds32_output_stack_pop): Support dwarf exception handling process.
334 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
335 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
336 exception handling process.
337 (nds32_compute_stack_frame): Likewise.
338 (nds32_return_addr_rtx): Likewise.
339 (nds32_initial_elimination_offset): Likewise.
340 (nds32_expand_prologue): Likewise.
341 (nds32_expand_epilogue): Likewise.
342 (nds32_dynamic_chain_address): New function.
343 * config/nds32/nds32.h (machine_function): Add fields for dwarf
345 (DYNAMIC_CHAIN_ADDRESS): Define.
346 (EH_RETURN_DATA_REGNO): Define.
347 (EH_RETURN_STACKADJ_RTX): Define.
348 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
349 patterns for dwarf exception handling.
351 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
353 * config/nds32/nds32.h: Clean up obsolete macros.
355 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
357 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
358 Add enum values for particular instructions.
359 * config/nds32/nds32-intrinsic.c: Implementation of expanding
360 particular intrinsic functions.
361 * config/nds32/nds32-intrinsic.md: Likewise.
362 * config/nds32/nds32_intrinsic.h: Likewise.
363 * config/nds32/nds32.h (nds32_builtins): Likewise.
364 * config/nds32/nds32.md (type): Add pbsad and pbsada.
365 (btst, ave): New patterns for particular instructions.
367 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
369 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
370 Add enum values for atomic load/store and memory sync.
371 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
373 * config/nds32/nds32-intrinsic.md: Likewise.
374 * config/nds32/nds32_intrinsic.h: Likewise.
375 * config/nds32/nds32.h (nds32_builtins): Likewise.
377 2018-04-07 Jakub Jelinek <jakub@redhat.com>
379 PR tree-optimization/85257
380 * fold-const.c (native_encode_vector): If not all elts could fit
381 and off is -1, return 0 rather than offset.
382 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
383 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
384 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
385 adjust buffer in native_interpret_expr call.
387 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
389 * config/nds32/constants.md (unspec_volatile_element): Add cache
391 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
392 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
393 * config/nds32/nds32.c (nds32_cctl_names): New.
394 (nds32_print_operand): Handle cache control register names.
395 * config/nds32/nds32.h (nds32_builtins): New enum values.
396 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
398 * config/nds32/nds32.md (type): Add mmu.
399 * config/nds32/pipelines.md (simple_insn): Add mmu.
401 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
403 * config/nds32/nds32.md (type): Remove call.
404 * config/nds32/pipelines.md (simple_insn): Likewise.
406 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
408 * config/nds32/constants.md (unspec_volatile_element): Add
409 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
410 UNSPEC_VOLATILE_FMFCFG.
411 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
412 description for fmfcfg and fmfcsr.
413 (bdesc_1arg): Add fmtcsr.
414 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
415 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
416 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
417 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
418 unspec_fmfcfg): New patterns.
419 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
420 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
421 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
422 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
423 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
424 __nds32__fmfcfg): Define.
426 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
428 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
429 intrinsic register names.
430 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
431 intrinsic register enum values and macros.
433 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
435 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
436 for load/store addressing form.
437 (nds32_print_operand_address): Likewise.
439 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
442 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
443 based on LABEL_REF. Remove useless assertion.
444 (pic_address_needs_scratch): Fix formatting.
445 (sparc_legitimize_pic_address): Minor tweaks.
446 (sparc_delegitimize_address): Adjust assertion accordingly.
447 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
448 into symbolic_operand.
449 (movsi_high_pic_label_ref): Likewise.
450 (movsi_lo_sum_pic_label_ref): Likewise.
451 (movdi_pic_label_ref): Likewise.
452 (movdi_high_pic_label_ref): Likewise.
453 (movdi_lo_sum_pic_label_ref): Likewise.
455 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
457 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
458 custom LIB_SPEC setup.
460 2018-04-06 Ruslan Bukin <br@bsdpad.com>
461 Kito Cheng <kito.cheng@gmail.com>
463 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
464 * config/riscv/freebsd.h: New.
466 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
468 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
469 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
472 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
473 Kito Cheng <kito.cheng@gmail.com>
475 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
476 nds32_output_call, nds32_symbol_binds_local_p): New functions.
477 * config/nds32/nds32-protos.h (nds32_output_call,
478 nds32_output_return): Declare.
479 * config/nds32/nds32.md: Refine all the call and return patterns.
481 2018-04-06 Jakub Jelinek <jakub@redhat.com>
484 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
485 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
487 PR rtl-optimization/84872
488 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
489 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
492 2018-04-06 Tamar Christina <tamar.christina@arm.com>
494 * expr.c (copy_blkmode_to_reg): Revert 254862.
495 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
497 2018-04-06 Richard Biener <rguenther@suse.de>
500 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
501 after seeing a component reference with an adjacent field. Treat
502 refs to arrays at struct end of external decls similar to
503 refs to unconstrained commons.
505 2018-04-06 Jakub Jelinek <jakub@redhat.com>
508 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
509 look through SAVE_EXPRs with non-side-effects argument. Adjust
511 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
514 2018-04-06 Richard Biener <rguenther@suse.de>
517 * alias.c (find_base_term): New wrapper around find_base_term
518 unwinding CSELIB_VAL_PTR changes.
519 (find_base_term): Do not restore CSELIB_VAL_PTR during the
522 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
524 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
526 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
527 constant definitions.
528 ("nop"): lr 0,0 -> nopr r0
529 ("nop_lr0", "nop_lr1"): New insn definitions.
531 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
533 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
534 NDS32_V3PUSH_AVAILABLE_P macro.
536 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
537 Chung-Ju Wu <jasonwucj@gmail.com>
539 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
540 (nds32*-*-*): Add float and fpu_config into supported_defaults.
541 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
542 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
543 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
544 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
545 * config/nds32/constraints.md: New constraints and checking for hard
547 * config/nds32/iterators.md: New mode iterator and attribute for hard
549 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
551 * config/nds32/nds32-fpu.md: New file.
552 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
553 deal with hard float code generation.
554 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
556 (abi_type, float_reg_number): New enum type.
557 * config/nds32/nds32-predicates.c: New predicates for hard float.
558 * config/nds32/nds32-protos.h: Declare functions for hard float.
559 * config/nds32/nds32.c: Implementation for hard float configuration.
560 * config/nds32/nds32.h: Definitions for hard float configuration.
561 * config/nds32/nds32.md: Include hard float machine description and
562 modify patterns for hard float configuration.
563 * config/nds32/nds32.opt: New options for hard float configuration.
564 * config/nds32/predicates.md: New predicates for hard float
567 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
569 * common/config/nds32/nds32-common.c
570 (nds32_option_optimization_table): Enable -mreleax-hint by default.
572 2018-04-05 Jakub Jelinek <jakub@redhat.com>
575 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
576 CONSTRUCTOR_ELT (ctor, ...)->value.
578 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
581 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
583 2018-04-05 Tom de Vries <tom@codesourcery.com>
586 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
589 2018-04-05 Shiva Chen <shiva0217@gmail.com>
590 Kito Cheng <kito.cheng@gmail.com>
592 * config/nds32/constraints.md (U33): Fine-tune checking condition.
593 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
594 * config/nds32/nds32.h (nds32_16bit_address_type): Add
595 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
597 2018-04-05 Shiva Chen <shiva0217@gmail.com>
598 Kito Cheng <kito.cheng@gmail.com>
600 * config/nds32/constraints.md (Ufe): New memory constraint.
601 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
602 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
603 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
605 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
606 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
608 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
610 * config/nds32/nds32.md: Use optimize_size in the condition for
611 alu-shift instructions.
613 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
615 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
617 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
619 * config/nds32/nds32.md (negsi2): Refine pattern.
621 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
622 Chung-Ju Wu <jasonwucj@gmail.com>
624 * config/nds32/iterators.md (shift_rotate): New code iterator.
625 (shift): New code attribute.
626 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
627 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
628 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
629 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
631 (andsi3, *andsi3): Ditto.
632 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
633 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
634 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
635 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
636 nds32_ior_operand, nds32_xor_operand): New predicates.
638 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
640 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
641 (addsi3, subsi3): ... this.
643 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
645 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
647 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
649 * config/nds32/nds32.md: Adjust indention.
651 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
653 * config/nds32/nds32.md (feature): New attribute.
655 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
657 * config/nds32/nds32.md (subtype): New attribute.
659 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
662 * config/arm/arm-builtins.c (arm_expand_builtin): Change
663 expansion to perform a bitwise AND of the argument followed by a
664 boolean negation of the result.
666 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
668 PR rtl-optimization/84878
669 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
670 the basic block. Assert the use reference is not artificial and that
671 it has an associated insn.
673 2018-04-04 Michael Matz <matz@suse.de>
675 * builtins.c (compute_objsize): Pass correct operand
676 to array_at_struct_end_p.
678 2018-04-04 Richard Biener <rguenther@suse.de>
681 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
682 from contexts for DINFO_LEVEL_TERSE and below.
684 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
686 * config/nds32/nds32-doubleword.md (move_<mode>): Require
687 resiter_operand condition.
688 * config/nds32/nds32.md (*move<mode>): Ditto.
690 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
691 Monk Chiang <sh.chiang04@gmail.com>
693 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
695 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
697 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
699 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
700 Kito Cheng <kito.cheng@gmail.com>
702 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
703 nds32_cond_code_str, output_cond_branch,
704 output_cond_branch_compare_zero, 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: New functions.
710 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
711 nds32_expand_cstore, nds32_expand_movcc,
712 nds32_output_cbranchsi4_equality_zero,
713 nds32_output_cbranchsi4_equality_reg,
714 nds32_output_cbranchsi4_equality_reg_or_const_int,
715 nds32_output_cbranchsi4_greater_less_zero): Declare.
716 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
717 nds32_rimm11s_operand): New predicates.
718 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
719 * config/nds32/nds32.md: Rewrite all the branch and conditional move
722 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
724 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
725 * config/nds32/nds32.md: Ditto.
726 * config/nds32/pipelines.md: Ditto.
728 2018-04-04 Richard Biener <rguenther@suse.de>
730 PR tree-optimization/85168
731 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
732 propagating abnormals.
734 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
736 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
738 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
739 Kito Cheng <kito.cheng@gmail.com>
741 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
742 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
743 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
744 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
745 * config/nds32/nds32.md (sibcall_internal): New.
746 (sibcall_register): Remove.
747 (sibcall_immediate): Remove.
748 (sibcall_value_internal): New.
749 (sibcall_value_register): Remove.
750 (sibcall_value_immediate): Remove.
751 * config/nds32/predicates.md (nds32_general_register_operand): New.
752 (nds32_call_address_operand): New.
754 2018-04-03 Jakub Jelinek <jakub@redhat.com>
756 PR rtl-optimization/85167
757 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
758 bb_defs if *split_p, instead preinitialize it to NULL.
760 PR tree-optimization/85156
761 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
762 evaluating the argument multiple times.
764 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
766 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
768 (_mm_cvtpd_ps): Likewise.
769 (_mm_cvttpd_epi32): Likewise.
770 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
771 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
772 vector, pixel, and bool following altivec.h include.
774 2018-04-03 Martin Sebor <msebor@redhat.com>
776 * doc/extend.texi (Common Function Attributes): Clarify.
777 (const attribute): Likewise.
778 (pure attribute): Likewise.
780 2018-04-03 Jakub Jelinek <jakub@redhat.com>
783 * config/i386/i386.c (ix86_expand_vector_set): Use
784 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
786 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
788 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
789 instructions when changing rounding bits to preserve precision bits
790 in the x87 control word.
792 2018-04-03 Martin Liska <mliska@suse.cz>
794 PR tree-optimization/82491
795 * rtl.h (strip_offset_and_add): Replace += suboffset with
796 poly_uint64 () + suboffset.
798 2018-03-29 Martin Liska <mliska@suse.cz>
799 Martin Jambor <mjambor@suse.cz>
802 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
803 param_type is not an integral or pointer type.
805 2018-04-03 Richard Biener <rguenther@suse.de>
807 * sese.h (recompute_all_dominators): Remove.
809 2018-04-02 Martin Sebor <msebor@redhat.com>
811 * doc/invoke.texi (-Wrestrict): Fix typos.
813 2018-04-02 Jim Wilson <jimw@sifive.com>
815 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
816 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
817 (<optab>di3, <optab>si3_extend): Likewise.
818 (<optab>si3_mask, <optab>si3_mask_1): New.
819 (<optab>di3_mask, <optab>di3_mask_1): New.
820 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
821 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
822 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
824 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
826 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
829 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
831 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
832 (nds32_canonicalize_comparison): New function.
834 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
835 Kito Cheng <kito.cheng@gmail.com>
836 Kuan-Lin Chen <kuanlinchentw@gmail.com>
838 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
839 * config/nds32/constants.md (unspec_volatile_element): Add
840 UNSPEC_VOLATILE_RELAX_GROUP.
841 * config/nds32/nds32-relax-opt.c: New file.
842 * config/nds32/nds32-predicates.c
843 (nds32_symbol_load_store_p): New function.
844 * config/nds32/nds32-protos.h
845 (nds32_symbol_load_store_p): Declare function.
846 (make_pass_nds32_relax_opt): Declare new rtl pass function.
847 * config/nds32/nds32.c
848 (nds32_register_pass): New function to register pass.
849 (nds32_register_passes): New function to register passes.
850 * config/nds32/nds32.md (relax_group): New pattern.
851 * config/nds32/nds32.opt (mrelax-hint): New option.
852 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
854 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
856 * config/nds32/t-nds32: Modify files dependency.
858 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
860 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
861 (PROFILE_HOOK): Define its implementation.
863 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
865 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
866 type and 32-bit size.
868 2018-04-01 Jakub Jelinek <jakub@redhat.com>
871 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
872 (V_128_256): New mode iterator.
873 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
874 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
875 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
877 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
878 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
880 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
883 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
884 NaN inputs correctly.
886 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
889 * config/rs6000/vsx.md (??r): New mode attribute.
890 (*vsx_mov<mode>_64bit): Use it.
891 (*vsx_mov<mode>_32bit): Likewise.
893 2018-03-30 Martin Sebor <msebor@redhat.com>
895 PR tree-optimization/84818
896 * builtins.c (check_access): Use warning_n.
898 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
901 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
903 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
906 2018-03-30 Julia Koval <julia.koval@intel.com>
909 * x86-tune.def (movx, partial_reg_dependency): Enable for
912 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
915 * lra-constraints.c (process_alt_operands): Move setting
916 this_alternative_matches below.
918 2018-03-29 Martin Liska <mliska@suse.cz>
921 * doc/invoke.texi: Document how LTO works with debug info.
922 Describe auto-load support of binutils. Mention 'x86-64'
923 as valid option value of -march option.
925 2018-03-29 Jakub Jelinek <jakub@redhat.com>
927 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
930 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
931 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
932 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
935 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
938 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
939 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
940 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
941 for RS6000_BTM_POWERPC64.
942 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
943 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
944 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
949 2018-03-28 Carl Love <cel@us.ibm.com>
952 2017-09-27 Carl Love <cel@us.ibm.com>
954 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
955 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
956 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
959 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
961 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
962 instead of __vector bool.
963 (_mm_max_pu8): Likewise.
964 (_mm_min_pi16): Likewise.
966 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
969 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
973 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
975 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
976 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
977 (div_extend): Likewise.
978 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
980 (__builtin_divweuo): Likewise.
981 (__builtin_divdeo): Likewise.
982 (__builtin_divdeuo): Likewise.
984 2018-03-28 Jakub Jelinek <jakub@redhat.com>
987 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
988 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
990 PR tree-optimization/82004
991 * gimple-match-head.c (optimize_pow_to_exp): New function.
992 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
993 Don't fold to exp if optimize_pow_to_exp is false.
995 2018-03-28 Martin Liska <mliska@suse.cz>
998 * calls.c (initialize_argument_information): Fix trailing space.
999 * common.opt: Fix typo and provide better explanation for
1000 -fsanitize-coverage option.
1001 * config/i386/i386.opt: Fix typo.
1003 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1004 Martin Liska <mliska@suse.cz>
1007 * gimplify.c (asan_poison_variable): Don't do the check for
1008 gimplify_omp_ctxp here.
1009 (gimplify_decl_expr): Do it here.
1010 (gimplify_target_expr): Likewise.
1012 2018-03-28 Martin Liska <mliska@suse.cz>
1015 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1016 chkp_type_bounds_count if MPX is not enabled.
1018 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1020 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1022 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1025 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1026 function to create the function decl for complex long double
1027 multiply and divide for -mabi=ieeelongdouble.
1028 (init_float128_ieee): Call it.
1030 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1033 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1034 -fcf-protection=branch -mibt.
1035 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1037 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1040 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1042 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1045 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1046 extern array declarations.
1048 2018-03-27 Richard Biener <rguenther@suse.de>
1051 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1052 explicit single_use checks.
1054 2018-03-27 Richard Biener <rguenther@suse.de>
1056 PR tree-optimization/85082
1057 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1060 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1062 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1063 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1064 Turn on fasynchronous-unwind-tables and funwind-tables.
1066 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1069 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1070 (*bmi_blsr_<mode>_ccz): Ditto.
1072 2018-03-26 Tom de Vries <tom@codesourcery.com>
1074 PR tree-optimization/85063
1075 * omp-general.c (offloading_function_p): New function. Factor out
1077 * omp-offload.c (pass_omp_target_link::gate): ... here.
1078 * omp-general.h (offloading_function_p): Declare.
1079 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1080 with attribute omp declare target for offloading functions.
1082 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1084 PR tree-optimization/84005
1085 * tree-data-ref.h (get_base_for_alignment): Declare.
1086 * tree-data-ref.c (get_base_for_alignment_1): New function.
1087 (get_base_for_alignment): Likewise.
1088 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1089 get_base_for_alignment to find a suitable base object, instead
1090 of always using drb->base_address.
1092 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1095 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1096 known size by default.
1098 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1101 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1102 and non BLKmode operands.
1104 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1107 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1108 Clean up attributes.
1110 2018-03-23 Richard Biener <rguenther@suse.de>
1113 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1114 we are going to emit early debug for LTO.
1116 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1119 * function.c (match_asm_constraints_1): Don't optimize if input
1120 doesn't satisfy general_operand predicate for output's mode.
1123 * alias.c (write_dependence_p): Don't require for x_canonicalized
1124 non-VOIDmode if x has VOIDmode.
1127 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1128 just don't try to optimize it rather than assert it never happens.
1130 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1132 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1133 macro expansions for definition of ST_INTERNAL_<mode> and
1134 LD_INTERNAL_<mode> builtins.
1135 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1137 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1139 (altivec_expand_st_builtin): Likewise.
1140 (altivec_expand_builtin): Remove calls to deleted functions.
1141 (rs6000_address_for_altivec): Delete this function.
1142 * config/rs6000/vector.md: Remove expands for
1143 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1145 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1148 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1149 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1150 re-computing once computed.
1151 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1152 (arm_init_machine_status): Initialize
1153 machine->static_chain_stack_bytes.
1155 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1158 * doc/extend.texi: Add four new prototypes for vec_ld.
1159 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1160 definitions for more logical presentation.
1161 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1162 entries for V1TI variants of __builtin_altivec_ld builtin.
1163 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1164 handling of V1TI variant of LVX icode pattern.
1165 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1166 (rs6000_gimple_fold_builtin): Likewise.
1167 (altivec_init_builtins): Add code to define
1168 __builtin_altivec_lvx_v1ti function.
1170 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1173 * function.c (match_asm_constraints_1): Don't do the optimization
1174 if input isn't a REG, SUBREG, MEM or constant.
1176 2018-03-22 Tom de Vries <tom@codesourcery.com>
1178 PR tree-optimization/84956
1179 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1180 bb_has_abnormal_pred.
1182 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1185 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1186 DECL_INITIAL (decl) to decl at the end.
1187 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1190 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1192 * doc/extend.texi (__builtin_tgmath): Document when complex
1193 integer types are treated as _Complex _Float64.
1195 2018-03-21 Tom de Vries <tom@codesourcery.com>
1197 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1199 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1201 PR tree-optimization/84960
1202 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1203 if it is ENTRY block, move them into single succ of ENTRY in that case.
1205 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1207 PR tree-optimization/84811
1208 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1209 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1210 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1211 (known_subrange_p): Update accordingly. Cast each value involved
1212 in the size comparison, rather than casting the result of the
1215 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1217 PR tree-optimization/84982
1218 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1219 by flipping the least significant bit rather than all bits from
1220 bitpos to bitpos + bitsize - 1.
1222 2018-03-21 Nathan Sidwell <nathan@acm.org>
1224 * doc/extend.texi (Deprecated Features): Remove mention of
1225 long-deleted deprecations.
1227 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1230 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1231 * configure: Regenerate.
1233 2018-03-21 Tom de Vries <tom@codesourcery.com>
1235 PR tree-optimization/83126
1236 * tree-parloops.c (num_phis): New function.
1237 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1239 2018-03-21 Nathan Sidwell <nathan@acm.org>
1241 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1242 mention anon-struct/union members and trailing attributes.
1244 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1246 PR tree-optimization/84969
1247 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1248 builtin memset partitions if they set different rhs values.
1250 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1252 PR rtl-optimization/84989
1253 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1254 VEC_DUPLICATE with scalar result mode.
1256 2018-03-21 Martin Liska <mliska@suse.cz>
1259 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1260 not intended return statement.
1262 2018-03-21 Martin Liska <mliska@suse.cz>
1265 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1266 (chkp_find_bound_slots_1): Limit number of iterations.
1268 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1271 * Minor grammar fixes for x86 options.
1273 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1276 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1277 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1279 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1282 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1283 (altivec_lvx_<mode>_1op): Likewise.
1284 (altivec_stvx_<mode>_2op): Likewise.
1285 (altivec_stvx_<mode>_1op): Likewise.
1286 (altivec_lvx_<VM2:mode>): New define_expand.
1287 (altivec_stvx_<VM2:mode>): Likewise.
1288 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1289 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1290 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1291 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1292 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1293 (rs6000_gen_lvx): Likewise.
1294 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1295 (altivec_expand_stv_builtin): Likewise.
1296 (altivec_expand_builtin): Likewise.
1297 * config/rs6000/vector.md: Likewise.
1299 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1302 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1305 2018-03-20 Richard Biener <rguenther@suse.de>
1308 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1309 sign-conversions as zero, fall back to standard scalar_stmt
1312 2018-03-20 Martin Liska <mliska@suse.cz>
1315 * predict.c (rebuild_frequencies): Handle case when we have
1316 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1318 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1321 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1322 flag_section_anchors.
1323 * varasm.c (use_blocks_for_decl_p): Remove hack for
1324 dw2_force_const_mem.
1327 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1329 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1330 be created, use lowpart_subreg of operands[0] rather than operands[0]
1332 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1333 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1334 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1335 and n constraint instead of aarch64_shift_imm_di and Usd.
1336 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1337 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1339 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1342 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1343 to favor GPR over NEON registers.
1344 (<shift>di3_neon): Likewise.
1346 2018-03-20 Tom de Vries <tom@codesourcery.com>
1349 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1350 (nvptx_process_pars): Emit bar.sync asap and alap.
1352 2018-03-20 Tom de Vries <tom@codesourcery.com>
1355 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1356 seen_label if seen_label is already set.
1358 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1361 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1362 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1363 Use 1U instead of 1. Formatting fixes.
1366 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1367 instead of TREE_TYPE (s1) for the return value.
1369 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1371 PR tree-optimization/84946
1372 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1373 bitsize + bitsize in poly_uint64 rather than poly_int64.
1376 * dwarf2asm.c: Include fold-const.c.
1377 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1378 of decl rather than decl itself.
1380 PR rtl-optimization/84643
1381 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1383 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1386 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1387 calling assemble_variable.
1389 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1392 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1393 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1395 2018-03-19 Jim Wilson <jimw@sifive.com>
1398 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1399 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1400 (riscv_first_stack_step): Likewise.
1401 (riscv_option_override): Use STACK_BOUNDARY instead of
1403 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1405 (BIGGEST_ALIGNMENT): Set to 128.
1406 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1407 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1410 2018-03-19 Richard Biener <rguenther@suse.de>
1412 PR tree-optimization/84933
1413 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1414 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1416 2018-03-19 Richard Biener <rguenther@suse.de>
1418 PR tree-optimization/84859
1419 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1420 (cond_if_else_store_replacement): Perform sinking operation on
1421 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1422 Generalize what a BB with a single eligible store is.
1424 2018-03-19 Richard Biener <rguenther@suse.de>
1426 PR tree-optimization/84929
1427 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1428 chrec_is_positive against non-chrec arg.
1430 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1433 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1435 2018-03-18 Martin Liska <mliska@suse.cz>
1437 PR rtl-optimization/84635
1438 * regrename.c (build_def_use): Use matches_mode only when
1441 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1443 PR tree-optimization/84913
1444 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1445 vectorize chains of COND_EXPRs.
1447 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1449 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1451 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1453 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1455 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1457 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1459 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1460 Kito Cheng <kito.cheng@gmail.com>
1462 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1463 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1464 (nds32_adjust_reg_alloc_order): New function.
1465 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1467 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1469 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1470 nds32_print_operand, nds32_print_operand_address): Use
1471 HOST_WIDE_INT_PRINT_DEC instead.
1473 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1475 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1477 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1480 * config/i386/i386.c (initial_ix86_tune_features,
1481 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1483 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1484 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1485 rather than 1u << ix86_tune. Formatting fix.
1486 (ix86_option_override_internal): Change ix86_arch_mask from
1487 unsigned int to unsigned HOST_WIDE_INT, initialize to
1488 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1489 (ix86_function_specific_restore): Likewise.
1491 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1494 * postreload.c (reload_combine_recognize_pattern): Perform
1495 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1496 truncate_int_for_mode the result for the destination's mode.
1499 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1500 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1502 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1505 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1508 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1509 regno_allocno_class_array and sorted_pseudos.
1510 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1511 insns where regno is used.
1513 2018-03-16 Martin Liska <mliska@suse.cz>
1516 * multiple_target.c (create_dispatcher_calls): Redirect
1517 reference in the symbol table.
1519 2018-03-16 Martin Liska <mliska@suse.cz>
1522 * multiple_target.c (create_dispatcher_calls): Redirect also
1525 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1529 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1530 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1531 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1533 2018-03-16 Julia Koval <julia.koval@intel.com>
1535 * doc/invoke.texi (Skylake Server): Add CLWB.
1536 Cannonlake): Remove CLWB.
1538 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1540 PR tree-optimization/84841
1541 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1543 (FLOAT_ONE_CONST_TYPE): Define.
1544 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1545 (sort_by_operand_rank): Put entries with higher constant_type last
1546 rather than first to match comments.
1548 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1550 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1553 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1556 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1557 check and use address of target always.
1559 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1562 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1563 (indirect_thunk_bnd_needed): Likewise.
1564 (indirect_thunks_used): Likewise.
1565 (indirect_thunks_bnd_used): Likewise.
1566 (indirect_return_needed): New.
1567 (indirect_return_bnd_needed): Likewise.
1568 (output_indirect_thunk_function): Add a bool argument for
1570 (output_indirect_thunk_function): Don't generate alias for
1571 function return thunk.
1572 (ix86_code_end): Call output_indirect_thunk_function to generate
1573 function return thunks.
1574 (ix86_output_function_return): Set indirect_return_bnd_needed
1575 and indirect_return_needed instead of indirect_thunk_bnd_needed
1576 and indirect_thunk_needed.
1578 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1580 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1581 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1582 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1584 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1585 Paul Hua <paul.hua.gm@gmail.com>
1588 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1590 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1592 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1593 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1596 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1599 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1600 instead of GET_MODE_SIZE when comparing Units.
1602 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1605 * varasm.c (hash_section): Return an unchangeble hash value
1606 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1607 Return !aarch64_can_use_per_function_literal_pools_p ().
1609 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1612 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1613 rather than address of cmode as last argument to prepare_cmp_insn.
1615 2018-03-15 Julia Koval <julia.koval@intel.com>
1617 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1618 F_AVX512VNNI, F_AVX512BITALG): New.
1620 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1623 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1624 insn for floating-point loads and stores.
1626 2018-03-14 Carl Love <cel@us.ibm.com>
1628 * config/rs6000/rs6000-c.c: Add macro definitions for
1629 ALTIVEC_BUILTIN_VEC_PERMXOR.
1630 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1631 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1632 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1633 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1635 * config/doc/extend.texi: Add prototypes for vec_permxor.
1637 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1640 * diagnostic-show-locus.c (class layout_point): Convert m_line
1641 from int to linenum_type.
1642 (line_span::comparator): Use linenum "compare" function when
1643 comparing line numbers.
1644 (test_line_span): New function.
1645 (layout_range::contains_point): Convert param "row" from int to
1647 (layout_range::intersects_line_p): Likewise.
1648 (layout::will_show_line_p): Likewise.
1649 (layout::print_source_line): Likewise.
1650 (layout::should_print_annotation_line_p): Likewise.
1651 (layout::print_annotation_line): Likewise.
1652 (layout::print_leading_fixits): Likewise.
1653 (layout::annotation_line_showed_range_p): Likewise.
1654 (struct line_corrections): Likewise for field m_row.
1655 (line_corrections::line_corrections): Likewise for param "row".
1656 (layout::print_trailing_fixits): Likewise.
1657 (layout::get_state_at_point): Likewise.
1658 (layout::get_x_bound_for_row): Likewise.
1659 (layout::print_line): Likewise.
1660 (diagnostic_show_locus): Likewise for locals "last_line" and
1662 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1663 * input.c (selftest::test_linenum_comparisons): New function.
1664 (selftest::input_c_tests): Call it.
1665 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1666 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1667 * selftest.h (ASSERT_GT): New macro.
1668 (ASSERT_GT_AT): New macro.
1669 (ASSERT_LT): New macro.
1670 (ASSERT_LT_AT): New macro.
1672 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1674 PR rtl-optimization/84780
1675 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1677 2018-03-14 Martin Liska <mliska@suse.cz>
1679 * tree.c (record_node_allocation_statistics): Use
1680 get_stats_node_kind.
1681 (get_stats_node_kind): New function extracted from
1682 record_node_allocation_statistics.
1683 (free_node): Use get_stats_node_kind.
1685 2018-03-14 Richard Biener <rguenther@suse.de>
1687 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1688 that the value-set of ANTIC_IN doesn't grow.
1691 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1693 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1694 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1696 2018-03-14 Julia Koval <julia.koval@intel.com>
1698 * config.gcc (icelake-client, icelake-server): New.
1700 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1701 (initial_ix86_arch_features): Ditto.
1702 (PTA_SKYLAKE): Add SGX.
1703 (PTA_ICELAKE): Remove.
1704 (PTA_ICELAKE_CLIENT): New.
1705 (PTA_ICELAKE_SERVER): New.
1706 (ix86_option_override_internal): Split up icelake on icelake client and
1708 (get_builtin_code_for_version): Ditto.
1709 (fold_builtin_cpu): Ditto.
1710 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1711 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1712 * config/i386/i386.h (processor_type): Ditto.
1713 * doc/invoke.texi: Ditto.
1715 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1718 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1719 INTEGER_CST offset, add it together with bitpos / 8 and
1720 sign extend based on POINTER_SIZE.
1724 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1727 * config/i386/constraints.md (Yc): New register constraint.
1728 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1729 Use Yc constraint for alternative 2 of operand 0. Remove
1730 preferred_for_speed attribute.
1732 2018-03-14 Richard Biener <rguenther@suse.de>
1734 PR tree-optimization/84830
1735 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1736 with the old one to avoid oscillations.
1738 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1741 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1743 (assign_by_spills): Return a flag of reload assignment failure.
1744 Do not process the reload assignment failures. Do not spill other
1745 reload pseudos if they has the same reg class. Update n if
1747 (lra_assign): Add a return arg. Set up from the result of
1748 assign_by_spills call.
1749 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1750 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1751 usage_insns if it is not NULL.
1752 (spill_hard_reg_in_range): New function.
1753 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1754 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1755 function prototypes.
1756 (lra_assign): Change prototype.
1757 * lra.c (lra): Add code to deal with fails by splitting hard reg
1760 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1762 * config/riscv/riscv.opt (mrelax): New option.
1763 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1764 "norelax" when riscv_mrelax is disabled.
1765 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1767 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1770 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1771 reassociation for int modes.
1773 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1775 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1776 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1778 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1779 * config/aarch64/aarch64-sve.md
1780 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1781 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1782 (*extend<mode><Vwide>2): Rename to...
1783 (aarch64_sve_extend<mode><Vwide>2): ...this.
1784 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1785 renaming the old pattern to...
1786 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1788 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1789 define_expand, renaming the old pattern to...
1790 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1791 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1792 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1793 account when deciding which SVE instruction the optab should use.
1794 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1796 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1798 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1799 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1800 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1801 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1802 (tlsdesc_small_<mode>): Turn a define_expand and use
1803 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1804 (tlsdesc_small_advsimd_<mode>): ...this.
1805 (tlsdesc_small_sve_<mode>): New pattern.
1807 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1809 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1810 (UNSPEC_UMUL_HIGHPART): New constants.
1811 (MUL_HIGHPART): New int iteraor.
1812 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1813 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1815 (*<su>mul<mode>3_highpart): New define_insn.
1817 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1820 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1823 2018-03-13 Martin Liska <mliska@suse.cz>
1826 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1828 (sem_item_optimizer::~sem_item_optimizer): Release it.
1829 (sem_item_optimizer::merge_classes): Register variable aliases.
1830 (sem_item_optimizer::fixup_pt_set): New function.
1831 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1832 * ipa-icf.h: Declare new variables and functions.
1834 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1837 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1838 integer_pow2p@2 and test integer_pow2p in condition.
1839 (A < 0 ? C : 0): Similarly for @1.
1842 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1843 characters starting at p contain '\0' character, don't look beyond
1847 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1848 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1851 * reg-stack.c (change_stack): Change update_end var from int to
1852 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1853 also call set_block_for_insn on the newly added insns and rescan.
1856 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1857 on the last operand.
1860 * tree.c (stabilize_reference_1): Return save_expr (e) for
1861 STATEMENT_LIST even if it doesn't have side-effects.
1863 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1865 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1867 2018-03-12 Renlin Li <renlin.li@arm.com>
1869 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1870 aarch64_output_scalar_simd_mov_immediate.
1872 2018-03-12 Martin Sebor <msebor@redhat.com>
1874 PR tree-optimization/83456
1875 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1876 for perfectly overlapping calls to memcpy.
1877 (gimple_fold_builtin_memory_chk): Same.
1878 (gimple_fold_builtin_strcpy): Handle no-warning.
1879 (gimple_fold_builtin_stxcpy_chk): Same.
1880 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1882 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1884 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1885 parameter. Use it for SFmode.
1886 (rs6000_function_arg_advance_1): Adjust.
1887 (rs6000_function_arg): Adjust.
1888 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1890 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1892 PR rtl-optimization/84169
1893 PR rtl-optimization/84780
1894 * combine.c (can_combine_p): Check for a 2-insn combination whether
1895 the destination register is used between the two insns, too.
1897 2018-03-12 Richard Biener <rguenther@suse.de>
1899 PR tree-optimization/84803
1900 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1901 for refs DR analysis didn't process.
1903 2018-03-12 Richard Biener <rguenther@suse.de>
1905 PR tree-optimization/84777
1906 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1907 force-vectorize loops ignore whether we are optimizing for size.
1909 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1911 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1912 (TARGET_MD_ASM_ADJUST): Define.
1914 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1915 Kito Cheng <kito.cheng@gmail.com>
1916 Chung-Ju Wu <jasonwucj@gmail.com>
1918 * config/nds32/nds32.c (nds32_compute_stack_frame,
1919 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1920 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1921 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1922 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1923 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1924 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1925 * config/nds32/nds32.md (prologue, epilogue): Use macro
1926 NDS32_V3PUSH_AVAILABLE_P to do checking.
1928 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1931 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1932 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1933 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1934 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1935 addition of most attributes on !orig_type_die or the attribute not
1936 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1938 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1939 Chung-Ju Wu <jasonwucj@gmail.com>
1941 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1943 * config/nds32/nds32.opt (mvh): New option.
1945 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1946 Chung-Ju Wu <jasonwucj@gmail.com>
1948 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1950 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1951 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1954 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1955 Chung-Ju Wu <jasonwucj@gmail.com>
1957 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1959 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1960 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1962 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1963 Kito Cheng <kito.cheng@gmail.com>
1964 Chung-Ju Wu <jasonwucj@gmail.com>
1966 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1967 UNSPEC_FFMISM and UNSPEC_FLMISM.
1968 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1969 for ffb, ffmism and flmism.
1970 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1971 (unspec_ffmism): Ditto.
1972 (unspec_flmism): Ditto.
1973 (nds32_expand_builtin_impl): Check if string extension is available.
1974 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1975 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1977 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1980 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1983 * lra-assigns.c (assign_by_spills): Return a flag of reload
1984 assignment failure. Do not process the reload assignment
1985 failures. Do not spill other reload pseudos if they has the same
1987 (lra_assign): Add a return arg. Set up from the result of
1988 assign_by_spills call.
1989 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1990 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1991 usage_insns if it is not NULL.
1992 (spill_hard_reg_in_range): New function.
1993 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1994 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1995 function prototypes.
1996 (lra_assign): Change prototype.
1997 * lra.c (lra): Add code to deal with fails by splitting hard reg
2000 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2003 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2005 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2008 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2009 (dw_val_node): Add val_symbolic_view.
2010 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2011 (symview_upper_bound): New.
2012 (new_line_info_table): Initialize symviews_since_reset.
2013 (dwarf2out_source_line): Count symviews_since_reset and set
2014 symview_upper_bound.
2015 (dw_val_equal_p): Handle symview.
2016 (add_AT_symview): New.
2017 (print_dw_val): Handle symview.
2018 (attr_checksum, attr_checksum_ordered): Likewise.
2019 (same_dw_val_p, size_of_die): Likewise.
2020 (value_format, output_die): Likewise.
2021 (add_high_low_attributes): Use add_AT_symview for entry_view.
2022 (dwarf2out_finish): Reset symview_upper_bound, clear
2025 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2028 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2029 Add strict argument and use it.
2030 (rs6000_split_multireg_move): Update for new strict argument.
2031 (mem_operand_gpr): Disallow all non-offsettable addresses.
2032 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2034 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2037 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2038 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2039 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2042 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2043 decl, use remap_type if we want to use the type.
2045 2018-03-09 Martin Sebor <msebor@redhat.com>
2047 PR tree-optimization/84526
2048 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2050 (builtin_access::generic_overlap): Be prepared to handle non-array
2053 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2055 PR rtl-optimization/84682
2056 * lra-constraints.c (process_address_1): Check is_address flag
2057 for address constraints.
2058 (process_alt_operands): Likewise.
2059 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2060 preprocess_constraints.
2061 * recog.h (preprocess_constraints): Add oploc parameter.
2063 * recog.c (preprocess_constraints): Test address_operand for
2064 CT_ADDRESS constraints.
2066 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2069 * lra-assigns.c (assign_by_spills): Return a flag of reload
2070 assignment failure. Do not process the reload assignment
2071 failures. Do not spill other reload pseudos if they has the same
2073 (lra_assign): Add a return arg. Set up from the result of
2074 assign_by_spills call.
2075 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2076 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2077 usage_insns if it is not NULL.
2078 (spill_hard_reg_in_range): New function.
2079 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2080 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2081 function prototypes.
2082 (lra_assign): Change prototype.
2083 * lra.c (lra): Add code to deal with fails by splitting hard reg
2086 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2089 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2090 Accept complain bool parameter. Only emit errors if it is true.
2091 (arm_parse_cpu_option_name): Likewise.
2092 (arm_target_thumb_only): Adjust callers of the above.
2093 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2094 prototype to take a default true bool parameter.
2095 (arm_parse_arch_option_name): Likewise.
2097 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2098 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2102 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2103 * configure: Regenerate.
2104 * configure.ac ("linker --version-script option"): New.
2105 ("linker soname option"): New.
2107 2018-03-09 Richard Biener <rguenther@suse.de>
2109 PR tree-optimization/84775
2110 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2111 immediate uses of predicate stmts and mark them modified.
2114 PR tree-optimization/84178
2115 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2117 (version_loop_for_if_conversion): Delay update_ssa call.
2118 (tree_if_conversion): Delay update_ssa until after predicate
2121 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2124 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2125 when the function accesses prior frames.
2127 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2130 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2131 gen_llsym, otherwise call maybe_gen_llsym.
2134 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2135 has ',' character inside of it.
2137 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2140 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2141 as clobbering CC_REGNUM.
2143 2018-03-08 Richard Biener <rguenther@suse.de>
2146 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2147 (follow_copies_to_constant): Do not follow SSA names registered
2150 2018-03-08 Richard Biener <rguenther@suse.de>
2152 PR tree-optimization/84178
2153 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2155 (version_loop_for_if_conversion): Delay update_ssa call.
2156 (tree_if_conversion): Delay update_ssa until after predicate
2159 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2161 PR tree-optimization/84178
2162 * tree-if-conv.c (release_bb_predicate): Remove the
2163 the assertion that the stmts have NULL use_ops.
2164 Discard the statements, asserting that they haven't
2165 yet been added to a BB.
2167 2018-03-08 Richard Biener <rguenther@suse.de>
2169 PR tree-optimization/84746
2170 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2171 (phi_translate): Pass in destination ANTIC_OUT set.
2172 (phi_translate_1): Likewise. For a simplified result lookup
2173 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2174 (phi_translate_set): Adjust.
2175 (do_pre_regular_insertion): Likewise.
2176 (do_pre_partial_partial_insertion): Likewise.
2178 2018-03-08 Martin Liska <mliska@suse.cz>
2180 PR gcov-profile/84735
2181 * doc/gcov.texi: Document usage of profile files.
2182 * gcov-io.h: Document changes in the format.
2184 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2188 * dwarf2out.c (struct dw_line_info_table): Update comments for
2190 (FORCE_RESET_NEXT_VIEW): New.
2191 (FORCE_RESETTING_VIEW_P): New.
2192 (RESETTING_VIEW_P): Check for -1 too.
2193 (ZERO_VIEW_P): Likewise.
2194 (new_line_info_table): Force-reset next view.
2195 (dwarf2out_begin_function): Likewise.
2196 (dwarf2out_source_line): Simplify zero_view_p initialization.
2197 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2198 view directly. Omit view when omitting .loc at line 0.
2200 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2202 PR tree-optimization/84740
2203 * tree-switch-conversion.c (process_switch): Call build_constructors
2204 only if info.phi_count is non-zero.
2206 PR tree-optimization/84739
2207 * tree-tailcall.c (find_tail_calls): Check call arguments against
2208 DECL_ARGUMENTS (current_function_decl) rather than
2209 DECL_ARGUMENTS (func) when checking for tail recursion.
2211 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2213 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2214 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2215 Volker Reichelt's entry and add entries for people that perform
2216 GCC fuzzy testing and report numerous bugs.
2218 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2221 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2222 readonly data in sdata, if that is disabled.
2223 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2224 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2225 -mreadonly-in-sdata option.
2227 2018-03-07 Martin Sebor <msebor@redhat.com>
2229 PR tree-optimization/84468
2230 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2231 basic block when looking for nul assignment.
2233 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2236 * except.h (output_function_exception_table): Adjust prototype.
2237 * except.c (output_function_exception_table): Remove FNNAME parameter
2238 and add SECTION parameter. Ouput one part of the table at a time.
2239 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2240 the first part of the exception table and emit unwind directives.
2241 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2242 (i386_pe_seh_cold_init): Likewise.
2243 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2244 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2245 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2246 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2247 * config/i386/winnt.c: Include except.h.
2248 (struct seh_frame_state): Add reg_offset, after_prologue and
2249 in_cold_section fields.
2250 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2251 (i386_pe_seh_cold_init): New function.
2252 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2253 to seh->in_cold_section.
2254 (seh_emit_push): Record the offset of the push.
2255 (seh_emit_save): Record the offet of the save.
2256 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2257 Test seh->after_prologue to disregard the epilogue.
2258 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2259 (i386_pe_end_cold_function): New function.
2261 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2264 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2265 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2268 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2270 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2271 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2274 * multiple_target.c: Include tree-inline.h and intl.h.
2275 (expand_target_clones): Diagnose and fail if node->definition and
2276 !tree_versionable_function_p (node->decl).
2278 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2280 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2282 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2283 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2284 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2286 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2289 * combine.c (try_combine): Use reg_or_subregno instead of handling
2290 just paradoxical SUBREGs and REGs.
2292 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2294 * config/arc/arc.c (arc_finalize_pic): Remove function.
2295 (arc_must_save_register): We use single base PIC register, remove
2296 checks to save/restore the PIC register.
2297 (arc_expand_prologue): Likewise.
2298 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2300 (arc_verify_short): Likewise.
2301 (arc_attr_type): Likewise.
2302 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2303 (walk_stores): Likewise.
2304 (arc_address_cost): Make it static.
2305 (arc_verify_short): Likewise.
2306 (branch_dest): Likewise.
2307 (arc_attr_type): Likewise.
2308 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2309 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2310 (arc_final_prescan_insn): Remove inserting the nops due to
2311 hardware hazards. It is done in reorg step.
2312 (insn_length_variant_t): Remove.
2313 (insn_length_parameters_t): Likewise.
2314 (arc_insn_length_parameters): Likewise.
2315 (arc_get_insn_variants): Likewise.
2316 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2318 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2321 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2324 PR tree-optimization/84687
2325 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2327 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2329 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2331 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2332 Rename to ppc_speculation_barrier.
2333 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2334 __builtin_ppc_speculation_barrier.
2336 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2339 * combine.c (combine_simplify_rtx): Don't try to simplify if
2340 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2343 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2345 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2346 to 32 bytes when compiling for POWER9.
2348 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2351 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2352 regparm >= 3 with no arg reg available also for calls with
2353 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2356 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2358 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2360 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2363 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2365 2018-03-05 Richard Biener <rguenther@suse.de>
2367 PR tree-optimization/84486
2368 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2369 When inserting a __builtin_assume_aligned call set the LHS
2370 SSA name alignment info accordingly.
2372 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2374 PR tree-optimization/84114
2375 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2376 Avoid reassociation of FLOAT_MODE addition.
2378 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2380 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2381 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2382 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2383 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2384 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2385 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2386 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2388 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2389 __builtin_ia32_wbinvd): New builtins.
2390 (SPECIAL_ARGS2): New.
2391 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2392 (SPECIAL_ARGS2): New.
2393 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2394 (ix86_valid_target_attribute_inner_p): Ditto.
2395 (ix86_init_mmx_sse_builtins): Add special_args2.
2396 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2397 TARGET_WBNOINVD_P): New.
2398 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2399 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2400 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2401 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2402 * config/i386/pconfigintrin.h: New file.
2403 * config/i386/wbnoinvdintrin.h: Ditto.
2404 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2405 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2407 2018-03-05 Richard Biener <rguenther@suse.de>
2409 PR tree-optimization/84670
2410 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2412 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2413 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2414 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2415 doesn't grow if all successors (recursively) were visited at least
2418 2018-03-05 Richard Biener <rguenther@suse.de>
2420 PR tree-optimization/84650
2421 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2422 if executed in the loop pipeline.
2424 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2426 * doc/configfiles.texi (Configuration Files): Move info about
2427 conditionalizing $target-protos.h to...
2428 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2429 differs from $target-protos.h.
2431 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2432 Chung-Ju Wu <jasonwucj@gmail.com>
2434 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2435 * config/nds32/nds32-multiple.md (setmemsi): Define.
2436 * config/nds32/nds32-memory-manipulation.c
2437 (nds32_gen_dup_4_byte_to_word_value): New.
2438 (emit_setmem_word_loop): New.
2439 (emit_setmem_byte_loop): New.
2440 (nds32_expand_setmem_loop): New.
2441 (nds32_expand_setmem_loop_v3m): New.
2442 (nds32_expand_setmem_unroll): New.
2443 (nds32_expand_setmem): New.
2445 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2446 Chung-Ju Wu <jasonwucj@gmail.com>
2448 * config/nds32/nds32-memory-manipulation.c
2449 (nds32_emit_load_store): New.
2450 (nds32_emit_post_inc_load_store): New.
2451 (nds32_emit_mem_move): New.
2452 (nds32_emit_mem_move_block): New.
2453 (nds32_expand_movmemsi_loop_unknown_size): New.
2454 (nds32_expand_movmemsi_loop_known_size): New.
2455 (nds32_expand_movmemsi_loop): New.
2456 (nds32_expand_movmemsi_unroll): New.
2457 (nds32_expand_movmemqi): Rename ...
2458 (nds32_expand_movmemsi): ... to this.
2459 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2460 (movmemsi): ... to this.
2461 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2462 (nds32_expand_movmemsi): ... to this.
2464 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2465 Monk Chiang <sh.chiang04@gmail.com>
2466 Chung-Ju Wu <jasonwucj@gmail.com>
2468 * config/nds32/nds32-protos.h
2469 (nds32_expand_load_multiple): New arguments.
2470 (nds32_expand_store_multiple): Ditto.
2471 (nds32_valid_multiple_load_store): Rename ...
2472 (nds32_valid_multiple_load_store_p): ... to this.
2473 * config/nds32/nds32-memory-manipulation.c
2474 (nds32_expand_load_multiple): Refine implementation.
2475 (nds32_expand_store_multiple): Ditto.
2476 * config/nds32/nds32-multiple.md
2477 (load_multiple): Update nds32_expand_load_multiple interface.
2478 (store_multiple): Update nds32_expand_store_multiple interface.
2479 * config/nds32/nds32-predicates.c
2480 (nds32_valid_multiple_load_store): Rename ...
2481 (nds32_valid_multiple_load_store_p): ... to this and refine
2483 * config/nds32/predicates.md
2484 (nds32_load_multiple_and_update_address_operation): New predicate.
2485 (nds32_store_multiple_and_update_address_operation): New predicate.
2487 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2488 Chung-Ju Wu <jasonwucj@gmail.com>
2490 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2491 (combo): New attribute.
2492 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2494 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2496 * config/nds32/nds32.opt: Change -mcmodel= default value.
2498 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2499 Monk Chiang <sh.chiang04@gmail.com>
2500 Chung-Ju Wu <jasonwucj@gmail.com>
2502 * config/nds32/constants.md (unspec_element): New enum.
2503 * config/nds32/constraints.md (Umw): New constraint.
2504 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2505 * config/nds32/nds32-intrinsic.md: Likewise.
2506 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2507 (nds32_valid_smw_lwm_base_p): New.
2508 (nds32_output_smw_single_word): New.
2509 (nds32_output_lmw_single_word): New.
2510 (nds32_expand_unaligned_load): New.
2511 (nds32_expand_unaligned_store): New.
2512 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2513 (nds32_output_smw_single_word): Declare.
2514 (nds32_output_lmw_single_word): Declare.
2515 (nds32_expand_unaligned_load): Declare.
2516 (nds32_expand_unaligned_store): Declare.
2517 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2518 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2519 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2520 NDS32_BUILTIN_UASTORE_DW.
2521 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2524 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2525 Kito Cheng <kito.cheng@gmail.com>
2526 Chung-Ju Wu <jasonwucj@gmail.com>
2528 * config/nds32/nds32-intrinsic.c
2529 (nds32_expand_builtin_null_ftype_reg): Delete.
2530 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2531 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2532 (nds32_read_argument): New.
2533 (nds32_legitimize_target): Ditto.
2534 (nds32_legitimize_argument): Ditto.
2535 (nds32_check_constant_argument): Ditto.
2536 (nds32_expand_unop_builtin): Ditto.
2537 (nds32_expand_unopimm_builtin): Ditto.
2538 (nds32_expand_binop_builtin): Ditto.
2539 (nds32_builtin_decl_impl): Ditto.
2540 (builtin_description): Ditto.
2541 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2542 (nds32_init_builtins_impl): Ditto.
2543 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2544 (nds32_builtin_decl): New.
2545 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2546 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2548 2018-03-02 Jeff Law <law@redhat.com>
2550 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2551 (redundant_insn, fill_simple_delay_slots): Likewise.
2552 (fill_slots_from_thread): Likewise.
2553 * resource.c (mark_referenced_resources): Likewise.
2554 (mark_set_resources, find_dead_or_set_registers): Likewise.
2556 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2558 * substring-locations.h (format_warning_va): Formatting fix for
2560 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2562 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2563 * substring-locations.c: Include intl.h.
2564 (format_warning_va): Turned into small wrapper around
2565 format_warning_n_va, renamed to ...
2566 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2567 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2569 (format_warning_at_substring_n): New function.
2570 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2571 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2572 format_warning_at_substring with just a shorter name instead of
2573 const function pointer.
2574 (fmtwarn_n): New function.
2575 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2576 appropriate, get rid of all the fmtstr temporaries, move conditionals
2577 with G_() wrapped string literals directly into fmtwarn arguments,
2578 cast dir.len to (int), formatting fixes.
2580 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2582 * doc/invoke.texi: Remove "Cilk Plus" references.
2584 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2585 Richard Biener <rguenther@suse.de>
2588 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2589 for error or warning attributes if CALL_FROM_THUNK_P is set.
2592 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2595 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2596 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2599 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2600 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2602 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2603 instead of -1U in last predictors element's probability member.
2605 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2608 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2609 arguments if they are comparable.
2611 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2613 PR tree-optimization/84634
2614 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2615 masks and masked_loop_p with a single loop_masks, making sure it's
2616 null for bb vectorization.
2618 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2620 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2621 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2622 loop->force_vectorize to check whether there is no alias.
2624 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2627 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2629 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2631 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2632 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2633 instead of a loop around prev_real_insn.
2634 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2638 * config/i386/i386.c (ix86_print_operand): Use conditional
2639 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2642 2018-03-02 Richard Biener <rguenther@suse.de>
2644 PR tree-optimization/84427
2645 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2646 (bitmap_set_subtract_values): Rewrite to handle multiple
2649 (prune_clobbered_mems): Likewise.
2650 (phi_translate): Take edge instead of pred/phiblock.
2651 (phi_translate_1): Likewise.
2652 (phi_translate_set): Likewise. Insert all translated
2653 exprs for a value into the set, keeping possibly multiple
2654 expressions per value.
2655 (compute_antic_aux): Adjust for phi_translate changes.
2656 When intersecting union the expressions and prune those
2657 not in the final value set, keeping possibly multiple
2658 expressions per value. Do not use value-insertion
2659 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2660 all expressions. Add verification that the value-sets
2661 only shrink during iteration.
2662 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2663 (do_pre_regular_insertion): Likewise.
2664 (do_pre_partial_partial_insertion): Likewise.
2666 2018-03-02 Richard Biener <rguenther@suse.de>
2669 * config/darwin.c (saved_debug_info_level): New static global.
2670 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2671 (darwin_asm_lto_end): Restore debug info generation settings.
2673 2018-03-01 Martin Liska <mliska@suse.cz>
2676 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2679 2018-03-01 Richard Biener <rguenther@suse.de>
2682 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2683 type annotation with LTO when debug was disabled at compile-time.
2685 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2687 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2689 (mips_final_postscan_insn): Likewise.
2691 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2693 PR rtl-optimization/84528
2694 * alias.c (init_alias_target): Add commentary.
2695 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2696 a unique base value if the frame pointer is not eliminated
2697 to the stack pointer.
2699 2018-03-01 Tom de Vries <tom@codesourcery.com>
2701 PR rtl-optimization/83327
2702 * lra-int.h (hard_regs_spilled_into): Declare.
2703 * lra.c (hard_regs_spilled_into): Define.
2704 (init_reg_info): Init hard_regs_spilled_into.
2705 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2706 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2707 (process_bb_lives): Handle hard_regs_spilled_into.
2708 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2709 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2711 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2713 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2714 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2715 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2716 * config/rs6000/aix72.h: New file.
2718 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2720 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2721 instead of warning_at with conditional singular and plural messages
2725 * stor-layout.c (update_alignment_for_field): For
2726 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2727 && !DECL_PACKED (field), do the alignment update, just use
2728 only desired_align instead of MAX (type_align, desired_align)
2730 (place_field): Don't do known_align < desired_align handling
2731 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2732 is non-NULL, instead do it after rli->prev_field handling and
2733 only if not within a bitfield word. For DECL_PACKED (field)
2734 use type_align of BITS_PER_UNIT.
2736 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2738 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2739 superfluous parentheses and trailing spaces.
2741 2018-02-28 Richard Biener <rguenther@suse.de>
2743 PR tree-optimization/84584
2744 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2745 SCoPs with fake exit edge.
2747 2018-02-28 Martin Liska <mliska@suse.cz>
2750 * timevar.c (timer::print): Fix format to properly print 100%
2753 2018-02-28 Richard Biener <rguenther@suse.de>
2756 * genmatch.c (capture_info::walk_match): Do not mark
2757 captured expressions without operands as expr_p given
2758 they act more like predicates and should be subject to
2759 "lost tail" side-effect preserving.
2761 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2763 PR rtl-optimization/81611
2764 * auto-inc-dec.c (attempt_change): Move dead note from
2765 mem_insn if it's the next use of regno
2766 (find_address): Take address use of reg holding
2767 non-incremented value. Add parm to limit search to the named
2769 (merge_in_block): Attempt to use a mem insn that is the next
2770 use of the original regno.
2772 2018-02-27 Martin Sebor <msebor@redhat.com>
2775 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2776 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2778 2018-02-27 Martin Sebor <msebor@redhat.com>
2780 PR translation/84207
2781 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2782 n argument to unsigned HOST_WIDE_INT.
2783 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2784 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2785 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2786 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2788 2018-02-27 Richard Biener <rguenther@suse.de>
2790 PR tree-optimization/84512
2791 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2792 Do not use the estimate returned from record_stmt_cost for
2793 the scalar iteration cost but sum properly using add_stmt_cost.
2795 2018-02-27 Richard Biener <rguenther@suse.de>
2797 PR tree-optimization/84466
2798 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2799 Adjust last change to less strictly validate use operands.
2801 2018-02-27 Martin Liska <mliska@suse.cz>
2803 PR gcov-profile/84548
2804 * gcov.c (process_file): Allow partial overlap and consider it
2805 also as group functions.
2806 (output_lines): Properly calculate range of lines for a group.
2808 2018-02-27 Martin Liska <mliska@suse.cz>
2810 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2811 'ggc' suffixes. Change first column width.
2812 (timer::print): Fix formatting of the column.
2814 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2816 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2817 preserve inline entry blocks for the sake of debug inline
2818 entry point markers alone.
2819 (remove_unused_locals): Suggest in comments a better place to
2820 force the preservation of inline entry blocks that are
2821 otherwise unused, but do not preserve them.
2823 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2825 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2827 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2830 * config/i386/constraints.md (Bs): Replace
2831 ix86_indirect_branch_register with
2832 TARGET_INDIRECT_BRANCH_REGISTER.
2834 * config/i386/i386.md (indirect_jump): Likewise.
2835 (tablejump): Likewise.
2836 (*sibcall_memory): Likewise.
2837 (*sibcall_value_memory): Likewise.
2838 Peepholes of indirect call and jump via memory: Likewise.
2839 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2840 (*sibcall_value_GOT_32): Likewise.
2841 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2842 (GOT_memory_operand): Likewise.
2843 (call_insn_operand): Likewise.
2844 (sibcall_insn_operand): Likewise.
2845 (GOT32_symbol_operand): Likewise.
2846 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2848 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2850 PR rtl-optimization/83496
2851 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2852 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2853 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2854 redundant insn, if any.
2855 (relax_delay_slots): Likewise.
2856 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2858 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2860 PR tree-optimization/83965
2861 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2862 that grouped statements are part of a reduction chain. Return
2863 true if the statement is not marked as a reduction itself but
2865 (vect_recog_dot_prod_pattern): Don't check whether the statement
2866 is part of a group here.
2867 (vect_recog_sad_pattern): Likewise.
2868 (vect_recog_widen_sum_pattern): Likewise.
2870 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2873 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2875 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2878 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2880 (ix86_output_indirect_function_return): New prototype.
2881 (ix86_split_simple_return_pop_internal): Likewise.
2882 * config/i386/i386.c (indirect_return_via_cx): New.
2883 (indirect_return_via_cx_bnd): Likewise.
2884 (indirect_thunk_name): Handle return va CX_REG.
2885 (output_indirect_thunk_function): Create alias for
2886 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2887 (ix86_output_indirect_jmp): Remove the bool argument.
2888 (ix86_output_indirect_function_return): New function.
2889 (ix86_split_simple_return_pop_internal): Likewise.
2890 * config/i386/i386.md (*indirect_jump): Don't pass false
2891 to ix86_output_indirect_jmp.
2892 (*tablejump_1): Likewise.
2893 (simple_return_pop_internal): Change it to define_insn_and_split.
2894 Call ix86_split_simple_return_pop_internal to split it for
2896 (simple_return_indirect_internal): Call
2897 ix86_output_indirect_function_return instead of
2898 ix86_output_indirect_jmp.
2900 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2903 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2904 memset and value initialization afterwards.
2906 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2908 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2910 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2913 * common/config/aarch64/aarch64-common.c
2914 (aarch_option_optimization_table[]): Switch
2915 off fomit-frame-pointer
2917 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2918 Chung-Ju Wu <jasonwucj@gmail.com>
2920 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2922 (store_multiple): Ditto.
2924 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2926 * config.gcc: Add --with-cpu support for nds32 target.
2927 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2928 * config/nds32/nds32.opt: Add -mcpu= option.
2930 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2932 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2933 isel=yes): Warn for these deprecated options.
2935 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2937 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2938 ISA_2_5_MASKS_EMBEDDED.
2940 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2942 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2943 p->max as pointers rather than using iterative_hash_expr.
2945 2018-02-23 Carl Love <cel@us.ibm.com>
2947 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2948 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2950 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2951 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2952 P8V_BUILTIN_VEC_VUNSIGNED2.
2954 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2957 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2958 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2960 * lra-constraints.c (curr_insn_transform): Set up
2961 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2963 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2964 * lra-eliminations.c (spill_pseudos): Ditto.
2965 (process_insn_for_elimination): Ditto.
2966 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2967 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2968 LRA_NON_CLOBBERED_ALT.
2970 2018-02-22 Martin Sebor <msebor@redhat.com>
2972 PR tree-optimization/84480
2973 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2974 to maybe_diag_stxncpy_trunc. Call it.
2975 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2976 from gimple_fold_builtin_strcpy. Print inlining stack.
2977 (handle_builtin_stxncpy): Print inlining stack.
2978 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2980 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2983 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2984 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2985 and -fcheck-pointer-bounds are used together.
2986 (indirect_thunk_prefix): New enum.
2987 (indirect_thunk_need_prefix): New function.
2988 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2989 "_nt" instead of "_bnd" for NOTRACK prefix.
2990 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2991 (output_indirect_thunk_function): Likewise.
2993 (ix86_code_end): Update output_indirect_thunk_function calls.
2994 (ix86_output_indirect_branch_via_reg): Replace
2995 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2996 (ix86_output_indirect_branch_via_push): Likewise.
2997 (ix86_output_function_return): Likewise.
2998 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2999 incompatible with -fcf-protection=branch and
3000 -fcheck-pointer-bounds.
3002 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3005 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3006 Change gcc_assert call to output_operand_lossage.
3008 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3010 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3012 2018-02-22 DJ Delorie <dj@redhat.com>
3013 Sebastian Perta <sebastian.perta@renesas.com>
3014 Oleg Endo <olegendo@gcc.gnu.org>
3016 * config/rx/rx.c (rx_rtx_costs): New function.
3017 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3019 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3021 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3023 2018-02-22 Martin Liska <mliska@suse.cz>
3026 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3027 Add "native" as a possible value.
3029 2018-02-22 Martin Liska <mliska@suse.cz>
3032 * config/i386/i386.c (ix86_option_override_internal):
3033 Add "native" as a possible value for -march and -mtune.
3035 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3038 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3039 to all type variants.
3041 PR tree-optimization/84503
3042 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3043 width as info->bitpos + info->bitsize - start.
3044 (merged_store_group::merge_overlapping): Simplify width computation.
3045 (check_no_overlap): New function.
3046 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3047 start + width and last_order of the group, fail if check_no_overlap
3049 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3050 to group if check_no_overlap fails.
3052 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3054 * config/rs6000/altivec.md: Delete contraint arguments to
3055 define_expand, define_split, and define_peephole2, and in
3056 define_insn_and_split if always unused.
3057 * config/rs6000/darwin.md: Ditto.
3058 * config/rs6000/dfp.md: Ditto.
3059 * config/rs6000/rs6000.md: Ditto.
3060 * config/rs6000/sync.md: Ditto.
3061 * config/rs6000/vector.md: Ditto.
3062 * config/rs6000/vsx.md: Ditto.
3064 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3066 * config/rs6000/altivec.md: Write output control strings as braced
3067 blocks instead of double-quoted strings.
3068 * config/rs6000/darwin.md: Ditto.
3069 * config/rs6000/rs6000.md: Ditto.
3070 * config/rs6000/vector.md: Ditto.
3071 * config/rs6000/vsx.md: Ditto.
3073 2018-02-21 Jason Merrill <jason@redhat.com>
3075 PR c++/84314 - ICE with templates and fastcall attribute.
3076 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3078 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3080 * ipa-cp.c (determine_versionability): Fix comment typos.
3082 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3085 * ipa-cp.c (determine_versionability): Do not version functions caling
3088 2018-02-21 Martin Liska <mliska@suse.cz>
3091 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3092 Add "native" as a possible value.
3093 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3094 the macro when native cpu detection is available.
3096 2018-02-21 Martin Liska <mliska@suse.cz>
3099 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3100 Add "native" as a possible value.
3101 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3102 when native cpu detection is available.
3104 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3105 Martin Sebor <msebor@redhat.com>
3107 PR tree-optimization/84478
3108 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3110 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3111 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3112 support which is conservatively correct, for 2 only stay conservative
3113 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3114 argument to the 2 argument get_range_strlen, adjust 6 arg
3115 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3117 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3118 (gimple_fold_builtin_strlen): Pass true as last argument to
3121 2018-02-20 Martin Sebor <msebor@redhat.com>
3124 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3125 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3126 (builtin_memref::builtin_memref): Factor out parts into
3127 set_base_and_offset and call it.
3129 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3132 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3133 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3134 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3135 search at the associated MODE_INT.
3137 2018-02-20 Jeff Law <law@redhat.com>
3140 PR tree-optimization/81592
3142 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3143 for range data rather than using global data.
3144 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3145 range data rather than using global data.
3146 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3147 pass it to children as needed.
3148 (struct directive::fmtresult): Similarly.
3149 (struct directive::set_width): Similarly.
3150 (struct directive::set_precision): Similarly.
3151 (format_integer, format_directive, parse_directive): Similarly.
3152 (format_none): Accept unnamed vr_values parameter.
3153 (format_percent, format_floating, format_character): Similarly.
3154 (format_string, format_plain): Similarly.
3155 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3156 the EVRP range analyzer for range data rather than using global data.
3157 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3158 gimple-ssa-evrp-analyze.h
3159 (class sprintf_dom_walker): Add after_dom_children member function.
3160 Add evrp_range_analyzer member.
3161 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3162 range analyzer as needed.
3163 (sprintf_dom_walker::after_dom_children): New member function.
3164 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3166 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3167 (evrp_range_analyzer::pop_to_marker): Likewise.
3169 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3171 PR tree-optimization/84419
3172 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3173 with the required type if its current type is compatible but
3176 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3179 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3180 after vectorization.
3182 2018-02-20 Martin Liska <mliska@suse.cz>
3185 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3186 possible values if we don't have a hint.
3188 2018-02-20 Martin Liska <mliska@suse.cz>
3192 * final.c (shorten_branches): Build align_tab array with one
3194 * opts.c (finish_options): Add alignment option limit check.
3195 (MAX_CODE_ALIGN): Likewise.
3196 (MAX_CODE_ALIGN_VALUE): Likewise.
3197 * doc/invoke.texi: Document maximum allowed option value for
3198 all -falign-* options.
3200 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3203 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3204 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3205 * var-tracking.c (emit_note_insn_var_location): Remove all references
3206 to NOTE_INSN_CALL_ARG_LOCATION.
3207 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3208 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3209 Use copy_rtx_if_shared.
3210 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3211 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3212 (dwarf2out_var_location): Remove handling of
3213 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3215 * final.c (final_scan_insn): Remove all references to
3216 NOTE_INSN_CALL_ARG_LOCATION.
3217 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3218 before dumping final insns.
3219 * except.c (emit_note_eh_region_end): Remove all references to
3220 NOTE_INSN_CALL_ARG_LOCATION.
3221 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3222 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3223 * config/arc/arc.c (hwloop_optimize): Likewise.
3224 * config/arm/arm.c (create_fix_barrier): Likewise.
3225 * config/s390/s390.c (s390_chunkify_start): Likewise.
3226 * config/sh/sh.c (find_barrier): Likewise.
3227 * config/i386/i386.c (rest_of_insert_endbranch,
3228 ix86_seh_fixup_eh_fallthru): Likewise.
3229 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3230 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3231 * config/frv/frv.c (frv_function_prologue): Likewise.
3232 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3234 (note_outside_basic_block_p): Remove all references to
3235 NOTE_INSN_CALL_ARG_LOCATION.
3236 * gengtype.c (adjust_field_rtx_def): Likewise.
3237 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3239 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3240 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3243 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3246 PR tree-optimization/84452
3247 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3248 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3251 2018-02-19 Martin Liska <mliska@suse.cz>
3254 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3256 2018-02-19 Martin Liska <mliska@suse.cz>
3257 Richard Sandiford <richard.sandiford@linaro.org>
3259 PR tree-optimization/82491
3260 * gimple-fold.c (get_base_constructor): Make earlier bail out
3263 2018-02-19 Carl Love <cel@us.ibm.com>
3265 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3266 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3268 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3269 P8V_BUILTIN_VEC_NEG.
3271 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3273 * config/rl78/rl78.md (movdf): New define expand.
3275 2018-02-19 Martin Liska <mliska@suse.cz>
3278 * doc/invoke.texi: Fix typo.
3279 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3281 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3283 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3284 handle rs6000_single_float and rs6000_double_float specially for
3287 2018-02-16 Jeff Law <law@redhat.com>
3289 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3292 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3295 * ipa-inline.c (inline_small_functions): Fix a typo.
3297 2018-02-16 Nathan Sidwell <nathan@acm.org>
3299 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3301 2018-02-16 Carl Love <cel@us.ibm.com>
3303 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3304 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3305 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3306 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3307 expansion to P8V_BUILTIN_VEC_FLOAT2.
3309 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3311 PR rtl-optimization/70023
3312 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3313 src_regno into account.
3315 2018-02-16 Carl Love <cel@us.ibm.com>
3317 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3318 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3319 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3320 * config/rs6000/rs6000.c: Remove case statements for
3321 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3322 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3323 and P9V_BUILTIN_VEC_VINSERT4B.
3324 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3325 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3326 * config/rs6000/vsx.md:
3327 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3330 2018-02-16 Carl Love <cel@us.ibm.com>
3332 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3334 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3336 * config/rs6000/rs6000-c.c: Add the definitions for
3337 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3338 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3339 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3340 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3341 definition for insert4b and define insn *insert3b_internal.
3342 * doc/extend.texi: Add documentation for vec_extract4b.
3344 2018-02-16 Nathan Sidwell <nathan@acm.org>
3346 * doc/extend.texi (Backwards Compatibility): Mention friend
3347 injection. Note for-scope is deprecated.
3348 * doc/invoke.texi (-ffriend-injection): Deprecate.
3350 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3352 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3353 that moved to I2, also allow destinations that are a paradoxical
3354 subreg (instead of a normal reg).
3356 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3359 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3362 2018-02-16 Richard Biener <rguenther@suse.de>
3364 PR tree-optimization/84037
3365 PR tree-optimization/84016
3367 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3368 Adjust vec_construct for the fact we need additional higher latency
3369 128bit inserts for AVX256 and AVX512 vector builds.
3370 (ix86_add_stmt_cost): Scale vector construction cost for
3373 2018-02-16 Richard Biener <rguenther@suse.de>
3375 PR tree-optimization/84417
3376 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3377 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3378 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3380 2018-02-16 Martin Liska <mliska@suse.cz>
3383 * internal-fn.def (ASAN_CHECK): Set proper flags.
3384 (ASAN_MARK): Likewise.
3386 2018-02-16 Julia Koval <julia.koval@intel.com>
3388 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3389 from PTA_CANNONLAKE.
3391 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3394 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3395 Use ++iter rather than iter++ for std::list iterators.
3396 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3397 defer deleting them until all nodes in the forest are processed. Do
3398 free even leaf nodes. Change to_process into auto_vec.
3401 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3402 * vec.h (vec_default_construct): Use memset instead of placement new
3403 if BROKEN_VALUE_INITIALIZATION is defined.
3404 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3405 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3408 PR rtl-optimization/83723
3409 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3410 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3411 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3413 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3415 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3417 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3419 PR rtl-optimization/81443
3420 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3421 from inner REGs to paradoxical SUBREGs.
3423 2018-02-16 Richard Biener <rguenther@suse.de>
3425 PR tree-optimization/84399
3426 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3427 For operands we can analyze at their definition make sure we can
3428 analyze them at each use as well.
3430 2018-02-16 Richard Biener <rguenther@suse.de>
3432 PR tree-optimization/84190
3433 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3434 volatile accesses if the decl isn't volatile.
3436 2018-02-15 Jason Merrill <jason@redhat.com>
3438 PR c++/84314 - ICE with templates and fastcall attribute.
3439 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3440 TYPE_CANONICAL on an existing type.
3442 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3444 PR tree-optimization/84383
3445 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3446 dstoff nor call operand_equal_p if dstbase is NULL.
3448 PR tree-optimization/84334
3449 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3450 also a CONSTANT_CLASS_P, punt.
3452 2018-02-14 Jim Wilson <jimw@sifive.com>
3454 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3455 first SMALL_OPERAND check. New local min_second_step. Move assert
3456 to where locals are set. Add TARGET_RVC support.
3457 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3459 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3461 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3463 2018-02-14 Martin Sebor <msebor@redhat.com>
3465 PR tree-optimization/83698
3466 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3467 arrays constrain the offset range to their bounds.
3468 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3469 (builtin_access::overlap): Avoid setting the size of overlap if it's
3471 (maybe_diag_overlap): Also consider arrays when deciding what values
3472 of offsets to include in diagnostics.
3474 2018-02-14 Martin Sebor <msebor@redhat.com>
3477 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3478 that correspond to the kind of a declaration.
3480 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3483 * config/pa/pa.md: Load address of PIC label using the linkage table
3484 if the label is nonlocal.
3486 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3488 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3489 warning message if user requests -maltivec=be.
3490 * doc/invoke.texi: Document deprecation of -maltivec=be.
3492 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3495 * config/rs6000/rs6000-c.c: Update definitions for
3496 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3497 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3499 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3502 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3503 add _get_ssp intrinsics. Remove argument from
3504 __builtin_ia32_rdssp[d|q].
3505 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3506 * config/i386/i386-builtin.def: Remove argument from
3507 __builtin_ia32_rdssp[d|q].
3508 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3509 ix86_expand_special_args_builtin for _rdssp[d|q].
3510 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3511 Clear register before usage.
3512 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3513 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3515 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3517 PR tree-optimization/84357
3518 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3519 operand 1 of an ARRAY_REF too.
3521 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3524 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3525 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3527 (set_of_reg): New struct.
3528 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3529 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3530 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3532 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3533 Split into bitclr, bitset, bitinvert patterns if appropriate.
3534 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3535 use rx_fuse_in_memory_bitop.
3536 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3537 to named insn, correct maximum insn length.
3539 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3542 * machmode.def: Define a complex mode for PARTIAL_INT.
3543 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3545 * doc/rtl.texi: Document CSPImode.
3546 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3548 (msp430_hard_regno_nregs_with_padding): Likewise.
3550 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3553 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3555 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3557 PR rtl-optimization/84169
3558 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3559 we generated a parallel as new i3 and we split that to new i2 and i3
3560 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3561 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3562 those to i2, not i1. Partially rewrite this scan code.
3564 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3567 * stor-layout.c (place_field): For variable length fields, adjust
3568 offset_align afterwards not just based on the field's alignment,
3569 but also on the size.
3572 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3573 of exps and logs in the use_exp2 case.
3575 2018-02-13 Jeff Law <law@redhat.com>
3577 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3580 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3583 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3587 * common.opt (gas-loc-support, gas-locview-support): New.
3588 (ginline-points, ginternal-reset-location-views): New.
3589 * doc/invoke.texi: Document them. Use @itemx where intended.
3590 (gvariable-location-views): Adjust.
3591 * target.def (reset_location_view): New.
3592 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3593 (TARGET_RESET_LOCATION_VIEW): New.
3594 * doc/tm.texi: Rebuilt.
3595 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3596 (dwarf2out_default_as_locview_support): New.
3597 (output_asm_line_debug_info): Use option variables.
3598 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3599 (output_loc_list): Likewise.
3600 (add_high_low_attributes): Check option variables.
3601 Don't output entry view attribute in strict mode.
3602 (gen_inlined_subroutine_die): Check option variables.
3603 (dwarf2out_inline_entry): Likewise.
3604 (init_sections_and_labels): Likewise.
3605 (dwarf2out_early_finish): Likewise.
3606 (maybe_reset_location_view): New, from...
3607 (dwarf2out_var_location): ... here. Call it.
3608 * debug.h (dwarf2out_default_as_loc_support): Declare.
3609 (dwarf2out_default_as_locview_support): Declare.
3610 * hooks.c (hook_int_rtx_insn_0): New.
3611 * hooks.h (hook_int_rtx_insn_0): Declare.
3612 * toplev.c (process_options): Take -gas-loc-support and
3613 -gas-locview-support from dwarf2out. Enable
3614 -gvariable-location-views by default only with locview
3615 assembler support. Enable -ginternal-reset-location-views by
3616 default only if the target defines the corresponding hook.
3617 Enable -ginline-points by default if location views are
3618 enabled; force it disabled if statement frontiers are
3620 * tree-inline.c (expand_call_inline): Check option variables.
3621 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3623 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3625 PR tree-optimization/84321
3626 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3627 handling. Also check whether the anti-range contains any values
3628 that satisfy the mask; switch to a VR_RANGE if not.
3630 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3633 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3635 2018-02-13 Martin Jambor <mjambor@suse.cz>
3638 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3639 of call statements, also set location of a load to a temporary.
3641 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3643 * config/rl78/rl78.c (add_vector_labels): New function.
3644 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3645 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3646 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3647 which checks that no arguments are passed.
3648 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3649 * doc/extend.texi: Documentation for the new attribute.
3651 2018-02-13 Andreas Schwab <schwab@suse.de>
3653 * config/riscv/linux.h (CPP_SPEC): Define.
3655 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3658 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3659 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3660 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3661 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3662 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3663 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3664 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3665 temporarily for AES and PCLMUL builtins.
3667 PR tree-optimization/84339
3668 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3669 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3673 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3674 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3675 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3677 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3679 * omp-simd-clone.h: New file.
3680 * omp-simd-clone.c: Include omp-simd-clone.h.
3681 (expand_simd_clones): No longer static.
3682 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3683 cgraph.h and omp-simd-clone.h.
3684 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3685 (vect_recog_widen_shift_pattern): Formatting fix.
3686 (vect_pattern_recog_1): Don't check optab for calls.
3689 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3690 operands[2] into a REG before using gen_lowpart on it.
3692 2018-02-12 Jeff Law <law@redhat.com>
3695 * config/sh/sh.c (find_barrier): Consider a sibling call
3698 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3699 successfully back substituting a reg.
3701 2018-02-12 Richard Biener <rguenther@suse.de>
3703 PR tree-optimization/84037
3704 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3705 parameter, move visited init to caller.
3706 (vect_slp_analyze_operations): Separate cost from validity
3707 check, initialize visited once for all instances.
3708 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3710 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3712 (vect_model_promotion_demotion_cost): Likewise.
3713 (vectorizable_bswap): Guard cost modeling with !slp_node
3714 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3716 (vectorizable_call): Likewise.
3717 (vectorizable_conversion): Likewise.
3718 (vectorizable_assignment): Likewise.
3719 (vectorizable_shift): Likewise.
3720 (vectorizable_operation): Likewise.
3721 (vectorizable_store): Likewise.
3722 (vectorizable_load): Likewise.
3723 (vectorizable_condition): Likewise.
3724 (vectorizable_comparison): Likewise.
3726 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3729 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3730 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3731 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3733 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3736 * params.def (inline-min-speedup): Increase from 8 to 15.
3737 (max-inline-insns-auto): Decrease from 40 to 30.
3738 * ipa-split.c (consider_split): Add some buffer for function to
3739 be considered inlining candidate.
3740 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3743 2018-02-12 Richard Biener <rguenther@suse.de>
3745 PR tree-optimization/84037
3746 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3747 matched stmts if we cannot swap the non-matched ones.
3749 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3751 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3752 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3753 _mm_maskz_scalef_round_ss): New intrinsics.
3754 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3755 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3756 __builtin_ia32_scalefss_round): Remove.
3757 (__builtin_ia32_scalefsd_mask_round,
3758 __builtin_ia32_scalefss_mask_round): New intrinsics.
3759 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3760 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3761 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3762 "<round_constraint>")): Changed to ...
3763 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3764 "<round_scalar_constraint>")): ... this.
3765 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3766 %0, %1, %2<round_op3>}"): Changed to ...
3767 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3768 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3769 %2<round_scalar_mask_op3>}"): ... this.
3770 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3772 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3774 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3775 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3776 (_mm_maskz_sqrt_round_ss): New intrinsics.
3777 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3778 (__builtin_ia32_sqrtsd_mask_round)
3779 (__builtin_ia32_sqrtss_mask_round): New builtins.
3780 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3781 (__builtin_ia32_sqrtss_round): Remove.
3782 (__builtin_ia32_sqrtsd_mask_round)
3783 (__builtin_ia32_sqrtss_mask_round): New builtins.
3784 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3785 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3786 ((match_operand:VF_128 1 "vector_operand"
3787 "xBm,<round_constraint>")): Changed to ...
3788 ((match_operand:VF_128 1 "vector_operand"
3789 "xBm,<round_scalar_constraint>")): ... this.
3790 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3791 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3792 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3793 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3794 %<iptr>1<round_scalar_mask_op3>}): ... this.
3795 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3796 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3798 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3801 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3802 Cast vec_cmpeq result to correct type.
3803 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3804 Cast vec_cmpgt result to correct type.
3806 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3808 * final.c (final_scan_insn_1): Renamed from...
3809 (final_scan_insn): ... this. New wrapper, to recover
3810 seen from the outermost call in recursive ones.
3811 * config/sparc/sparc.c (output_return): Drop seen from call.
3812 (output_sibcall): Likewise.
3813 * config/visium/visium.c (output_branch): Likewise.
3815 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3817 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3820 2018-02-10 Alan Modra <amodra@gmail.com>
3823 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3824 Specify LR as an input.
3826 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3829 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3830 remove_member_access_dummy_vars): New functions.
3831 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3832 lower_omp_1, execute_lower_omp): Use them.
3834 PR rtl-optimization/84308
3835 * shrink-wrap.c (spread_components): Release todo vector.
3837 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3839 PR rtl-optimization/57193
3840 * ira-color.c (struct allocno_color_data): Add member
3841 conflict_allocno_hard_prefs.
3842 (update_conflict_allocno_hard_prefs): New.
3843 (bucket_allocno_compare_func): Add a preference based on
3844 conflict_allocno_hard_prefs.
3845 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3846 (color_allocnos): Remove a dead code. Initiate
3847 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3849 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3852 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3853 constraint from =wa to wa. Avoid a subreg on the output operand,
3854 instead use a pseudo and subreg it in a move.
3855 (p9_xxbrd_<mode>): Changed to ...
3856 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3857 (p9_xxbrd_v2df): New expander.
3858 (p9_xxbrw_<mode>): Changed to ...
3859 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3860 (p9_xxbrw_v4sf): New expander.
3862 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3864 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3866 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3869 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3870 multiply in 32-bit mode.
3871 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3872 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3875 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3877 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3878 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3879 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3880 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3882 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3885 * dwarf2out.c (is_trivial_indirect_ref): New function.
3886 (dwarf2out_late_global_decl): Do not generate a location
3887 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3888 and that are not defined in the current unit.
3890 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3892 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3893 instead of a libcall for UNORDERED.
3895 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3898 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3899 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3901 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3904 * config/s390/s390.c (s390_set_current_function): Invoke
3905 s390_indirect_branch_settings also if fndecl didn't change.
3907 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3909 * config/rs6000/rs6000.md (blockage): Set length to zero.
3911 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3913 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3915 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3918 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3919 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3923 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3924 PARALLEL incoming that failed vt_get_decl_and_offset check.
3927 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3928 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3930 (get_variable_section): For decls in named .bss* sections pass true as
3931 second argument to bss_initializer_p.
3933 2018-02-09 Marek Polacek <polacek@redhat.com>
3934 Jakub Jelinek <jakub@redhat.com>
3937 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3938 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3939 Sync some changes from cxx_fold_indirect_ref.
3941 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3943 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3945 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3946 (BLOCK_INLINE_ENTRY_LABEL): New.
3947 (dwarf2out_var_location): Disregard inline entry markers.
3948 (inline_entry_data): New struct.
3949 (inline_entry_data_hasher): New hashtable type.
3950 (inline_entry_data_hasher::hash): New.
3951 (inline_entry_data_hasher::equal): New.
3952 (inline_entry_data_table): New variable.
3953 (add_high_low_attributes): Add DW_AT_entry_pc and
3954 DW_AT_GNU_entry_view attributes if a pending entry is found
3955 in inline_entry_data_table. Add old entry_pc attribute only
3956 if debug nonbinding markers are disabled.
3957 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3958 markers are enabled.
3959 (block_within_block_p, dwarf2out_inline_entry): New.
3960 (dwarf2out_finish): Check that no entries remained in
3961 inline_entry_data_table.
3962 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3963 (final_scan_insn, notice_source_line): Likewise.
3964 (rest_of_clean_state): Skip inline entry markers.
3965 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3967 * gimple.c (gimple_build_debug_inline_entry): New.
3968 * gimple.h (enum gimple_debug_subcode): Add
3969 GIMPLE_DEBUG_INLINE_ENTRY.
3970 (gimple_build_debug_inline_entry): Declare.
3971 (gimple_debug_inline_entry_p): New.
3972 (gimple_debug_nonbind_marker_p): Adjust.
3973 * insn-notes.def (INLINE_ENTRY): New.
3974 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3975 inline entry marker notes.
3976 (print_insn): Likewise.
3977 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3978 (INSN_DEBUG_MARKER_KIND): Likewise.
3979 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3980 * tree-inline.c (expand_call_inline): Build and insert
3981 debug_inline_entry stmt.
3982 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3983 inline entry blocks early, if nonbind markers are enabled.
3984 (dump_scope_block): Dump fragment info.
3985 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3986 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3987 (gimple_build_debug_inline_entry): New.
3988 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3989 Enable/disable inline entry points too.
3990 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3991 (DEBUG_INSN): Describe inline entry markers.
3993 * common.opt (gvariable-location-views): New.
3994 (gvariable-location-views=incompat5): New.
3995 * config.in: Rebuilt.
3996 * configure: Rebuilt.
3997 * configure.ac: Test assembler for view support.
3998 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3999 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4000 * dwarf2out.c (var_loc_view): New typedef.
4001 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4002 (dwarf2out_locviews_in_attribute): New.
4003 (dwarf2out_locviews_in_loclist): New.
4004 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4005 (enum dw_line_info_opcode): Add LI_adv_address.
4006 (struct dw_line_info_table): Add view.
4007 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4008 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4009 (zero_view_p): New variable.
4010 (ZERO_VIEW_P): New macro.
4011 (output_asm_line_debug_info): New.
4012 (struct var_loc_node): Add view.
4013 (add_AT_view_list, AT_loc_list): New.
4014 (add_var_loc_to_decl): Add view param. Test it against last.
4015 (new_loc_list): Add view params. Record them.
4016 (AT_loc_list_ptr): Handle loc and view lists.
4017 (view_list_to_loc_list_val_node): New.
4018 (print_dw_val): Handle dw_val_class_view_list.
4019 (size_of_die): Likewise.
4020 (value_format): Likewise.
4021 (loc_list_has_views): New.
4022 (gen_llsym): Set vl_symbol too.
4023 (maybe_gen_llsym, skip_loc_list_entry): New.
4024 (dwarf2out_maybe_output_loclist_view_pair): New.
4025 (output_loc_list): Output view list or entries too.
4026 (output_view_list_offset): New.
4027 (output_die): Handle dw_val_class_view_list.
4028 (output_dwarf_version): New.
4029 (output_compilation_unit_header): Use it.
4030 (output_skeleton_debug_sections): Likewise.
4031 (output_rnglists, output_line_info): Likewise.
4032 (output_pubnames, output_aranges): Update version comments.
4033 (output_one_line_info_table): Output view numbers in asm comments.
4034 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4035 Call maybe_gen_llsym.
4036 (loc_list_from_tree_1): Adjust.
4037 (add_AT_location_description): Create view list attribute if
4038 needed, check it's absent otherwise.
4039 (convert_cfa_to_fb_loc_list): Adjust.
4040 (maybe_emit_file): Call output_asm_line_debug_info for test.
4041 (dwarf2out_var_location): Reset views as needed. Precompute
4042 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4043 attribute. Set view.
4044 (new_line_info_table): Reset next view.
4045 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4046 (dwarf2out_source_line): Likewise. Output view resets and labels to
4047 the assembler, or select appropriate line info opcodes.
4048 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4049 (optimize_string_length): Catch it. Adjust.
4050 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4051 dw_val_class_view_list, and remove it if no longer needed.
4052 (hash_loc_list): Hash view numbers.
4053 (loc_list_hasher::equal): Compare them.
4054 (optimize_location_lists): Check whether a view list symbol is
4055 needed, and whether the locview attribute is present, and
4056 whether they match. Remove the locview attribute if no longer
4058 (index_location_lists): Call skip_loc_list_entry for test.
4059 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4060 Use output_dwarf_version.
4061 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4062 (struct dw_val_node): Add val_view_list.
4063 * final.c (SEEN_NEXT_VIEW): New.
4064 (set_next_view_needed): New.
4065 (clear_next_view_needed): New.
4066 (maybe_output_next_view): New.
4067 (final_start_function): Rename to...
4068 (final_start_function_1): ... this. Take pointer to FIRST,
4069 add SEEN parameter. Emit param bindings in the initial view.
4070 (final_start_function): Reintroduce SEEN-less interface.
4071 (final): Rename to...
4072 (final_1): ... this. Take SEEN parameter. Output final pending
4073 next view at the end.
4074 (final): Reintroduce seen-less interface.
4075 (final_scan_insn): Output pending next view before switching
4076 sections or ending a block. Mark the next view as needed when
4077 outputting variable locations. Notify debug backend of section
4078 changes, and of location view changes.
4079 (rest_of_handle_final): Adjust.
4080 * toplev.c (process_options): Autodetect value for debug variable
4081 location views option. Warn on incompat5 without -gdwarf-5.
4082 * doc/invoke.texi (gvariable-location-views): New.
4083 (gvariable-location-views=incompat5): New.
4084 (gno-variable-location-views): New.
4086 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4088 PR tree-optimization/84136
4089 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4090 that the result of find_edge is non-NULL.
4092 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4095 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4096 storing integer register in SImode. Fix cost of 256 and 512
4097 byte aligned SSE register store.
4099 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4101 * config/i386/i386.c (ix86_multiplication_cost): Fix
4102 multiplication cost for TARGET_AVX512DQ.
4104 2018-02-08 Marek Polacek <polacek@redhat.com>
4106 PR tree-optimization/84238
4107 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4110 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4112 PR tree-optimization/84265
4113 * tree-vect-stmts.c (vectorizable_store): Don't treat
4114 VMAT_CONTIGUOUS accesses as grouped.
4115 (vectorizable_load): Likewise.
4117 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4119 PR tree-optimization/81635
4120 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4121 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4122 (test_round_for_mask): New functions.
4123 (wide_int_cc_tests): Call test_round_for_mask.
4124 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4125 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4126 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4127 range returned by get_range_info.
4129 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4132 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4133 * symtab.c: Include builtins.h
4134 (symtab_node::output_to_lto_symbol_table_p): Move here
4135 from lto-streamer-out.c:output_symbol_p.
4136 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4137 (output_symbol_p): Move all logic to symtab.c
4138 (produce_symtab): Update.
4140 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4142 * config/s390/s390-opts.h (enum indirect_branch): Define.
4143 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4144 (s390_indirect_branch_via_thunk)
4145 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4146 (enum s390_indirect_branch_type): Define.
4147 * config/s390/s390.c (struct s390_frame_layout, struct
4148 machine_function): Remove.
4149 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4150 (indirect_branch_table_label_no, indirect_branch_table_name):
4152 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4153 (enum s390_indirect_branch_option): Define.
4154 (s390_return_addr_from_memory): New function.
4155 (s390_handle_string_attribute): New function.
4156 (s390_attribute_table): Add new attribute handler.
4157 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4158 (s390_indirect_branch_via_thunk): New function.
4159 (s390_indirect_branch_via_inline_thunk): New function.
4160 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4161 sibling call optimization for non z10 compiles.
4162 (s390_emit_call): Force indirect branch target to be a single
4163 register. Add r1 clobber for non-z10 compiles.
4164 (s390_emit_epilogue): Emit return jump via return_use expander.
4165 (s390_reorg): Handle JUMP_INSNs as execute targets.
4166 (s390_option_override_internal): Perform validity checks for the
4167 new command line options.
4168 (s390_indirect_branch_attrvalue): New function.
4169 (s390_indirect_branch_settings): New function.
4170 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4171 (s390_output_indirect_thunk_function): New function.
4172 (s390_code_end): Implement target hook.
4173 (s390_case_values_threshold): Implement target hook.
4174 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4176 * config/s390/s390.h (struct s390_frame_layout)
4177 (struct machine_function): Move here from s390.c.
4178 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4179 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4180 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4181 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4182 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4183 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4184 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4185 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4186 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4187 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4188 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4189 (mnemonic attribute): Add values which aren't recognized
4191 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4192 pattern for branch conversion. Fix mnemonic attribute.
4193 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4194 indirect branch via thunk if requested.
4195 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4196 ("*indirect_jump"): Disable for branch conversion using out of
4198 ("indirect_jump_via_thunk<mode>_z10")
4199 ("indirect_jump_via_thunk<mode>")
4200 ("indirect_jump_via_inlinethunk<mode>_z10")
4201 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4202 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4203 ("casesi_jump_via_inlinethunk<mode>_z10")
4204 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4205 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4206 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4207 ("*indirect2_jump"): Disable for branch conversion.
4208 ("casesi_jump"): Turn into expander and expand patterns for branch
4210 ("return_use"): New expander.
4211 ("*return"): Emit return via thunk and rename it to ...
4212 ("*return<mode>"): ... this one.
4213 * config/s390/s390.opt: Add new options and and enum for the
4216 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4218 * lra-constraints.c (match_reload): Unconditionally use
4219 gen_lowpart_SUBREG, rather than selecting between that
4220 and equivalent gen_rtx_SUBREG code.
4222 2018-02-08 Richard Biener <rguenther@suse.de>
4224 PR tree-optimization/84233
4225 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4226 changed flag instead of boguously re-using phi_inserted.
4228 2018-02-08 Martin Jambor <mjambor@suse.cz>
4230 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4231 static local variables.
4233 2018-02-08 Richard Biener <rguenther@suse.de>
4235 PR tree-optimization/84278
4236 * tree-vect-stmts.c (vectorizable_store): When looking for
4237 smaller vector types to perform grouped strided loads/stores
4238 make sure the mode is supported by the target.
4239 (vectorizable_load): Likewise.
4241 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4243 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4244 Increase LDP/STP opportunities by adding adjacent callee-saves.
4246 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4248 PR rtl-optimization/84068
4249 PR rtl-optimization/83459
4250 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4252 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4254 PR tree-optimization/84224
4255 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4256 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4259 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4262 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4263 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4264 count, remove one USE.
4266 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4268 * doc/install.texi (Configuration): Document the
4269 --with-long-double-format={ibm,ieee} PowerPC configuration
4273 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4274 Convert from define_expand to be define_insn_and_split. Rework
4275 float/double/_Float128 conversions to QI/HI/SImode to work with
4276 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4277 conversions to QI/HImode types did a store and then a load to
4278 truncate the value. For conversions to VSX registers, don't split
4279 the insn, instead emit the code directly. Use the code iterator
4280 any_fix to combine signed and unsigned conversions.
4281 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4282 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4283 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4284 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4285 (fix_<mode>di2_hw): Likewise.
4286 (fixuns_<mode>di2_hw): Likewise.
4287 (fix_<mode>si2_hw): Likewise.
4288 (fixuns_<mode>si2_hw): Likewise.
4289 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4290 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4291 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4292 fix<uns>_trunc<SFDF:mode>si2_p8.
4293 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4295 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4296 (fix<uns>_<mode>_mem): Likewise.
4297 (fctiw<u>z_<mode>_mem): Likewise.
4298 (fix<uns>_<mode>_mem): Likewise.
4299 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4300 the register allocator from doing a direct move to the GPRs to do
4301 a store, and instead use the ISA 3.0 store byte/half-word from
4302 vector register instruction. For IEEE 128-bit floating point,
4303 also optimize stores of 32-bit ints.
4304 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4306 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4308 * genextract.c (push_pathstr_operand): New function to support
4310 (walk_rtx): Call push_pathstr_operand.
4311 (print_path): Support [a-zA-Z].
4313 2018-02-07 Richard Biener <rguenther@suse.de>
4315 PR tree-optimization/84037
4316 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4317 (cse_and_gimplify_to_preheader): Declare.
4318 (vect_get_place_in_interleaving_chain): Likewise.
4319 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4321 (_loop_vec_info::~_loop_vec_info): Delete it.
4322 (cse_and_gimplify_to_preheader): New function.
4323 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4324 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4325 (vectorizable_load): Likewise. For grouped stores always base
4326 the IV on the first element.
4327 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4328 condition before gimplifying.
4330 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4332 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4333 *DIV_EXPR and *MOD_EXPR.
4335 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4338 * config/i386/i386.c (ix86_option_override_internal): Mask out
4339 the CF_SET bit when checking -fcf-protection.
4341 2018-02-07 Tom de Vries <tom@codesourcery.com>
4344 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4347 2018-02-07 Richard Biener <rguenther@suse.de>
4349 PR tree-optimization/84204
4350 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4353 PR tree-optimization/84205
4354 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4355 special-case isl_ast_op_zdiv_r.
4357 PR tree-optimization/84223
4358 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4359 Only add conditions from within the region.
4360 (gather_bbs::after_dom_children): Adjust.
4362 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4365 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4366 * config/avr/avr.md: Only post-reload split REG-REG moves if
4367 either register is GENERAL_REG_P.
4369 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4371 PR tree-optimization/84235
4372 * tree-ssa-scopedtables.c
4373 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4374 if the subtraction is performed in floating point type where NaNs are
4375 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4376 build 1. Formatting fix.
4378 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4381 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4382 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4383 and skip it regardless of bb boundaries. Use CALL_P macro,
4384 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4385 unnecessarily, formatting fix.
4387 2018-02-06 Michael Collison <michael.collison@arm.com>
4389 * config/arm/thumb2.md:
4390 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4391 (*thumb_mov_notscc): Ditto.
4393 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4396 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4397 unsigned_fix, not "s".
4399 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4401 * configure.ac (gcc_fn_eh_frame_ro): New function.
4402 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4403 correct .eh_frame permissions.
4404 * configure: Regenerate.
4406 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4408 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4411 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4413 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4414 Display warning message for -mno-speculate-indirect-jumps.
4416 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4418 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4420 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4422 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4424 PR tree-optimization/84225
4425 * tree-eh.c (find_trapping_overflow): Only call
4426 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4428 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4431 * config/i386/i386.c: Reimplement the check of possible options
4432 -mibt/-mshstk conbination. Change error messages.
4433 * doc/invoke.texi: Fix a typo: remove extra '='.
4435 2018-02-06 Marek Polacek <polacek@redhat.com>
4437 PR tree-optimization/84228
4438 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4440 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4443 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4444 emitted arch directives.
4445 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4446 __ARM_FEATURE_COPROC before changing architectures.
4448 2018-02-06 Richard Biener <rguenther@suse.de>
4450 * config/i386/i386.c (print_reg): Fix typo.
4451 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4453 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4455 * configure: Regenerate.
4457 2018-02-05 Martin Sebor <msebor@redhat.com>
4459 PR tree-optimization/83369
4460 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4463 2018-02-05 Martin Liska <mliska@suse.cz>
4465 * doc/invoke.texi: Cherry-pick upstream r323995.
4467 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4469 * ira.c (ira_init_register_move_cost): Adjust comment.
4471 2018-02-05 Martin Liska <mliska@suse.cz>
4473 PR gcov-profile/84137
4474 * doc/gcov.texi: Fix typo in documentation.
4476 2018-02-05 Martin Liska <mliska@suse.cz>
4478 PR gcov-profile/83879
4479 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4480 using dlopen functionality.
4482 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4484 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4485 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4486 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4487 _mm_maskz_range_round_ss): New intrinsics.
4488 (__builtin_ia32_rangesd128_round)
4489 (__builtin_ia32_rangess128_round): Remove.
4490 (__builtin_ia32_rangesd128_mask_round,
4491 __builtin_ia32_rangess128_mask_round): New builtins.
4492 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4493 __builtin_ia32_rangess128_round): Remove.
4494 (__builtin_ia32_rangesd128_mask_round,
4495 __builtin_ia32_rangess128_mask_round): New builtins.
4496 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4497 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4498 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4499 "<round_saeonly_constraint>")): Changed to ...
4500 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4501 "<round_saeonly_scalar_constraint>")): ... this.
4502 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4503 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4504 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4505 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4506 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4508 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4510 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4512 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4513 Remove all values except native, 8540 and 8548.
4515 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4517 * config/i386/i386.c (ix86_output_function_return): Pass
4518 INVALID_REGNUM, instead of -1, as invalid register number to
4519 indirect_thunk_name and output_indirect_thunk.
4521 2018-02-02 Julia Koval <julia.koval@intel.com>
4523 * config.gcc: Add -march=icelake.
4524 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4525 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4526 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4527 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4528 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4529 (processor_target_table): Add icelake.
4530 (ix86_option_override_internal): Handle new PTAs.
4531 (get_builtin_code_for_version): Handle icelake.
4532 (M_INTEL_COREI7_ICELAKE): New.
4533 (fold_builtin_cpu): Handle icelake.
4534 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4535 * doc/invoke.texi: Add -march=icelake.
4537 2018-02-02 Julia Koval <julia.koval@intel.com>
4539 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4540 to wide_int_bitmask.
4541 * wide-int-bitmask.h: New.
4543 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4546 * config/i386/i386.md: Replace Pmode with word_mode in
4547 builtin_setjmp_setup and builtin_longjmp to support x32.
4549 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4553 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4555 (rs6000_supported_cpu_names): New static variable.
4556 (linux_cpu_translation_table): Likewise.
4557 (elf_platform) <cpu>: Define new static variable and use it.
4558 Translate kernel AT_PLATFORM name to canonical name if needed.
4559 Error if platform name is unknown.
4561 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4564 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4566 2018-02-01 Jeff Law <law@redhat.com>
4569 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4570 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4571 the scratch if RELEASE_VIA_POP is false.
4572 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4573 If we have to save a temporary register, decrement SIZE appropriately.
4574 Pass new arguments to release_scratch_register_on_entry.
4575 (ix86_adjust_stack_and_probe): Likewise.
4576 (ix86_emit_probe_stack_range): Pass new arguments to
4577 release_scratch_register_on_entry.
4579 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4581 PR rtl-optimization/84157
4582 * combine.c (change_zero_ext): Use REG_P predicate in
4583 front of HARD_REGISTER_P predicate.
4585 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4587 * config/avr/avr.c (avr_option_override): Move disabling of
4588 -fdelete-null-pointer-checks to...
4589 * common/config/avr/avr-common.c (avr_option_optimization_table):
4592 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4594 PR tree-optimization/81635
4595 * tree-data-ref.c (split_constant_offset_1): For types that
4596 wrap on overflow, try to use range info to prove that wrapping
4599 2018-02-01 Renlin Li <renlin.li@arm.com>
4602 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4604 (aarch64_register_move_cost): Likewise.
4605 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4607 (REG_CLASS_NAMES): Likewise.
4608 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4609 TAILCALL_ADDR_REGS. Remove IP registers.
4610 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4612 2018-02-01 Richard Biener <rguenther@suse.de>
4614 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4615 for specifying RPO order and allow NULL for that.
4616 * domwalk.c (dom_walker::dom_walker): Likewise.
4617 (dom_walker::walk): Handle NULL RPO order.
4618 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4620 (rewrite_update_dom_walker): Likewise.
4621 (mark_def_dom_walker): Likewise.
4623 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4625 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4626 (aarch64_maybe_expand_sve_subreg_move): Declare.
4627 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4628 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4630 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4631 that are semantically a reverse operation.
4632 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4633 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4634 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4636 (aarch64_can_change_mode_class): For big-endian, forbid changes
4637 between two SVE modes if they have different element sizes.
4639 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4641 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4642 the TImode handling for big-endian targets.
4644 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4646 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4647 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4649 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4650 Remove BSWAP handing for big-endian targets and use the form of
4651 LD1RQ appropariate for the mode.
4653 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4655 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4656 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4659 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4662 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4663 check for operands that need to go through aarch64_sve_reload_be.
4665 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4667 PR tree-optimization/81661
4668 PR tree-optimization/84117
4669 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4670 * tree-eh.c: Include gimplify.h.
4671 (find_trapping_overflow, replace_trapping_overflow,
4672 rewrite_to_non_trapping_overflow): New functions.
4673 * tree-vect-loop.c: Include tree-eh.h.
4674 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4675 * tree-data-ref.c: Include tree-eh.h.
4676 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4678 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4680 PR rtl-optimization/84123
4681 * combine.c (change_zero_ext): Check if hard register satisfies
4682 can_change_dest_mode before calling gen_lowpart_SUBREG.
4684 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4687 * ira.c (ira_init_register_move_cost): Remove assert.
4689 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4691 PR rtl-optimization/84071
4692 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4693 * doc/tm.texi: Regenerate.
4695 2018-01-31 Richard Biener <rguenther@suse.de>
4697 PR tree-optimization/84132
4698 * tree-data-ref.c (analyze_miv_subscript): Properly
4699 check whether evolution_function_is_affine_multivariate_p
4700 before calling gcd_of_steps_may_divide_p.
4702 2018-01-31 Julia Koval <julia.koval@intel.com>
4705 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4706 * config/i386/i386.md (rdpid_rex64) New.
4707 (rdpid): Make 32bit only.
4709 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4712 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4713 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4715 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4718 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4720 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4722 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4724 PR rtl-optimization/84071
4725 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4726 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4728 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4730 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4731 (arc_attribute_table): Add 'aux' attribute.
4732 (arc_in_small_data_p): Consider aux like variables.
4733 (arc_is_aux_reg_p): New function.
4734 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4735 (arc_get_aux_arg): New function.
4736 (prepare_move_operands): Handle aux-register access.
4737 (arc_handle_aux_attribute): New function.
4738 * doc/extend.texi (ARC Variable attributes): Add subsection.
4740 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4742 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4743 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4744 (arc_attribute_table): Add 'uncached' attribute.
4745 (arc_print_operand): Print '.di' flag for uncached memory
4747 (arc_in_small_data_p): Do not consider for small data the uncached
4749 (arc_is_uncached_mem_p): New function.
4750 * config/arc/predicates.md (compact_store_memory_operand): Check
4751 for uncached memory accesses.
4752 (nonvol_nonimm_operand): Likewise.
4753 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4755 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4758 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4759 falign-loops=): Add Optimization flag.
4761 2018-01-30 Jeff Law <law@redhat.com>
4764 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4765 INT_REGISTERS_SAVED. Check it prior to calling
4766 get_scratch_register_on_entry.
4767 (ix86_adjust_stack_and_probe): Similarly.
4768 (ix86_emit_probe_stack_range): Similarly.
4769 (ix86_expand_prologue): Corresponding changes.
4771 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4774 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4775 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4777 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4780 * lra-constraints.c (curr_insn_transform): Process AND in the
4783 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4785 PR rtl-optimization/83986
4786 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4787 dependence against last_pending_memory_flush in addition to
4790 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4792 PR tree-optimization/81611
4793 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4796 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4799 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4802 2018-01-30 Richard Biener <rguenther@suse.de>
4803 Jakub Jelinek <jakub@redhat.com>
4805 PR tree-optimization/84111
4806 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4807 inner loops added during recursion, as they don't have up-to-date
4810 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4813 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4814 (can_inline_edge_by_limits_p): ... here.
4815 (can_early_inline_edge_p, check_callers,
4816 update_caller_keys, update_callee_keys, recursive_inlining,
4817 add_new_edges_to_heap, speculation_useful_p,
4818 inline_small_functions,
4819 inline_small_functions, flatten_function,
4820 inline_to_all_callers_1): Update.
4822 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4824 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4827 2018-01-30 Richard Biener <rguenther@suse.de>
4829 PR tree-optimization/83008
4830 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4831 invariant and constant vector uses in stmts when they need
4834 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4837 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4838 * configure: Regenerate.
4840 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4842 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4844 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4845 Use gen_rtx_REG rather than gen_lowpart.
4847 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4849 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4850 rather than 0 when creating partial subregs.
4852 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4854 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4857 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4860 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4861 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4862 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4863 flags. This restores the settings used before the 2017-07-24.
4864 Turning off pre increment/decrement/modify allows IVOPTS to
4865 optimize DF/SF loops where the index is an int.
4867 2018-01-29 Richard Biener <rguenther@suse.de>
4868 Kelvin Nilsen <kelvin@gcc.gnu.org>
4871 * tree-vect-stmts.c (vectorizable_call): Don't call
4872 targetm.vectorize_builtin_md_vectorized_function if callee is
4875 2018-01-22 Carl Love <cel@us.ibm.com>
4877 * doc/extend.tex: Fix typo in second arg in
4878 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4880 2018-01-29 Richard Biener <rguenther@suse.de>
4882 PR tree-optimization/84086
4883 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4884 (flush_ssaname_freelist): When SSA names were released reset
4885 the SCEV hash table.
4887 2018-01-29 Richard Biener <rguenther@suse.de>
4889 PR tree-optimization/84057
4890 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4891 removed paths when removing edges.
4893 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4895 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4896 -mfunction-return=@var{choice}.
4898 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4901 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4903 (layout::print_source_line): Likewise.
4904 (test_get_line_width_without_trailing_whitespace): Add test cases.
4906 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4909 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4912 2018-01-26 Jim Wilson <jimw@sifive.com>
4914 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4916 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4919 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4921 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4922 and CMP + SUB-immediate -> SUBS.
4924 2018-01-26 Martin Sebor <msebor@redhat.com>
4926 PR tree-optimization/83896
4927 * tree-ssa-strlen.c (get_string_len): Rename...
4928 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4929 Avoid assuming length is constant.
4930 (handle_char_store): Use HOST_WIDE_INT for string length.
4932 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4935 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4936 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4938 2018-01-26 Richard Biener <rguenther@suse.de>
4940 PR rtl-optimization/84003
4941 * dse.c (record_store): Only record redundant stores when
4942 the earlier store aliases at least all accesses the later one does.
4944 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4946 PR rtl-optimization/83985
4947 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4948 REG_CFA_RESTORE insns.
4949 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4950 REG_CFA_RESTORE insns here.
4953 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4954 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4956 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4958 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4959 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4960 (arc_init): Likewise.
4961 (arc_override_options): Likewise.
4962 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4964 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4966 * config/arc/arc.h (TARGET_DBNZ): Define.
4967 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4968 properly set the tune attribute.
4969 (dbnz): Use TARGET_DBNZ guard.
4970 * config/arc/arc.opt (mtune): Add core3 option.
4972 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4974 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4975 recognize new pic like addresses.
4976 (arc_delegitimize_address): Clean up.
4978 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4980 * config/arc/arc-arches.def: Option mrf16 valid for all
4982 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4983 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4984 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4985 * config/arc/arc-tables.opt: Regenerate.
4986 * config/arc/arc.c (arc_conditional_register_usage): Handle
4987 reduced register file case.
4988 (arc_file_start): Set must have build attributes.
4989 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4991 * config/arc/arc.opt (mrf16): Add new option.
4992 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4993 * config/arc/genmultilib.awk: Handle new mrf16 option.
4994 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4995 * config/arc/t-multilib: Regenerate.
4996 * doc/invoke.texi (ARC Options): Document mrf16 option.
4998 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5000 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5001 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5002 (arc_attribute_table): Add 'secure_call' attribute.
5003 (arc_print_operand): Print secure call operand.
5004 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5006 (arc_is_secure_call_p): New function. * config/arc/arc.md
5007 (call_i): Add support for sjli instruction.
5008 (call_value_i): Likewise.
5009 * config/arc/constraints.md (Csc): New constraint.
5011 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5012 John Eric Martin <John.Martin@emmicro-us.com>
5014 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5015 * config/arc/arc.c (_arc_jli_section): New struct.
5016 (arc_jli_section): New type.
5017 (rc_jli_sections): New static variable.
5018 (arc_handle_jli_attribute): New function.
5019 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5020 (arc_file_end): New function.
5021 (TARGET_ASM_FILE_END): Define.
5022 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5023 (arc_add_jli_section): New function.
5024 (jli_call_scan): Likewise.
5025 (arc_reorg): Call jli_call_scan.
5026 (arc_output_addsi): Remove 'S' from printing asm operand.
5027 (arc_is_jli_call_p): New function.
5028 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5030 (movhi_insn): Likewise.
5031 (movsi_insn): Likewise.
5032 (movsi_set_cc_insn): Likewise.
5033 (loadqi_update): Likewise.
5034 (load_zeroextendqisi_update): Likewise.
5035 (load_signextendqisi_update): Likewise.
5036 (loadhi_update): Likewise.
5037 (load_zeroextendhisi_update): Likewise.
5038 (load_signextendhisi_update): Likewise.
5039 (loadsi_update): Likewise.
5040 (loadsf_update): Likewise.
5041 (movsicc_insn): Likewise.
5042 (bset_insn): Likewise.
5043 (bxor_insn): Likewise.
5044 (bclr_insn): Likewise.
5045 (bmsk_insn): Likewise.
5046 (bicsi3_insn): Likewise.
5047 (cmpsi_cc_c_insn): Likewise.
5048 (movsi_ne): Likewise.
5049 (movsi_cond_exec): Likewise.
5050 (clrsbsi2): Likewise.
5061 (call_i): Remove 'S' asm letter, add jli instruction.
5062 (call_value_i): Likewise.
5063 * config/arc/arc.op (mjli-always): New option.
5064 * config/arc/constraints.md (Cji): New constraint.
5065 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5067 (subsf3_fpx): Likewise.
5068 (mulsf3_fpx): Likewise.
5069 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5071 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5073 * doc/invoke.texi (ARC): Document mjli-always option.
5075 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5077 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5078 avoid addition with 0 and use incw and decw where possible.
5080 2018-01-26 Richard Biener <rguenther@suse.de>
5082 PR tree-optimization/81082
5083 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5084 association if it requires casting to unsigned.
5085 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5086 from fold_plusminus_mult_expr to catch important cases late when
5087 range info is available.
5089 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5091 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5092 * configure.ac (hidden_linkonce): New test.
5093 * configure: Regenerate.
5094 * config.in: Regenerate.
5096 2018-01-26 Julia Koval <julia.koval@intel.com>
5098 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5099 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5100 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5101 _mm_mask_bitshuffle_epi64_mask): Fix type.
5102 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5103 USI_FTYPE_V4DI_V4DI_USI): Remove.
5104 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5105 __builtin_ia32_vpshufbitqmb256_mask,
5106 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5107 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5108 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5110 2018-01-26 Alan Modra <amodra@gmail.com>
5113 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5114 UNSPEC_VBPERMQ. Sort other unspecs.
5116 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5118 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5120 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5123 * predict.c (drop_profile): Do not push/pop cfun; update also
5125 (handle_missing_profiles): Fix logic looking for zero profiles.
5127 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5130 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5131 on functions with #pragma omp declare simd or functions with simd
5133 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5134 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5135 Remove trailing \n from warning_at calls.
5137 2018-01-25 Tom de Vries <tom@codesourcery.com>
5140 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5141 for neutered workers.
5143 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5146 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5147 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5149 2018-01-24 Jeff Law <law@redhat.com>
5152 * i386.c (get_probe_interval): Move to earlier point.
5153 (ix86_compute_frame_layout): If -fstack-clash-protection and
5154 the frame is larger than the probe interval, then use pushes
5155 to save registers rather than reg->mem moves.
5156 (ix86_expand_prologue): Remove conditional for int_registers_saved
5159 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5162 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5163 min/max for never referenced object.
5165 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5168 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5170 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5171 attributes from DECL_ATTRIBUTES (decl) without affecting
5172 DECL_ATTRIBUTES (current_function_decl).
5173 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5174 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5176 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5178 PR tree-optimization/83979
5179 * fold-const.c (fold_comparison): Use constant_boolean_node
5180 instead of boolean_{true,false}_node.
5182 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5184 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5187 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5189 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5190 Simplify the clause that sets the length attribute.
5191 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5192 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5193 clause that sets the length attribute.
5194 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5196 2018-01-24 Tom de Vries <tom@codesourcery.com>
5199 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5200 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5201 Add strict parameter.
5202 (prevent_branch_around_nothing): Insert dummy insn between branch to
5203 label and label with no ptx insn inbetween.
5204 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5206 2018-01-24 Tom de Vries <tom@codesourcery.com>
5209 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5210 for neutered threads in warp.
5211 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5213 2018-01-24 Richard Biener <rguenther@suse.de>
5215 PR tree-optimization/83176
5216 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5219 2018-01-24 Richard Biener <rguenther@suse.de>
5221 PR tree-optimization/82819
5222 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5223 code generating pluses that are no-ops in the target precision.
5225 2018-01-24 Richard Biener <rguenther@suse.de>
5228 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5230 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5232 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5233 to merge probabilities.
5234 * predict.c (probably_never_executed): Also mark as cold functions
5235 with global 0 profile and guessed local profile.
5236 * profile-count.c (profile_probability::combine_with_count): New
5238 * profile-count.h (profile_probability::operator*,
5239 profile_probability::operator*=, profile_probability::operator/,
5240 profile_probability::operator/=): Reduce precision to adjusted
5241 and set value to guessed on contradictory divisions.
5242 (profile_probability::combine_with_freq): Remove.
5243 (profile_probability::combine_wiht_count): Declare.
5244 (profile_count::force_nonzero):: Set to adjusted.
5245 (profile_count::probability_in):: Set quality to adjusted.
5246 * tree-ssa-tail-merge.c (replace_block_by): Use
5249 2018-01-23 Andrew Waterman <andrew@sifive.com>
5250 Jim Wilson <jimw@sifive.com>
5252 * config/riscv/riscv.c (riscv_stack_boundary): New.
5253 (riscv_option_override): Set riscv_stack_boundary. Handle
5254 riscv_preferred_stack_boundary_arg.
5255 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5256 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5257 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5258 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5259 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5260 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5262 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5265 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5266 of struct ix86_frame.
5267 (ix86_expand_epilogue): Likewise. Add a local variable for
5268 the reg_save_offset field in struct ix86_frame.
5270 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5272 PR tree-optimization/82604
5273 * tree-loop-distribution.c (enum partition_kind): New enum item
5274 PKIND_PARTIAL_MEMSET.
5275 (partition_builtin_p): Support above new enum item.
5276 (generate_code_for_partition): Ditto.
5277 (compute_access_range): Differentiate cases that equality can be
5278 proven at all loops, the innermost loops or no loops.
5279 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5280 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5281 (finalize_partitions, distribute_loop): Don't fuse partition of
5282 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5283 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5286 2018-01-23 Martin Liska <mliska@suse.cz>
5288 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5289 order to ignore the predictor.
5290 (PRED_POLYMORPHIC_CALL): Likewise.
5291 (PRED_RECURSIVE_CALL): Likewise.
5293 2018-01-23 Martin Liska <mliska@suse.cz>
5295 * tree-profile.c (tree_profiling): Print function header to
5296 aware reader which function we are working on.
5297 * value-prof.c (gimple_find_values_to_profile): Do not print
5298 not interesting value histograms.
5300 2018-01-23 Martin Liska <mliska@suse.cz>
5302 * profile-count.h (enum profile_quality): Add
5303 profile_uninitialized as the first value. Do not number values
5304 as they are zero based.
5305 (profile_count::verify): Update sanity check.
5306 (profile_probability::verify): Likewise.
5308 2018-01-23 Nathan Sidwell <nathan@acm.org>
5310 * doc/invoke.texi (ffor-scope): Deprecate.
5312 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5314 PR tree-optimization/83510
5315 * domwalk.c (set_all_edges_as_executable): New function.
5316 (dom_walker::dom_walker): Convert bool param
5317 "skip_unreachable_blocks" to enum reachability. Move setup of
5318 edge flags to set_all_edges_as_executable and only do it when
5319 reachability is REACHABLE_BLOCKS.
5320 * domwalk.h (enum dom_walker::reachability): New enum.
5321 (dom_walker::dom_walker): Convert bool param
5322 "skip_unreachable_blocks" to enum reachability.
5323 (set_all_edges_as_executable): New decl.
5324 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5325 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5327 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5328 but converting true to REACHABLE_BLOCKS.
5329 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5331 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5332 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5333 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5335 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5336 if check_all_array_refs will be called.
5338 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5340 * tree.c (selftest::test_location_wrappers): Add more test
5343 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5345 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5346 (selftest::test_bit_in_range): Likewise.
5348 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5351 * doc/sourcebuild.texi (vect_float): Say that the selector
5352 only describes the situation when -funsafe-math-optimizations is on.
5353 (vect_float_strict): Document.
5355 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5357 PR tree-optimization/83965
5358 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5359 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5360 instead of checking only for a reduction.
5361 (vect_recog_widen_sum_pattern): Likewise.
5363 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5365 * predict.c (probably_never_executed): Only use precise profile info.
5366 (compute_function_frequency): Skip after inlining hack since we now
5367 have quality checking.
5369 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5371 * profile-count.h (profile_probability::very_unlikely,
5372 profile_probability::unlikely, profile_probability::even): Set
5373 precision to guessed.
5375 2018-01-23 Richard Biener <rguenther@suse.de>
5377 PR tree-optimization/83963
5378 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5379 Properly terminate dominator walk when crossing the exit edge not
5380 when visiting its source block.
5382 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5385 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5386 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5388 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5390 PR tree-optimization/83957
5391 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5392 semicolon after for body surrounded by braces.
5394 PR tree-optimization/83081
5395 * profile-count.h (profile_probability::split): New method.
5396 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5397 Use profile_probability::split.
5398 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5399 when splitting a single conditional jump into 2.
5401 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5403 PR tree-optimization/69452
5404 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5407 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5409 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5410 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5411 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5413 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5415 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5416 * config/rl78/rl78.md (movdi): New define_expand.
5417 * config/rl78/rl78.c (rl78_split_movdi): New function.
5419 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5422 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5424 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5425 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5426 128-bit to produce an UNSPEC move to get the double word with the
5427 signbit and then a shift directly to do signbit.
5428 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5429 implementation with a new version that just does either a direct
5430 move or a regular move. Move memory interface to separate insns.
5431 Move insns so they are next to the expander.
5432 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5433 with signbit move. Split big and little endian case.
5434 (signbit<mode>2_dm_mem_le): Likewise.
5435 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5436 (signbit<mode>2_dm2): Likewise.
5438 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5440 * config/rl78/rl78.md (anddi3): New define_expand.
5442 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5444 * config/rl78/rl78.md (umindi3): New define_expand.
5446 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5448 * config/rl78/rl78.md (smindi3): New define_expand.
5450 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5452 * config/rl78/rl78.md (smaxdi3): New define_expand.
5454 2018-01-22 Carl Love <cel@us.ibm.com>
5456 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5457 LVX_V1TI): Add macro expansion.
5458 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5459 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5460 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5461 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5462 Change check to determine if the instruction is a byte reversing
5463 entry. Fix typo in comment.
5464 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5465 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5466 Add def_builtin calls for new builtins.
5467 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5468 Add define_insn expansion.
5470 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5472 * config/rl78/rl78.md (umaxdi3): New define_expand.
5474 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5476 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5477 for non-QImode registers.
5479 2018-01-22 Richard Biener <rguenther@suse.de>
5481 PR tree-optimization/83963
5482 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5483 including the loop exit block.
5484 (scop_detection::merge_sese): Likewise.
5485 (scop_detection::add_scop): Do it here instead.
5487 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5489 * doc/sourcebuild.texi (arm_softfloat): Document.
5491 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5494 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5495 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5496 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5498 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5499 David Edelsohn <dje.gcc@gmail.com>
5502 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5503 Change "crset eq" to "crset 2".
5504 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5505 (*call_indirect_aix<mode>_nospec): Likewise.
5506 (*call_value_indirect_aix<mode>_nospec): Likewise.
5507 (*call_indirect_elfv2<mode>_nospec): Likewise.
5508 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5509 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5510 change assembly output from . to $.
5511 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5512 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5513 (*tablejump<mode>_internal1_nospec): Likewise.
5515 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5518 * config/sh/sh_optimize_sett_clrt.cc:
5519 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5521 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5523 PR tree-optimization/83940
5524 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5525 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5526 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5527 use it to pass back the definition type.
5528 (vect_check_store_rhs): Likewise rhs_dt_out.
5529 (vect_build_gather_load_calls): Add a mask_dt argument and use
5530 it instead of a call to vect_is_simple_use.
5531 (vectorizable_store): Update calls to vect_check_load_store_mask
5532 and vect_check_store_rhs. Use the dt returned by the latter instead
5533 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5534 instead of calls to vect_is_simple_use. Pass the scalar rather
5535 than the vector operand to vect_is_simple_use when handling
5536 second and subsequent copies of an rhs value.
5537 (vectorizable_load): Update calls to vect_check_load_store_mask
5538 and vect_build_gather_load_calls. Use the cached mask_dt and
5539 gs_info.offset_dt instead of calls to vect_is_simple_use.
5541 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5544 * tree-emutls.c: Include gimplify.h.
5545 (lower_emutls_2): New function.
5546 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5547 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5548 it before further processing.
5551 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5552 UINTVAL (trueop1) instead of INTVAL (op1).
5554 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5558 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5559 INCOMING_FRAME_SP_OFFSET if not defined.
5560 (scan_trace): Add ENTRY argument. If true and
5561 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5562 emit a note to adjust the CFA offset.
5563 (create_cfi_notes): Adjust scan_trace callers.
5564 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5565 INCOMING_FRAME_SP_OFFSET in the CIE.
5566 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5567 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5569 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5570 * doc/tm.texi: Regenerated.
5572 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5574 PR rtl-optimization/83147
5575 * lra-constraints.c (remove_inheritance_pseudos): Use
5576 lra_substitute_pseudo_within_insn.
5578 2018-01-19 Tom de Vries <tom@codesourcery.com>
5579 Cesar Philippidis <cesar@codesourcery.com>
5582 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5584 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5587 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5588 spaces for function labels.
5590 2018-01-19 Martin Liska <mliska@suse.cz>
5592 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5593 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5594 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5595 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5596 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5597 (PRED_CONST_RETURN): Change from 69 to 65.
5598 (PRED_NULL_RETURN): Change from 91 to 71.
5599 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5600 (PRED_LOOP_GUARD): Change from 66 to 73.
5602 2018-01-19 Martin Liska <mliska@suse.cz>
5604 * predict.c (predict_insn_def): Add new assert.
5605 (struct branch_predictor): Change type to signed integer.
5606 (test_prediction_value_range): Amend test to cover
5608 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5609 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5610 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5611 (PRED_LOOP_IV_COMPARE): Likewise.
5612 * predict.h (PROB_UNINITIALIZED): Define new constant.
5614 2018-01-19 Martin Liska <mliska@suse.cz>
5616 * predict.c (dump_prediction): Add new format for
5617 analyze_brprob.py script which is enabled with -details
5619 * profile-count.h (precise_p): New function.
5621 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5623 PR tree-optimization/83922
5624 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5625 there are no statements that need masking.
5626 (vect_active_double_reduction_p): New function.
5627 (vect_analyze_loop_operations): Use it when handling phis that
5628 are not in the loop header.
5630 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5632 PR tree-optimization/83914
5633 * tree-vect-loop.c (vectorizable_induction): Don't convert
5634 init_expr or apply the peeling adjustment for inductions
5635 that are nested within the vectorized loop.
5637 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5639 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5642 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5646 * function.h (gimplify_parameters): Add gimple_seq * argument.
5647 * function.c: Include gimple.h and options.h.
5648 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5649 for the added local temporaries if needed.
5650 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5651 if there are any parameter cleanups, wrap whole body into a
5652 try/finally with the cleanups.
5654 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5657 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5658 Use GET_MODE_CLASS for scalar floating point.
5660 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5664 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5665 Fix call of call_cgraph_insertion_hooks.
5667 2018-01-18 Martin Sebor <msebor@redhat.com>
5669 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5671 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5674 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5677 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5680 * common.opt: (-ffile-prefix-map): New option.
5681 * opts.c (common_handle_option): Defer it.
5682 * opts-global.c (handle_common_deferred_options): Handle it.
5683 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5684 * file-prefix-map.h: New file.
5685 (remap_debug_filename, add_debug_prefix_map): ...here.
5686 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5687 * final.c (debug_prefix_map, add_debug_prefix_map
5688 remap_debug_filename): Move to...
5689 * file-prefix-map.c: New file.
5690 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5691 generalize, get rid of alloca(), use strrchr() instead of strchr().
5692 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5693 Implement in terms of add_prefix_map().
5694 (remap_macro_filename, remap_debug_filename): Implement in term of
5696 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5697 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5698 * dbxout.c: Include file-prefix-map.h.
5699 * varasm.c: Likewise.
5700 * vmsdbgout.c: Likewise.
5701 * xcoffout.c: Likewise.
5702 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5703 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5704 * doc/invoke.texi (-ffile-prefix-map): Document.
5705 (-fdebug-prefix-map): Update description.
5707 2018-01-18 Martin Liska <mliska@suse.cz>
5709 * config/i386/i386.c (indirect_thunk_name): Document that also
5711 (output_indirect_thunk): Document why both instructions
5712 (pause and lfence) are generated.
5714 2018-01-18 Richard Biener <rguenther@suse.de>
5716 PR tree-optimization/83887
5717 * graphite-scop-detection.c
5718 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5719 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5720 (scop_detection::merge_sese): Re-implement with a flood-fill
5721 algorithm that properly finds a SESE region if it exists.
5723 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5726 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5727 pointer_diff optimizations use view_convert instead of convert.
5729 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5731 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5732 Generate different code for -mno-speculate-indirect-jumps.
5733 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5734 (*call_indirect_aix<mode>): Disable for
5735 -mno-speculate-indirect-jumps.
5736 (*call_indirect_aix<mode>_nospec): New define_insn.
5737 (*call_value_indirect_aix<mode>): Disable for
5738 -mno-speculate-indirect-jumps.
5739 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5740 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5741 -mno-speculate-indirect-jumps.
5742 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5744 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5746 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5747 long double type, set the flags for noting the default long double
5748 type, even if we don't pass or return a long double type.
5750 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5753 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5756 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5758 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5759 support for merge[hl].
5760 (fold_mergehl_helper): New helper function.
5761 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5762 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5763 (altivec_vmrglw_direct): Add xxmrglw insn.
5765 2018-01-17 Andrew Waterman <andrew@sifive.com>
5767 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5768 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5770 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5773 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5774 call the lto_location_cache before reading the
5775 DECL_SOURCE_LOCATION of the types.
5777 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5778 Richard Sandiford <richard.sandiford@linaro.org>
5780 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5781 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5782 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5783 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5784 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5786 * config/aarch64/constraints.md (aarch64_movti_operand):
5788 * config/aarch64/predicates.md (Uti): Add new constraint.
5790 2018-01-17 Carl Love <cel@us.ibm.com>
5792 * config/rs6000/vsx.md (define_expand xl_len_r,
5793 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5794 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5796 (define_expand, define_insn): Move the shift left from the
5797 define_insn to the define_expand for lxvl and stxvl instructions.
5798 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5799 and XL_LEN_R definitions to PURE.
5801 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5803 * config/i386/i386.c (indirect_thunk_name): Declare regno
5804 as unsigned int. Compare regno with INVALID_REGNUM.
5805 (output_indirect_thunk): Ditto.
5806 (output_indirect_thunk_function): Ditto.
5807 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5808 in the call to output_indirect_thunk_function.
5810 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5813 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5814 rather than the size of inner_type to determine the stack slot size
5815 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5817 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5820 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5823 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5825 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5826 endian Linux systems to optionally enable multilibs for selecting
5827 the long double type if the user configured an explicit type.
5828 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5829 have no long double multilibs if not defined.
5830 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5831 warn if the user used -mabi={ieee,ibm}longdouble and we built
5832 multilibs for long double.
5833 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5834 appropriate multilib option.
5835 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5837 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5838 for building long double multilibs.
5839 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5841 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5843 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5846 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5848 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5851 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5854 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5857 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5859 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5860 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5862 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5864 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5865 different rtl trees depending on TARGET_64BIT.
5866 (rs6000_gen_lvx): Likewise.
5868 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5870 * config/visium/visium.md (nop): Tweak comment.
5871 (hazard_nop): Likewise.
5873 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5875 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5876 -mspeculate-indirect-jumps.
5877 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5878 for -mno-speculate-indirect-jumps.
5879 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5880 (*call_value_indirect_elfv2<mode>): Disable for
5881 -mno-speculate-indirect-jumps.
5882 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5883 (indirect_jump): Emit different RTL for
5884 -mno-speculate-indirect-jumps.
5885 (*indirect_jump<mode>): Disable for
5886 -mno-speculate-indirect-jumps.
5887 (*indirect_jump<mode>_nospec): New define_insn.
5888 (tablejump): Emit different RTL for
5889 -mno-speculate-indirect-jumps.
5890 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5891 (tablejumpsi_nospec): New define_expand.
5892 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5893 (tablejumpdi_nospec): New define_expand.
5894 (*tablejump<mode>_internal1): Disable for
5895 -mno-speculate-indirect-jumps.
5896 (*tablejump<mode>_internal1_nospec): New define_insn.
5897 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5900 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5902 * caller-save.c (insert_save): Drop unnecessary parameter. All
5905 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5906 Richard Biener <rguenth@suse.de>
5909 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5910 return early, inline manually is_gimple_sizepos. Make sure if we
5911 call gimplify_expr we don't end up with a gimple constant.
5912 * tree.c (variably_modified_type_p): Don't return true for
5913 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5914 * gimplify.h (is_gimple_sizepos): Remove.
5916 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5918 PR tree-optimization/83857
5919 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5920 vectorizable_live_operation for pure SLP statements.
5921 (vectorizable_live_operation): Handle PHIs.
5923 2018-01-16 Richard Biener <rguenther@suse.de>
5925 PR tree-optimization/83867
5926 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5927 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5929 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5932 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5933 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5934 If off is not INTEGER_CST, issue a may not be aligned warning
5935 rather than isn't aligned. Use isn%'t rather than isn't.
5936 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5938 <case MULT_EXPR>: Improve the case when bottom and one of the
5939 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5940 operand, in that case check if the other operand is multiple of
5941 bottom divided by the INTEGER_CST operand.
5943 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5946 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5947 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5948 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5949 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5950 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5951 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5952 (pa_function_arg_size): New function.
5954 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5956 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5957 in a separate statement.
5959 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5961 PR tree-optimization/83847
5962 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5963 group gathers and scatters.
5965 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5967 PR rtl-optimization/86620
5968 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5970 PR rtl-optimization/83213
5971 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5972 to last if both are JUMP_INSNs.
5974 PR tree-optimization/83843
5975 * gimple-ssa-store-merging.c
5976 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5977 store_immediate_info for bswap/nop orig_stores.
5979 2018-01-15 Andrew Waterman <andrew@sifive.com>
5981 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5983 <UDIV>: Increase cost if !TARGET_DIV.
5985 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5987 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5988 (define_attr "cr_logical_3op"): New.
5989 (cceq_ior_compare): Adjust.
5990 (cceq_ior_compare_complement): Adjust.
5991 (*cceq_rev_compare): Adjust.
5992 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5993 (is_cracked_insn): Adjust.
5994 (insn_must_be_first_in_group): Adjust.
5995 * config/rs6000/40x.md: Adjust.
5996 * config/rs6000/440.md: Adjust.
5997 * config/rs6000/476.md: Adjust.
5998 * config/rs6000/601.md: Adjust.
5999 * config/rs6000/603.md: Adjust.
6000 * config/rs6000/6xx.md: Adjust.
6001 * config/rs6000/7450.md: Adjust.
6002 * config/rs6000/7xx.md: Adjust.
6003 * config/rs6000/8540.md: Adjust.
6004 * config/rs6000/cell.md: Adjust.
6005 * config/rs6000/e300c2c3.md: Adjust.
6006 * config/rs6000/e500mc.md: Adjust.
6007 * config/rs6000/e500mc64.md: Adjust.
6008 * config/rs6000/e5500.md: Adjust.
6009 * config/rs6000/e6500.md: Adjust.
6010 * config/rs6000/mpc.md: Adjust.
6011 * config/rs6000/power4.md: Adjust.
6012 * config/rs6000/power5.md: Adjust.
6013 * config/rs6000/power6.md: Adjust.
6014 * config/rs6000/power7.md: Adjust.
6015 * config/rs6000/power8.md: Adjust.
6016 * config/rs6000/power9.md: Adjust.
6017 * config/rs6000/rs64.md: Adjust.
6018 * config/rs6000/titan.md: Adjust.
6020 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6022 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6023 ix86_indirect_branch_register logic.
6025 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6027 * config/i386/constraints.md (Bs): Update
6028 ix86_indirect_branch_register check. Don't check
6029 ix86_indirect_branch_register with GOT_memory_operand.
6031 * config/i386/predicates.md (GOT_memory_operand): Don't check
6032 ix86_indirect_branch_register here.
6033 (GOT32_symbol_operand): Likewise.
6035 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6037 * config/i386/predicates.md (constant_call_address_operand):
6038 Rewrite ix86_indirect_branch_register logic.
6039 (sibcall_insn_operand): Likewise.
6041 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6043 * config/i386/constraints.md (Bs): Replace
6044 ix86_indirect_branch_thunk_register with
6045 ix86_indirect_branch_register.
6047 * config/i386/i386.md (indirect_jump): Likewise.
6048 (tablejump): Likewise.
6049 (*sibcall_memory): Likewise.
6050 (*sibcall_value_memory): Likewise.
6051 Peepholes of indirect call and jump via memory: Likewise.
6052 * config/i386/i386.opt: Likewise.
6053 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6054 (GOT_memory_operand): Likewise.
6055 (call_insn_operand): Likewise.
6056 (sibcall_insn_operand): Likewise.
6057 (GOT32_symbol_operand): Likewise.
6059 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6062 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6063 type rather than type addr's type points to.
6064 (expand_omp_atomic_mutex): Likewise.
6065 (expand_omp_atomic): Likewise.
6067 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6070 * config/i386/i386.c (output_indirect_thunk_function): Use
6071 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6072 for __x86_return_thunk.
6074 2018-01-15 Richard Biener <rguenther@suse.de>
6077 * expmed.c (extract_bit_field_1): Fix typo.
6079 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6082 * config/arm/iterators.md (VF): New mode iterator.
6083 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6084 Remove integer-related logic from pattern.
6085 (neon_vabd<mode>_3): Likewise.
6087 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6090 * common.opt (fstrict-overflow): No longer an alias.
6091 (fwrapv-pointer): New option.
6092 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6093 also for pointer types based on flag_wrapv_pointer.
6094 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6095 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6096 opts->x_flag_wrapv got set.
6097 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6098 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6099 POINTER_TYPE_OVERFLOW_UNDEFINED.
6100 * match.pd: Likewise in address comparison pattern.
6101 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6103 2018-01-15 Richard Biener <rguenther@suse.de>
6106 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6107 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6108 Reset type names to their identifier if their TYPE_DECL doesn't
6109 have linkage (and thus is used for ODR and devirt).
6110 (save_debug_info_for_decl): Remove.
6111 (save_debug_info_for_type): Likewise.
6112 (add_tree_to_fld_list): Adjust.
6113 * tree-pretty-print.c (dump_generic_node): Make dumping of
6114 type names more robust.
6116 2018-01-15 Richard Biener <rguenther@suse.de>
6118 * BASE-VER: Bump to 8.0.1.
6120 2018-01-14 Martin Sebor <msebor@redhat.com>
6123 * builtins.c (check_access): Avoid warning when the no-warning bit
6126 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6128 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6129 * ira-color (allocno_hard_regs_compare): Likewise.
6131 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6134 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6135 Use .pushsection/.popsection.
6137 2018-01-14 Martin Sebor <msebor@redhat.com>
6140 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6142 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6144 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6145 entry from extra_headers.
6146 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6147 extra_headers, make the list bitwise identical to the i?86-*-* one.
6149 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6151 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6152 -mcmodel=large with -mindirect-branch=thunk,
6153 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6154 -mfunction-return=thunk-extern.
6155 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6156 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6157 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6159 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6161 * config/i386/i386.c (print_reg): Print the name of the full
6162 integer register without '%'.
6163 (ix86_print_operand): Handle 'V'.
6164 * doc/extend.texi: Document 'V' modifier.
6166 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6168 * config/i386/constraints.md (Bs): Disallow memory operand for
6169 -mindirect-branch-register.
6171 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6172 (GOT_memory_operand): Likewise.
6173 (call_insn_operand): Likewise.
6174 (sibcall_insn_operand): Likewise.
6175 (GOT32_symbol_operand): Likewise.
6176 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6177 for -mindirect-branch-register.
6178 (tablejump): Likewise.
6179 (*sibcall_memory): Likewise.
6180 (*sibcall_value_memory): Likewise.
6181 Disallow peepholes of indirect call and jump via memory for
6182 -mindirect-branch-register.
6183 (*call_pop): Replace m with Bw.
6184 (*call_value_pop): Likewise.
6185 (*sibcall_pop_memory): Replace m with Bs.
6186 * config/i386/i386.opt (mindirect-branch-register): New option.
6187 * doc/invoke.texi: Document -mindirect-branch-register option.
6189 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6191 * config/i386/i386-protos.h (ix86_output_function_return): New.
6192 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6193 set function_return_type.
6194 (indirect_thunk_name): Add ret_p to indicate thunk for function
6196 (output_indirect_thunk_function): Pass false to
6197 indirect_thunk_name.
6198 (ix86_output_indirect_branch_via_reg): Likewise.
6199 (ix86_output_indirect_branch_via_push): Likewise.
6200 (output_indirect_thunk_function): Create alias for function
6201 return thunk if regno < 0.
6202 (ix86_output_function_return): New function.
6203 (ix86_handle_fndecl_attribute): Handle function_return.
6204 (ix86_attribute_table): Add function_return.
6205 * config/i386/i386.h (machine_function): Add
6206 function_return_type.
6207 * config/i386/i386.md (simple_return_internal): Use
6208 ix86_output_function_return.
6209 (simple_return_internal_long): Likewise.
6210 * config/i386/i386.opt (mfunction-return=): New option.
6211 (indirect_branch): Mention -mfunction-return=.
6212 * doc/extend.texi: Document function_return function attribute.
6213 * doc/invoke.texi: Document -mfunction-return= option.
6215 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6217 * config/i386/i386-opts.h (indirect_branch): New.
6218 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6219 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6220 with local indirect jump when converting indirect call and jump.
6221 (ix86_set_indirect_branch_type): New.
6222 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6223 (indirectlabelno): New.
6224 (indirect_thunk_needed): Likewise.
6225 (indirect_thunk_bnd_needed): Likewise.
6226 (indirect_thunks_used): Likewise.
6227 (indirect_thunks_bnd_used): Likewise.
6228 (INDIRECT_LABEL): Likewise.
6229 (indirect_thunk_name): Likewise.
6230 (output_indirect_thunk): Likewise.
6231 (output_indirect_thunk_function): Likewise.
6232 (ix86_output_indirect_branch_via_reg): Likewise.
6233 (ix86_output_indirect_branch_via_push): Likewise.
6234 (ix86_output_indirect_branch): Likewise.
6235 (ix86_output_indirect_jmp): Likewise.
6236 (ix86_code_end): Call output_indirect_thunk_function if needed.
6237 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6239 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6240 (ix86_attribute_table): Add indirect_branch.
6241 * config/i386/i386.h (machine_function): Add indirect_branch_type
6242 and has_local_indirect_jump.
6243 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6245 (tablejump): Likewise.
6246 (*indirect_jump): Use ix86_output_indirect_jmp.
6247 (*tablejump_1): Likewise.
6248 (simple_return_indirect_internal): Likewise.
6249 * config/i386/i386.opt (mindirect-branch=): New option.
6250 (indirect_branch): New.
6253 (thunk-inline): Likewise.
6254 (thunk-extern): Likewise.
6255 * doc/extend.texi: Document indirect_branch function attribute.
6256 * doc/invoke.texi: Document -mindirect-branch= option.
6258 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6261 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6263 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6265 * ipa-inline.c (want_inline_small_function_p): Return false if
6266 inlining has already failed with CIF_FINAL_ERROR.
6267 (update_caller_keys): Call want_inline_small_function_p before
6269 (update_callee_keys): Likewise.
6271 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6273 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6275 (rs6000_quadword_masked_address_p): Likewise.
6276 (quad_aligned_load_p): Likewise.
6277 (quad_aligned_store_p): Likewise.
6278 (const_load_sequence_p): Add comment to describe the outer-most loop.
6279 (mimic_memory_attributes_and_flags): New function.
6280 (rs6000_gen_stvx): Likewise.
6281 (replace_swapped_aligned_store): Likewise.
6282 (rs6000_gen_lvx): Likewise.
6283 (replace_swapped_aligned_load): Likewise.
6284 (replace_swapped_load_constant): Capitalize argument name in
6285 comment describing this function.
6286 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6287 and stores that access quad-word aligned addresses and replace
6288 with stvx or lvx instructions when appropriate.
6289 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6290 New function prototype.
6291 (rs6000_quadword_masked_address_p): Likewise.
6292 (rs6000_gen_lvx): Likewise.
6293 (rs6000_gen_stvx): Likewise.
6294 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6295 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6296 when memory address is aligned.
6297 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6298 this split to select lvx instruction when memory address is aligned.
6299 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6300 instruction when memory address is aligned.
6301 (*vsx_le_perm_load_v16qi): Likewise.
6302 (four unnamed splitters): Modify to select the stvx instruction
6303 when memory is aligned.
6305 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6307 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6308 which appears in the queue multiple times.
6310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6311 Alan Hayward <alan.hayward@arm.com>
6312 David Sherwood <david.sherwood@arm.com>
6314 * tree-vectorizer.h (vec_lower_bound): New structure.
6315 (_loop_vec_info): Add check_nonzero and lower_bounds.
6316 (LOOP_VINFO_CHECK_NONZERO): New macro.
6317 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6318 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6319 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6320 fields. Make seg_len the distance travelled, not including the
6322 (dr_direction_indicator): Declare.
6323 (dr_zero_step_indicator): Likewise.
6324 (dr_known_forward_stride_p): Likewise.
6325 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6327 (runtime_alias_check_p): Allow runtime alias checks with
6329 (operator ==): Compare access_size and align.
6330 (prune_runtime_alias_test_list): Rework for new distinction between
6331 the access_size and seg_len.
6332 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6334 (get_segment_min_max): New function.
6335 (create_intersect_range_checks): Use it.
6336 (dr_step_indicator): New function.
6337 (dr_direction_indicator): Likewise.
6338 (dr_zero_step_indicator): Likewise.
6339 (dr_known_forward_stride_p): Likewise.
6340 * tree-loop-distribution.c (data_ref_segment_size): Return
6341 DR_STEP * (niters - 1).
6342 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6344 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6345 (vect_preserves_scalar_order_p): New function, split out from...
6346 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6347 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6348 (vect_vfa_access_size): New function.
6349 (vect_vfa_align): Likewise.
6350 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6351 (dump_lower_bound): New function.
6352 (vect_check_lower_bound): Likewise.
6353 (vect_small_gap_p): Likewise.
6354 (vectorizable_with_step_bound_p): Likewise.
6355 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6356 depencies if the vectorization factor is 1. Convert the checks
6357 for nonzero steps into checks on the bounds of DR_STEP. Try using
6358 a bunds check for variable steps if the minimum required step is
6359 relatively small. Update calls to the dr_with_seg_len
6360 constructor and to vect_compile_time_alias.
6361 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6363 (vect_loop_versioning): Call it.
6364 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6366 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6368 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6369 Alan Hayward <alan.hayward@arm.com>
6370 David Sherwood <david.sherwood@arm.com>
6372 * doc/sourcebuild.texi (vect_scatter_store): Document.
6373 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6375 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6377 * genopinit.c (main): Add supports_vec_scatter_store and
6378 supports_vec_scatter_store_cached to target_optabs.
6379 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6380 IFN_MASK_SCATTER_STORE.
6381 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6383 * internal-fn.h (internal_store_fn_p): Declare.
6384 (internal_fn_stored_value_index): Likewise.
6385 * internal-fn.c (scatter_store_direct): New macro.
6386 (expand_scatter_store_optab_fn): New function.
6387 (direct_scatter_store_optab_supported_p): New macro.
6388 (internal_store_fn_p): New function.
6389 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6390 IFN_MASK_SCATTER_STORE.
6391 (internal_fn_mask_index): Likewise.
6392 (internal_fn_stored_value_index): New function.
6393 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6395 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6396 * optabs-query.c (supports_vec_scatter_store_p): New function.
6397 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6398 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6399 true for scatter stores.
6400 (vect_gather_scatter_fn_p): Handle scatter stores too.
6401 (vect_check_gather_scatter): Consider using scatter stores if
6402 supports_vec_scatter_store_p.
6403 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6405 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6406 internal_fn_stored_value_index.
6407 (check_load_store_masking): Handle scatter stores too.
6408 (vect_get_store_rhs): Make public.
6409 (vectorizable_call): Use internal_store_fn_p.
6410 (vectorizable_store): Handle scatter store internal functions.
6411 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6412 when deciding whether the end of the group has been reached.
6413 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6414 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6415 (mask_scatter_store<mode>): New insns.
6417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6418 Alan Hayward <alan.hayward@arm.com>
6419 David Sherwood <david.sherwood@arm.com>
6421 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6422 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6423 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6425 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6426 Use vect_truncate_gather_scatter_offset if we can't treat the
6427 operation as a normal gather load or scatter store.
6428 (get_group_load_store_type): Take the gather_scatter_info
6429 as argument. Try using a gather load or scatter store for
6430 single-element groups.
6431 (get_load_store_type): Update calls to get_group_load_store_type
6432 and vect_use_strided_gather_scatters_p.
6434 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6435 Alan Hayward <alan.hayward@arm.com>
6436 David Sherwood <david.sherwood@arm.com>
6438 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6439 optional tree argument.
6440 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6442 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6443 but continue to use the current value as a fallback.
6444 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6445 to compare the updates.
6446 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6447 (get_load_store_type): Use it when handling a strided access.
6448 (vect_get_strided_load_store_ops): New function.
6449 (vect_get_data_ptr_increment): Likewise.
6450 (vectorizable_load): Handle strided gather loads. Always pass
6451 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6453 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6454 Alan Hayward <alan.hayward@arm.com>
6455 David Sherwood <david.sherwood@arm.com>
6457 * doc/md.texi (gather_load@var{m}): Document.
6458 (mask_gather_load@var{m}): Likewise.
6459 * genopinit.c (main): Add supports_vec_gather_load and
6460 supports_vec_gather_load_cached to target_optabs.
6461 * optabs-tree.c (init_tree_optimization_optabs): Use
6462 ggc_cleared_alloc to allocate target_optabs.
6463 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6464 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6466 * internal-fn.h (internal_load_fn_p): Declare.
6467 (internal_gather_scatter_fn_p): Likewise.
6468 (internal_fn_mask_index): Likewise.
6469 (internal_gather_scatter_fn_supported_p): Likewise.
6470 * internal-fn.c (gather_load_direct): New macro.
6471 (expand_gather_load_optab_fn): New function.
6472 (direct_gather_load_optab_supported_p): New macro.
6473 (direct_internal_fn_optab): New function.
6474 (internal_load_fn_p): Likewise.
6475 (internal_gather_scatter_fn_p): Likewise.
6476 (internal_fn_mask_index): Likewise.
6477 (internal_gather_scatter_fn_supported_p): Likewise.
6478 * optabs-query.c (supports_at_least_one_mode_p): New function.
6479 (supports_vec_gather_load_p): Likewise.
6480 * optabs-query.h (supports_vec_gather_load_p): Declare.
6481 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6482 and memory_type field.
6483 (NUM_PATTERNS): Bump to 15.
6484 * tree-vect-data-refs.c: Include internal-fn.h.
6485 (vect_gather_scatter_fn_p): New function.
6486 (vect_describe_gather_scatter_call): Likewise.
6487 (vect_check_gather_scatter): Try using internal functions for
6488 gather loads. Recognize existing calls to a gather load function.
6489 (vect_analyze_data_refs): Consider using gather loads if
6490 supports_vec_gather_load_p.
6491 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6492 (vect_get_gather_scatter_offset_type): Likewise.
6493 (vect_convert_mask_for_vectype): Likewise.
6494 (vect_add_conversion_to_patterm): Likewise.
6495 (vect_try_gather_scatter_pattern): Likewise.
6496 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6497 (vect_vect_recog_func_ptrs): Add it.
6498 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6499 internal_fn_mask_index and internal_gather_scatter_fn_p.
6500 (check_load_store_masking): Take the gather_scatter_info as an
6501 argument and handle gather loads.
6502 (vect_get_gather_scatter_ops): New function.
6503 (vectorizable_call): Check internal_load_fn_p.
6504 (vectorizable_load): Likewise. Handle gather load internal
6506 (vectorizable_store): Update call to check_load_store_masking.
6507 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6508 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6509 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6510 (aarch64_gather_scale_operand_d): New predicates.
6511 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6512 (mask_gather_load<mode>): New insns.
6514 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6515 Alan Hayward <alan.hayward@arm.com>
6516 David Sherwood <david.sherwood@arm.com>
6518 * optabs.def (fold_left_plus_optab): New optab.
6519 * doc/md.texi (fold_left_plus_@var{m}): Document.
6520 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6521 * internal-fn.c (fold_left_direct): Define.
6522 (expand_fold_left_optab_fn): Likewise.
6523 (direct_fold_left_optab_supported_p): Likewise.
6524 * fold-const-call.c (fold_const_fold_left): New function.
6525 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6526 * tree-parloops.c (valid_reduction_p): New function.
6527 (gather_scalar_reductions): Use it.
6528 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6529 (vect_finish_replace_stmt): Declare.
6530 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6531 (needs_fold_left_reduction_p): New function, split out from...
6532 (vect_is_simple_reduction): ...here. Accept reductions that
6533 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6534 (vect_force_simple_reduction): Also store the reduction type in
6535 the assignment's STMT_VINFO_REDUC_TYPE.
6536 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6537 (merge_with_identity): New function.
6538 (vect_expand_fold_left): Likewise.
6539 (vectorize_fold_left_reduction): Likewise.
6540 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6541 scalar phi in place for it. Check for target support and reject
6542 cases that would reassociate the operation. Defer the transform
6543 phase to vectorize_fold_left_reduction.
6544 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6545 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6546 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6550 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6551 call to ifc_temp_var.
6553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6554 Alan Hayward <alan.hayward@arm.com>
6555 David Sherwood <david.sherwood@arm.com>
6557 * target.def (legitimize_address_displacement): Take the original
6558 offset as a poly_int.
6559 * targhooks.h (default_legitimize_address_displacement): Update
6561 * targhooks.c (default_legitimize_address_displacement): Likewise.
6562 * doc/tm.texi: Regenerate.
6563 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6564 as an argument, moving assert of ad->disp == ad->disp_term to...
6565 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6566 Try calling targetm.legitimize_address_displacement before expanding
6567 the address rather than afterwards, and adjust for the new interface.
6568 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6569 Match the new hook interface. Handle SVE addresses.
6570 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6575 * Makefile.in (OBJS): Add early-remat.o.
6576 * target.def (select_early_remat_modes): New hook.
6577 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6578 * doc/tm.texi: Regenerate.
6579 * targhooks.h (default_select_early_remat_modes): Declare.
6580 * targhooks.c (default_select_early_remat_modes): New function.
6581 * timevar.def (TV_EARLY_REMAT): New timevar.
6582 * passes.def (pass_early_remat): New pass.
6583 * tree-pass.h (make_pass_early_remat): Declare.
6584 * early-remat.c: New file.
6585 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6587 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6589 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6590 Alan Hayward <alan.hayward@arm.com>
6591 David Sherwood <david.sherwood@arm.com>
6593 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6594 vfm1 with a bound_epilog parameter.
6595 (vect_do_peeling): Update calls accordingly, and move the prologue
6596 call earlier in the function. Treat the base bound_epilog as 0 for
6597 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6598 this base when peeling for gaps.
6599 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6600 with fully-masked loops.
6601 (vect_estimate_min_profitable_iters): Handle the single peeled
6602 iteration in that case.
6604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6605 Alan Hayward <alan.hayward@arm.com>
6606 David Sherwood <david.sherwood@arm.com>
6608 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6609 single-element interleaving even if the size is not a power of 2.
6610 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6611 accesses for single-element interleaving if the group size is
6614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6615 Alan Hayward <alan.hayward@arm.com>
6616 David Sherwood <david.sherwood@arm.com>
6618 * doc/md.texi (fold_extract_last_@var{m}): Document.
6619 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6620 * optabs.def (fold_extract_last_optab): New optab.
6621 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6622 * internal-fn.c (fold_extract_direct): New macro.
6623 (expand_fold_extract_optab_fn): Likewise.
6624 (direct_fold_extract_optab_supported_p): Likewise.
6625 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6626 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6627 EXTRACT_LAST_REDUCTION.
6628 (get_initial_def_for_reduction): Do not create an initial vector
6629 for EXTRACT_LAST_REDUCTION reductions.
6630 (vectorizable_reduction): Leave the scalar phi in place for
6631 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6632 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6633 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6634 transform phase to vectorizable_condition.
6635 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6637 (vect_finish_stmt_generation): ...here.
6638 (vect_finish_replace_stmt): New function.
6639 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6640 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6642 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6645 Alan Hayward <alan.hayward@arm.com>
6646 David Sherwood <david.sherwood@arm.com>
6648 * doc/md.texi (extract_last_@var{m}): Document.
6649 * optabs.def (extract_last_optab): New optab.
6650 * internal-fn.def (EXTRACT_LAST): New internal function.
6651 * internal-fn.c (cond_unary_direct): New macro.
6652 (expand_cond_unary_optab_fn): Likewise.
6653 (direct_cond_unary_optab_supported_p): Likewise.
6654 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6655 loops using EXTRACT_LAST.
6656 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6657 (extract_last_<mode>): ...this optab.
6658 (vec_extract<mode><Vel>): Update accordingly.
6660 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6661 Alan Hayward <alan.hayward@arm.com>
6662 David Sherwood <david.sherwood@arm.com>
6664 * target.def (empty_mask_is_expensive): New hook.
6665 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6666 * doc/tm.texi: Regenerate.
6667 * targhooks.h (default_empty_mask_is_expensive): Declare.
6668 * targhooks.c (default_empty_mask_is_expensive): New function.
6669 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6670 if the target says that empty masks are expensive.
6671 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6673 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6675 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6676 Alan Hayward <alan.hayward@arm.com>
6677 David Sherwood <david.sherwood@arm.com>
6679 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6680 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6681 (vect_use_loop_mask_for_alignment_p): New function.
6682 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6683 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6684 niters_skip argument. Make sure that the first niters_skip elements
6685 of the first iteration are inactive.
6686 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6687 Update call to vect_set_loop_masks_directly.
6688 (get_misalign_in_elems): New function, split out from...
6689 (vect_gen_prolog_loop_niters): ...here.
6690 (vect_update_init_of_dr): Take a code argument that specifies whether
6691 the adjustment should be added or subtracted.
6692 (vect_update_init_of_drs): Likewise.
6693 (vect_prepare_for_masked_peels): New function.
6694 (vect_do_peeling): Skip prologue peeling if we're using a mask
6695 instead. Update call to vect_update_inits_of_drs.
6696 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6698 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6699 alignment. Do not include the number of peeled iterations in
6700 the minimum threshold in that case.
6701 (vectorizable_induction): Adjust the start value down by
6702 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6703 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6704 Take the number of skipped iterations into account when calculating
6706 * tree-vect-stmts.c (vect_gen_while_not): New function.
6708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6709 Alan Hayward <alan.hayward@arm.com>
6710 David Sherwood <david.sherwood@arm.com>
6712 * doc/sourcebuild.texi (vect_fully_masked): Document.
6713 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6715 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6717 (vect_analyze_loop_2): ...here. Don't check the vectorization
6718 factor against the number of loop iterations if the loop is
6721 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6722 Alan Hayward <alan.hayward@arm.com>
6723 David Sherwood <david.sherwood@arm.com>
6725 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6726 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6727 (dump_groups): Update accordingly.
6728 (iv_use::mem_type): New member variable.
6729 (address_p): New function.
6730 (record_use): Add a mem_type argument and initialize the new
6732 (record_group_use): Add a mem_type argument. Use address_p.
6733 Remove obsolete null checks of base_object. Update call to record_use.
6734 (find_interesting_uses_op): Update call to record_group_use.
6735 (find_interesting_uses_cond): Likewise.
6736 (find_interesting_uses_address): Likewise.
6737 (get_mem_type_for_internal_fn): New function.
6738 (find_address_like_use): Likewise.
6739 (find_interesting_uses_stmt): Try find_address_like_use before
6740 calling find_interesting_uses_op.
6741 (addr_offset_valid_p): Use the iv mem_type field as the type
6742 of the addressed memory.
6743 (add_autoinc_candidates): Likewise.
6744 (get_address_cost): Likewise.
6745 (split_small_address_groups_p): Use address_p.
6746 (split_address_groups): Likewise.
6747 (add_iv_candidate_for_use): Likewise.
6748 (autoinc_possible_for_pair): Likewise.
6749 (rewrite_groups): Likewise.
6750 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6751 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6752 (get_alias_ptr_type_for_ptr_address): New function.
6753 (rewrite_use_address): Rewrite address uses in calls that were
6754 identified by find_address_like_use.
6756 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6757 Alan Hayward <alan.hayward@arm.com>
6758 David Sherwood <david.sherwood@arm.com>
6760 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6762 * gimple-expr.h (is_gimple_addressable: Likewise.
6763 * gimple-expr.c (is_gimple_address): Likewise.
6764 * internal-fn.c (expand_call_mem_ref): New function.
6765 (expand_mask_load_optab_fn): Use it.
6766 (expand_mask_store_optab_fn): Likewise.
6768 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6769 Alan Hayward <alan.hayward@arm.com>
6770 David Sherwood <david.sherwood@arm.com>
6772 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6773 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6774 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6775 (cond_umax@var{mode}): Document.
6776 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6777 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6778 (cond_umin_optab, cond_umax_optab): New optabs.
6779 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6780 (COND_IOR, COND_XOR): New internal functions.
6781 * internal-fn.h (get_conditional_internal_fn): Declare.
6782 * internal-fn.c (cond_binary_direct): New macro.
6783 (expand_cond_binary_optab_fn): Likewise.
6784 (direct_cond_binary_optab_supported_p): Likewise.
6785 (get_conditional_internal_fn): New function.
6786 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6787 Cope with reduction statements that are vectorized as calls rather
6789 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6790 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6791 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6792 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6793 (UNSPEC_COND_EOR): New unspecs.
6794 (optab): Add mappings for them.
6795 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6796 (sve_int_op, sve_fp_op): New int attributes.
6798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6799 Alan Hayward <alan.hayward@arm.com>
6800 David Sherwood <david.sherwood@arm.com>
6802 * optabs.def (while_ult_optab): New optab.
6803 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6804 * internal-fn.def (WHILE_ULT): New internal function.
6805 * internal-fn.h (direct_internal_fn_supported_p): New override
6806 that takes two types as argument.
6807 * internal-fn.c (while_direct): New macro.
6808 (expand_while_optab_fn): New function.
6809 (convert_optab_supported_p): Likewise.
6810 (direct_while_optab_supported_p): New macro.
6811 * wide-int.h (wi::udiv_ceil): New function.
6812 * tree-vectorizer.h (rgroup_masks): New structure.
6813 (vec_loop_masks): New typedef.
6814 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6816 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6817 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6818 (vect_max_vf): New function.
6819 (slpeel_make_loop_iterate_ntimes): Delete.
6820 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6821 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6822 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6823 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6824 internal-fn.h, stor-layout.h and optabs-query.h.
6825 (vect_set_loop_mask): New function.
6826 (add_preheader_seq): Likewise.
6827 (add_header_seq): Likewise.
6828 (interleave_supported_p): Likewise.
6829 (vect_maybe_permute_loop_masks): Likewise.
6830 (vect_set_loop_masks_directly): Likewise.
6831 (vect_set_loop_condition_masked): Likewise.
6832 (vect_set_loop_condition_unmasked): New function, split out from
6833 slpeel_make_loop_iterate_ntimes.
6834 (slpeel_make_loop_iterate_ntimes): Rename to..
6835 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6836 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6837 (vect_do_peeling): Update call accordingly.
6838 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6840 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6841 mask_compare_type, can_fully_mask_p and fully_masked_p.
6842 (release_vec_loop_masks): New function.
6843 (_loop_vec_info): Use it to free the loop masks.
6844 (can_produce_all_loop_masks_p): New function.
6845 (vect_get_max_nscalars_per_iter): Likewise.
6846 (vect_verify_full_masking): Likewise.
6847 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6848 retries, and free the mask rgroups before retrying. Check loop-wide
6849 reasons for disallowing fully-masked loops. Make the final decision
6850 about whether use a fully-masked loop or not.
6851 (vect_estimate_min_profitable_iters): Do not assume that peeling
6852 for the number of iterations will be needed for fully-masked loops.
6853 (vectorizable_reduction): Disable fully-masked loops.
6854 (vectorizable_live_operation): Likewise.
6855 (vect_halve_mask_nunits): New function.
6856 (vect_double_mask_nunits): Likewise.
6857 (vect_record_loop_mask): Likewise.
6858 (vect_get_loop_mask): Likewise.
6859 (vect_transform_loop): Handle the case in which the final loop
6860 iteration might handle a partial vector. Call vect_set_loop_condition
6861 instead of slpeel_make_loop_iterate_ntimes.
6862 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6863 (check_load_store_masking): New function.
6864 (prepare_load_store_mask): Likewise.
6865 (vectorizable_store): Handle fully-masked loops.
6866 (vectorizable_load): Likewise.
6867 (supportable_widening_operation): Use vect_halve_mask_nunits for
6869 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6870 (vect_gen_while): New function.
6871 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6872 (aarch64_uqdec<mode>): New insn.
6874 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6875 Alan Hayward <alan.hayward@arm.com>
6876 David Sherwood <david.sherwood@arm.com>
6878 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6879 (reduc_xor_scal_optab): New optabs.
6880 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6881 (reduc_xor_scal_@var{m}): Document.
6882 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6883 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6885 * fold-const-call.c (fold_const_call): Handle them.
6886 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6887 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6888 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6889 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6890 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6891 (UNSPEC_XORV): New unspecs.
6892 (optab): Add entries for them.
6893 (BITWISEV): New int iterator.
6894 (bit_reduc_op): New int attributes.
6896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6897 Alan Hayward <alan.hayward@arm.com>
6898 David Sherwood <david.sherwood@arm.com>
6900 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6901 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6902 * optabs.def (vec_shl_insert_optab): New optab.
6903 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6904 (duplicate_and_interleave): Likewise.
6905 * tree-vect-loop.c: Include internal-fn.h.
6906 (neutral_op_for_slp_reduction): New function, split out from
6907 get_initial_defs_for_reduction.
6908 (get_initial_def_for_reduction): Handle option 2 for variable-length
6909 vectors by loading the neutral value into a vector and then shifting
6910 the initial value into element 0.
6911 (get_initial_defs_for_reduction): Replace the code argument with
6912 the neutral value calculated by neutral_op_for_slp_reduction.
6913 Use gimple_build_vector for constant-length vectors.
6914 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6915 but the first group_size elements have a neutral value.
6916 Use duplicate_and_interleave otherwise.
6917 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6918 Update call to get_initial_defs_for_reduction. Handle SLP
6919 reductions for variable-length vectors by creating one vector
6920 result for each scalar result, with the elements associated
6921 with other scalar results stubbed out with the neutral value.
6922 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6923 Require IFN_VEC_SHL_INSERT for double reductions on
6924 variable-length vectors, or SLP reductions that have
6925 a neutral value. Require can_duplicate_and_interleave_p
6926 support for variable-length unchained SLP reductions if there
6927 is no neutral value, such as for MIN/MAX reductions. Also require
6928 the number of vector elements to be a multiple of the number of
6929 SLP statements when doing variable-length unchained SLP reductions.
6930 Update call to vect_create_epilog_for_reduction.
6931 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6932 and remove initial values.
6933 (duplicate_and_interleave): Make public.
6934 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6935 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6938 Alan Hayward <alan.hayward@arm.com>
6939 David Sherwood <david.sherwood@arm.com>
6941 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6942 (can_duplicate_and_interleave_p): New function.
6943 (vect_get_and_check_slp_defs): Take the vector of statements
6944 rather than just the current one. Remove excess parentheses.
6945 Restriction rejectinon of vect_constant_def and vect_external_def
6946 for variable-length vectors to boolean types, or types for which
6947 can_duplicate_and_interleave_p is false.
6948 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6949 (duplicate_and_interleave): New function.
6950 (vect_get_constant_vectors): Use gimple_build_vector for
6951 constant-length vectors and suitable variable-length constant
6952 vectors. Use duplicate_and_interleave for other variable-length
6953 vectors. Don't defer the update when inserting new statements.
6955 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6956 Alan Hayward <alan.hayward@arm.com>
6957 David Sherwood <david.sherwood@arm.com>
6959 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6960 min_profitable_iters doesn't go negative.
6962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6963 Alan Hayward <alan.hayward@arm.com>
6964 David Sherwood <david.sherwood@arm.com>
6966 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6967 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6968 * optabs.def (vec_mask_load_lanes_optab): New optab.
6969 (vec_mask_store_lanes_optab): Likewise.
6970 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6971 (MASK_STORE_LANES): Likewise.
6972 * internal-fn.c (mask_load_lanes_direct): New macro.
6973 (mask_store_lanes_direct): Likewise.
6974 (expand_mask_load_optab_fn): Handle masked operations.
6975 (expand_mask_load_lanes_optab_fn): New macro.
6976 (expand_mask_store_optab_fn): Handle masked operations.
6977 (expand_mask_store_lanes_optab_fn): New macro.
6978 (direct_mask_load_lanes_optab_supported_p): Likewise.
6979 (direct_mask_store_lanes_optab_supported_p): Likewise.
6980 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6982 (vect_load_lanes_supported): Likewise.
6983 * tree-vect-data-refs.c (strip_conversion): New function.
6984 (can_group_stmts_p): Likewise.
6985 (vect_analyze_data_ref_accesses): Use it instead of checking
6986 for a pair of assignments.
6987 (vect_store_lanes_supported): Take a masked_p parameter.
6988 (vect_load_lanes_supported): Likewise.
6989 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6990 vect_store_lanes_supported and vect_load_lanes_supported.
6991 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6992 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6993 parameter. Don't allow gaps for masked accesses.
6994 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6995 and vect_load_lanes_supported.
6996 (get_load_store_type): Take a masked_p parameter and update
6997 call to get_group_load_store_type.
6998 (vectorizable_store): Update call to get_load_store_type.
6999 Handle IFN_MASK_STORE_LANES.
7000 (vectorizable_load): Update call to get_load_store_type.
7001 Handle IFN_MASK_LOAD_LANES.
7003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7004 Alan Hayward <alan.hayward@arm.com>
7005 David Sherwood <david.sherwood@arm.com>
7007 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7009 * config/aarch64/aarch64-protos.h
7010 (aarch64_sve_struct_memory_operand_p): Declare.
7011 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7012 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7013 (VPRED, vpred): Handle SVE structure modes.
7014 * config/aarch64/constraints.md (Utx): New constraint.
7015 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7016 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7017 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7018 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7019 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7020 structure modes. Split into pieces after RA.
7021 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7022 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7024 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7025 SVE structure modes.
7026 (aarch64_classify_address): Likewise.
7027 (sizetochar): Move earlier in file.
7028 (aarch64_print_operand): Handle SVE register lists.
7029 (aarch64_array_mode): New function.
7030 (aarch64_sve_struct_memory_operand_p): Likewise.
7031 (TARGET_ARRAY_MODE): Redefine.
7033 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7034 Alan Hayward <alan.hayward@arm.com>
7035 David Sherwood <david.sherwood@arm.com>
7037 * target.def (array_mode): New target hook.
7038 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7039 * doc/tm.texi: Regenerate.
7040 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7041 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7042 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7044 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7048 Alan Hayward <alan.hayward@arm.com>
7049 David Sherwood <david.sherwood@arm.com>
7051 * fold-const.c (fold_binary_loc): Check the argument types
7052 rather than the result type when testing for a vector operation.
7054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7056 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7057 * doc/tm.texi: Regenerate.
7059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7060 Alan Hayward <alan.hayward@arm.com>
7061 David Sherwood <david.sherwood@arm.com>
7063 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7064 (sve): Document new AArch64 extension.
7065 * doc/md.texi (w): Extend the description of the AArch64
7066 constraint to include SVE vectors.
7067 (Upl, Upa): Document new AArch64 predicate constraints.
7068 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7070 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7071 (msve-vector-bits=): New option.
7072 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7073 SVE when these are disabled.
7074 (sve): New extension.
7075 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7076 modes. Adjust their number of units based on aarch64_sve_vg.
7077 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7078 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7079 aarch64_addr_query_type.
7080 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7081 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7082 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7083 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7084 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7085 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7086 (aarch64_simd_imm_zero_p): Delete.
7087 (aarch64_check_zero_based_sve_index_immediate): Declare.
7088 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7089 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7090 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7091 (aarch64_sve_float_mul_immediate_p): Likewise.
7092 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7094 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7095 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7096 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7097 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7098 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7099 (aarch64_regmode_natural_size): Likewise.
7100 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7101 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7103 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7104 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7105 for VG and the SVE predicate registers.
7106 (V_ALIASES): Add a "z"-prefixed alias.
7107 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7108 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7109 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7110 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7111 (REG_CLASS_NAMES): Add entries for them.
7112 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7113 and the predicate registers.
7114 (aarch64_sve_vg): Declare.
7115 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7116 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7117 (REGMODE_NATURAL_SIZE): Define.
7118 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7120 * config/aarch64/aarch64.c: Include cfgrtl.h.
7121 (simd_immediate_info): Add a constructor for series vectors,
7122 and an associated step field.
7123 (aarch64_sve_vg): New variable.
7124 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7125 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7126 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7127 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7128 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7129 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7130 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7131 (aarch64_get_mask_mode): New functions.
7132 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7133 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7134 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7135 predicate modes and predicate registers. Explicitly restrict
7136 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7137 to store a vector mode if it is recognized by
7138 aarch64_classify_vector_mode.
7139 (aarch64_regmode_natural_size): New function.
7140 (aarch64_hard_regno_caller_save_mode): Return the original mode
7142 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7143 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7144 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7145 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7147 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7148 does not overlap dest if the function is frame-related. Handle
7150 (aarch64_split_add_offset): New function.
7151 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7152 them aarch64_add_offset.
7153 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7154 and update call to aarch64_sub_sp.
7155 (aarch64_add_cfa_expression): New function.
7156 (aarch64_expand_prologue): Pass extra temporary registers to the
7157 functions above. Handle the case in which we need to emit new
7158 DW_CFA_expressions for registers that were originally saved
7159 relative to the stack pointer, but now have to be expressed
7160 relative to the frame pointer.
7161 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7163 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7164 IP0 and IP1 values for SVE frames.
7165 (aarch64_expand_vec_series): New function.
7166 (aarch64_expand_sve_widened_duplicate): Likewise.
7167 (aarch64_expand_sve_const_vector): Likewise.
7168 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7169 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7170 into the register, rather than emitting a SET directly.
7171 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7172 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7173 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7174 (offset_9bit_signed_scaled_p): New functions.
7175 (aarch64_replicate_bitmask_imm): New function.
7176 (aarch64_bitmask_imm): Use it.
7177 (aarch64_cannot_force_const_mem): Reject expressions involving
7178 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7179 (aarch64_classify_index): Handle SVE indices, by requiring
7180 a plain register index with a scale that matches the element size.
7181 (aarch64_classify_address): Handle SVE addresses. Assert that
7182 the mode of the address is VOIDmode or an integer mode.
7183 Update call to aarch64_classify_symbol.
7184 (aarch64_classify_symbolic_expression): Update call to
7185 aarch64_classify_symbol.
7186 (aarch64_const_vec_all_in_range_p): New function.
7187 (aarch64_print_vector_float_operand): Likewise.
7188 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7189 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7190 and the FP immediates 1.0 and 0.5.
7191 (aarch64_print_address_internal): Handle SVE addresses.
7192 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7193 (aarch64_regno_regclass): Handle predicate registers.
7194 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7196 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7197 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7198 (aarch64_convert_sve_vector_bits): New function.
7199 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7200 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7202 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7203 Handle SVE vector and predicate modes. Accept VL-based constants
7204 that need only one temporary register, and VL offsets that require
7205 no temporary registers.
7206 (aarch64_conditional_register_usage): Mark the predicate registers
7207 as fixed if SVE isn't available.
7208 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7209 Return true for SVE vector and predicate modes.
7210 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7211 rather than an unsigned int. Handle SVE modes.
7212 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7214 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7216 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7217 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7218 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7219 (aarch64_sve_float_mul_immediate_p): New functions.
7220 (aarch64_sve_valid_immediate): New function.
7221 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7222 Explicitly reject structure modes. Check for INDEX constants.
7223 Handle PTRUE and PFALSE constants.
7224 (aarch64_check_zero_based_sve_index_immediate): New function.
7225 (aarch64_simd_imm_zero_p): Delete.
7226 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7227 vector modes. Accept constants in the range of CNT[BHWD].
7228 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7229 ask for an Advanced SIMD mode.
7230 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7231 (aarch64_simd_vector_alignment): Handle SVE predicates.
7232 (aarch64_vectorize_preferred_vector_alignment): New function.
7233 (aarch64_simd_vector_alignment_reachable): Use it instead of
7235 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7236 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7238 (MAX_VECT_LEN): Delete.
7239 (expand_vec_perm_d): Add a vec_flags field.
7240 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7241 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7242 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7244 (aarch64_evpc_rev): Rename to...
7245 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7246 (aarch64_evpc_rev_global): New function.
7247 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7248 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7250 (aarch64_evpc_sve_tbl): New function.
7251 (aarch64_expand_vec_perm_const_1): Update after rename of
7252 aarch64_evpc_rev. Handle SVE permutes too, trying
7253 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7254 than aarch64_evpc_tbl.
7255 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7256 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7257 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7258 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7259 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7260 (aarch64_expand_sve_vcond): New functions.
7261 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7262 of aarch64_vector_mode_p.
7263 (aarch64_dwarf_poly_indeterminate_value): New function.
7264 (aarch64_compute_pressure_classes): Likewise.
7265 (aarch64_can_change_mode_class): Likewise.
7266 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7267 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7268 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7269 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7270 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7271 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7272 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7273 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7275 (Dn, Dl, Dr): Accept const as well as const_vector.
7276 (Dz): Likewise. Compare against CONST0_RTX.
7277 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7278 of "vector" where appropriate.
7279 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7280 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7281 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7282 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7283 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7284 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7285 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7286 (v_int_equiv): Extend to SVE modes.
7287 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7289 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7290 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7291 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7292 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7293 (SVE_COND_FP_CMP): New int iterators.
7294 (perm_hilo): Handle the new unpack unspecs.
7295 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7297 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7298 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7299 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7300 (aarch64_equality_operator, aarch64_constant_vector_operand)
7301 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7302 (aarch64_sve_nonimmediate_operand): Likewise.
7303 (aarch64_sve_general_operand): Likewise.
7304 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7305 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7306 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7307 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7308 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7309 (aarch64_sve_float_arith_immediate): Likewise.
7310 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7311 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7312 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7313 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7314 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7315 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7316 (aarch64_sve_float_arith_operand): Likewise.
7317 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7318 (aarch64_sve_float_mul_operand): Likewise.
7319 (aarch64_sve_vec_perm_operand): Likewise.
7320 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7321 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7322 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7323 as well as const_vector.
7324 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7325 in file. Use CONST0_RTX and CONSTM1_RTX.
7326 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7327 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7328 Use aarch64_simd_imm_zero.
7329 * config/aarch64/aarch64-sve.md: New file.
7330 * config/aarch64/aarch64.md: Include it.
7331 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7332 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7333 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7334 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7335 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7336 (sve): New attribute.
7337 (enabled): Disable instructions with the sve attribute unless
7339 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7340 aarch64_expand_mov_immediate.
7341 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7342 CNT[BHSD] immediates.
7343 (movti): Split CONST_POLY_INT moves into two halves.
7344 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7345 Split additions that need a temporary here if the destination
7346 is the stack pointer.
7347 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7348 (*add<mode>3_poly_1): New instruction.
7349 (set_clobber_cc): New expander.
7351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7353 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7354 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7355 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7356 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7357 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7358 Change innermode from fixed_mode_size to machine_mode.
7359 (simplify_subreg): Update call accordingly. Handle a constant-sized
7360 subreg of a variable-length CONST_VECTOR.
7362 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7363 Alan Hayward <alan.hayward@arm.com>
7364 David Sherwood <david.sherwood@arm.com>
7366 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7367 (add_offset_to_base): New function, split out from...
7368 (create_mem_ref): ...here. When handling a scale other than 1,
7369 check first whether the address is valid without the offset.
7370 Add it into the base if so, leaving the index and scale as-is.
7372 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7375 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7376 fold_for_warn before checking if arg2 is INTEGER_CST.
7378 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7380 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7381 (store_multiple_operation): Delete.
7382 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7383 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7384 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7385 guarded by TARGET_STRING.
7386 (rs6000_output_load_multiple): Delete.
7387 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7388 OPTION_MASK_STRING / TARGET_STRING handling.
7389 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7390 (const rs6000_opt_masks) <"string">: Change mask to 0.
7391 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7392 (MASK_STRING): Delete.
7393 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7395 (load_multiple): Delete.
7402 (store_multiple): Delete.
7409 (movmemsi_8reg): Delete.
7410 (corresponding unnamed define_insn): Delete.
7411 (movmemsi_6reg): Delete.
7412 (corresponding unnamed define_insn): Delete.
7413 (movmemsi_4reg): Delete.
7414 (corresponding unnamed define_insn): Delete.
7415 (movmemsi_2reg): Delete.
7416 (corresponding unnamed define_insn): Delete.
7417 (movmemsi_1reg): Delete.
7418 (corresponding unnamed define_insn): Delete.
7419 * config/rs6000/rs6000.opt (mno-string): New.
7420 (mstring): Replace by deprecation warning stub.
7421 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7423 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7425 * regrename.c (regrename_do_replace): If replacing the same
7426 reg multiple times, try to reuse last created gen_raw_REG.
7429 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7430 main to workaround a bug in GDB.
7432 2018-01-12 Tom de Vries <tom@codesourcery.com>
7435 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7437 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7439 PR rtl-optimization/80481
7440 * ira-color.c (get_cap_member): New function.
7441 (allocnos_conflict_by_live_ranges_p): Use it.
7442 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7443 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7445 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7448 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7449 (*saddl_se_1): Ditto.
7451 (*ssubl_se_1): Ditto.
7453 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7455 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7456 rather than wi::to_widest for DR_INITs.
7457 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7458 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7459 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7461 (vect_analyze_group_access_1): Note that here.
7463 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7465 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7466 polynomial type sizes.
7468 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7470 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7471 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7472 (gimple_add_tmp_var): Likewise.
7474 2018-01-12 Martin Liska <mliska@suse.cz>
7476 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7477 (gimple_alloc_sizes): Likewise.
7478 (dump_gimple_statistics): Use PRIu64 in printf format.
7479 * gimple.h: Change uint64_t to int.
7481 2018-01-12 Martin Liska <mliska@suse.cz>
7483 * tree-core.h: Use uint64_t instead of int.
7484 * tree.c (tree_node_counts): Likewise.
7485 (tree_node_sizes): Likewise.
7486 (dump_tree_statistics): Use PRIu64 in printf format.
7488 2018-01-12 Martin Liska <mliska@suse.cz>
7490 * Makefile.in: As qsort_chk is implemented in vec.c, add
7491 vec.o to linkage of gencfn-macros.
7492 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7493 passing the info to record_node_allocation_statistics.
7494 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7496 * ggc-common.c (struct ggc_usage): Add operator== and use
7497 it in operator< and compare function.
7498 * mem-stats.h (struct mem_usage): Likewise.
7499 * vec.c (struct vec_usage): Remove operator< and compare
7500 function. Can be simply inherited.
7502 2018-01-12 Martin Jambor <mjambor@suse.cz>
7505 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7506 * tree-ssa-math-opts.c: Include domwalk.h.
7507 (convert_mult_to_fma_1): New function.
7508 (fma_transformation_info): New type.
7509 (fma_deferring_state): Likewise.
7510 (cancel_fma_deferring): New function.
7511 (result_of_phi): Likewise.
7512 (last_fma_candidate_feeds_initial_phi): Likewise.
7513 (convert_mult_to_fma): Added deferring logic, split actual
7514 transformation to convert_mult_to_fma_1.
7515 (math_opts_dom_walker): New type.
7516 (math_opts_dom_walker::after_dom_children): New method, body moved
7517 here from pass_optimize_widening_mul::execute, added deferring logic
7519 (pass_optimize_widening_mul::execute): Moved most of code to
7520 math_opts_dom_walker::after_dom_children.
7521 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7522 * config/i386/i386.c (ix86_option_override_internal): Added
7523 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7525 2018-01-12 Richard Biener <rguenther@suse.de>
7528 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7529 inline instance vars.
7531 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7534 * config/rx/rx.c (rx_is_restricted_memory_address):
7537 2018-01-12 Richard Biener <rguenther@suse.de>
7539 PR tree-optimization/80846
7540 * target.def (split_reduction): New target hook.
7541 * targhooks.c (default_split_reduction): New function.
7542 * targhooks.h (default_split_reduction): Declare.
7543 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7544 target requests first reduce vectors by combining low and high
7546 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7547 (get_vectype_for_scalar_type_and_size): Export.
7548 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7549 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7550 * doc/tm.texi: Regenerate.
7551 * config/i386/i386.c (ix86_split_reduction): Implement
7552 TARGET_VECTORIZE_SPLIT_REDUCTION.
7554 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7557 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7558 in PIC mode except for TARGET_VXWORKS_RTP.
7559 * config/sparc/sparc.c: Include cfgrtl.h.
7560 (TARGET_INIT_PIC_REG): Define.
7561 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7562 (sparc_pic_register_p): New predicate.
7563 (sparc_legitimate_address_p): Use it.
7564 (sparc_legitimize_pic_address): Likewise.
7565 (sparc_delegitimize_address): Likewise.
7566 (sparc_mode_dependent_address_p): Likewise.
7567 (gen_load_pcrel_sym): Remove 4th parameter.
7568 (load_got_register): Adjust call to above. Remove obsolete stuff.
7569 (sparc_expand_prologue): Do not call load_got_register here.
7570 (sparc_flat_expand_prologue): Likewise.
7571 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7572 (sparc_use_pseudo_pic_reg): New function.
7573 (sparc_init_pic_reg): Likewise.
7574 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7575 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7577 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7579 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7580 Add item for branch_cost.
7582 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7584 PR rtl-optimization/83565
7585 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7586 not extend the result to a larger mode for rotate operations.
7587 (num_sign_bit_copies1): Likewise.
7589 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7592 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7594 Use values-Xc.o for -pedantic.
7595 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7597 2018-01-12 Martin Liska <mliska@suse.cz>
7600 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7602 (possible_polymorphic_call_targets): Use it.
7603 (ipa_devirt): Likewise.
7605 2018-01-12 Martin Liska <mliska@suse.cz>
7607 * profile-count.h (enum profile_quality): Use 0 as invalid
7608 enum value of profile_quality.
7610 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7612 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7613 -mext-string options.
7615 2018-01-12 Richard Biener <rguenther@suse.de>
7617 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7618 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7619 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7621 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7623 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7625 * configure.ac (--with-long-double-format): Add support for the
7626 configuration option to change the default long double format on
7628 * config.gcc (powerpc*-linux*-*): Likewise.
7629 * configure: Regenerate.
7630 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7631 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7632 used without modification.
7634 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7636 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7637 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7638 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7639 MISC_BUILTIN_SPEC_BARRIER.
7640 (rs6000_init_builtins): Likewise.
7641 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7643 (speculation_barrier): New define_insn.
7644 * doc/extend.texi: Document __builtin_speculation_barrier.
7646 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7649 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7650 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7651 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7653 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7654 integral modes instead of "ss" and "sd".
7655 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7656 vectors with 32-bit and 64-bit elements.
7657 (vecdupssescalarmodesuffix): New mode attribute.
7658 (vec_dup<mode>): Use it.
7660 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7663 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7664 frame if argument is passed on stack.
7666 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7669 * ree.c (combine_reaching_defs): Optimize also
7670 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7671 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7673 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7676 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7678 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7681 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7682 after they are computed.
7684 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7686 PR tree-optimization/83695
7687 * gimple-loop-linterchange.cc
7688 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7689 reset cached scev information after interchange.
7690 (pass_linterchange::execute): Remove call to scev_reset_htab.
7692 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7694 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7695 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7696 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7697 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7698 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7699 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7700 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7701 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7702 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7703 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7704 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7705 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7706 (V_lane_reg): Likewise.
7707 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7709 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7710 (vfmal_lane_low<mode>_intrinsic,
7711 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7712 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7713 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7714 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7715 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7716 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7718 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7720 * config/arm/arm-cpus.in (fp16fml): New feature.
7721 (ALL_SIMD): Add fp16fml.
7722 (armv8.2-a): Add fp16fml as an option.
7723 (armv8.3-a): Likewise.
7724 (armv8.4-a): Add fp16fml as part of fp16.
7725 * config/arm/arm.h (TARGET_FP16FML): Define.
7726 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7728 * config/arm/arm-modes.def (V2HF): Define.
7729 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7730 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7731 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7732 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7733 vfmsl_low, vfmsl_high): New set of builtins.
7734 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7735 (vfml_op): New code attribute.
7736 (VFMLHALVES): New int iterator.
7737 (VFML, VFMLSEL): New mode attributes.
7738 (V_reg): Define mapping for V2HF.
7739 (V_hi, V_lo): New mode attributes.
7740 (VF_constraint): Likewise.
7741 (vfml_half, vfml_half_selector): New int attributes.
7742 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7744 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7745 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7747 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7748 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7749 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7750 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7752 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7753 Document new effective target and option set.
7755 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7757 * config/arm/arm-cpus.in (armv8_4): New feature.
7758 (ARMv8_4a): New fgroup.
7759 (armv8.4-a): New arch.
7760 * config/arm/arm-tables.opt: Regenerate.
7761 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7762 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7763 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7764 Add matching rules for -march=armv8.4-a and extensions.
7765 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7767 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7770 * config/rx/rx.md (BW): New mode attribute.
7771 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7773 2018-01-11 Richard Biener <rguenther@suse.de>
7775 PR tree-optimization/83435
7776 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7777 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7778 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7780 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7781 Alan Hayward <alan.hayward@arm.com>
7782 David Sherwood <david.sherwood@arm.com>
7784 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7786 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7787 (aarch64_print_address_internal): Use it to check for a zero offset.
7789 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7790 Alan Hayward <alan.hayward@arm.com>
7791 David Sherwood <david.sherwood@arm.com>
7793 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7794 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7795 Return a poly_int64 rather than a HOST_WIDE_INT.
7796 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7797 rather than a HOST_WIDE_INT.
7798 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7799 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7800 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7801 final_offset from HOST_WIDE_INT to poly_int64.
7802 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7803 to_constant when getting the number of units in an Advanced SIMD
7805 (aarch64_builtin_vectorized_function): Check for a constant number
7807 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7809 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7810 attribute instead of GET_MODE_NUNITS.
7811 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7812 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7813 GET_MODE_SIZE for fixed-size registers.
7814 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7815 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7816 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7817 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7818 (aarch64_print_operand, aarch64_print_address_internal)
7819 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7820 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7821 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7822 Handle polynomial GET_MODE_SIZE.
7823 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7824 wider than SImode without modification.
7825 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7826 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7827 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7828 passing and returning SVE modes.
7829 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7830 rather than GEN_INT.
7831 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7832 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7833 (aarch64_allocate_and_probe_stack_space): Likewise.
7834 (aarch64_layout_frame): Cope with polynomial offsets.
7835 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7836 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7838 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7839 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7840 poly_int64 rather than a HOST_WIDE_INT.
7841 (aarch64_get_separate_components, aarch64_process_components)
7842 (aarch64_expand_prologue, aarch64_expand_epilogue)
7843 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7844 (aarch64_anchor_offset): New function, split out from...
7845 (aarch64_legitimize_address): ...here.
7846 (aarch64_builtin_vectorization_cost): Handle polynomial
7847 TYPE_VECTOR_SUBPARTS.
7848 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7850 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7851 number of elements from the PARALLEL rather than the mode.
7852 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7853 rather than GET_MODE_BITSIZE.
7854 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7855 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7856 (aarch64_expand_vec_perm_const_1): Handle polynomial
7857 d->perm.length () and d->perm elements.
7858 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7859 Apply to_constant to d->perm elements.
7860 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7861 polynomial CONST_VECTOR_NUNITS.
7862 (aarch64_move_pointer): Take amount as a poly_int64 rather
7864 (aarch64_progress_pointer): Avoid temporary variable.
7865 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7866 the mode attribute instead of GET_MODE.
7868 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7869 Alan Hayward <alan.hayward@arm.com>
7870 David Sherwood <david.sherwood@arm.com>
7872 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7873 x exists before using it.
7874 (aarch64_add_constant_internal): Rename to...
7875 (aarch64_add_offset_1): ...this. Replace regnum with separate
7876 src and dest rtxes. Handle the case in which they're different,
7877 including when the offset is zero. Replace scratchreg with an rtx.
7878 Use 2 additions if there is no spare register into which we can
7879 move a 16-bit constant.
7880 (aarch64_add_constant): Delete.
7881 (aarch64_add_offset): Replace reg with separate src and dest
7882 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7883 Use aarch64_add_offset_1.
7884 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7885 an rtx rather than an int. Take the delta as a poly_int64
7886 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7887 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7888 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7889 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7890 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7892 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7893 aarch64_add_constant.
7895 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7897 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7898 Use scalar_float_mode.
7900 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7902 * config/aarch64/aarch64-simd.md
7903 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7904 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7905 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7906 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7907 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7908 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7909 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7910 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7911 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7912 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7914 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7917 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7918 targ_options->x_arm_arch_string is non NULL.
7920 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7922 * config/aarch64/aarch64.h
7923 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7925 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7928 * expmed.c (emit_store_flag_force): Swap if const op0
7929 and change VOIDmode to mode of op0.
7931 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7933 PR rtl-optimization/83761
7934 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7935 than bytes to mode_for_size.
7937 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7940 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7941 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7944 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7947 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7948 when in layout mode.
7949 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7950 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7953 2018-01-10 Michael Collison <michael.collison@arm.com>
7955 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7956 * config/aarch64/aarch64-option-extension.def: Add
7957 AARCH64_OPT_EXTENSION of 'fp16fml'.
7958 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7959 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7960 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7961 * config/aarch64/constraints.md (Ui7): New constraint.
7962 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7963 (VFMLA_SEL_W): Ditto.
7966 (VFMLA16_LOW): New int iterator.
7967 (VFMLA16_HIGH): Ditto.
7968 (UNSPEC_FMLAL): New unspec.
7969 (UNSPEC_FMLSL): Ditto.
7970 (UNSPEC_FMLAL2): Ditto.
7971 (UNSPEC_FMLSL2): Ditto.
7972 (f16mac): New code attribute.
7973 * config/aarch64/aarch64-simd-builtins.def
7974 (aarch64_fmlal_lowv2sf): Ditto.
7975 (aarch64_fmlsl_lowv2sf): Ditto.
7976 (aarch64_fmlalq_lowv4sf): Ditto.
7977 (aarch64_fmlslq_lowv4sf): Ditto.
7978 (aarch64_fmlal_highv2sf): Ditto.
7979 (aarch64_fmlsl_highv2sf): Ditto.
7980 (aarch64_fmlalq_highv4sf): Ditto.
7981 (aarch64_fmlslq_highv4sf): Ditto.
7982 (aarch64_fmlal_lane_lowv2sf): Ditto.
7983 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7984 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7985 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7986 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7987 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7988 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7989 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7990 (aarch64_fmlal_lane_highv2sf): Ditto.
7991 (aarch64_fmlsl_lane_highv2sf): Ditto.
7992 (aarch64_fmlal_laneq_highv2sf): Ditto.
7993 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7994 (aarch64_fmlalq_lane_highv4sf): Ditto.
7995 (aarch64_fmlsl_lane_highv4sf): Ditto.
7996 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7997 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7998 * config/aarch64/aarch64-simd.md:
7999 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8000 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8001 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8002 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8003 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8004 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8005 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8006 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8007 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8008 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8009 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8010 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8011 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8012 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8013 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8014 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8015 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8016 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8017 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8018 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8019 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8020 (vfmlsl_low_u32): Ditto.
8021 (vfmlalq_low_u32): Ditto.
8022 (vfmlslq_low_u32): Ditto.
8023 (vfmlal_high_u32): Ditto.
8024 (vfmlsl_high_u32): Ditto.
8025 (vfmlalq_high_u32): Ditto.
8026 (vfmlslq_high_u32): Ditto.
8027 (vfmlal_lane_low_u32): Ditto.
8028 (vfmlsl_lane_low_u32): Ditto.
8029 (vfmlal_laneq_low_u32): Ditto.
8030 (vfmlsl_laneq_low_u32): Ditto.
8031 (vfmlalq_lane_low_u32): Ditto.
8032 (vfmlslq_lane_low_u32): Ditto.
8033 (vfmlalq_laneq_low_u32): Ditto.
8034 (vfmlslq_laneq_low_u32): Ditto.
8035 (vfmlal_lane_high_u32): Ditto.
8036 (vfmlsl_lane_high_u32): Ditto.
8037 (vfmlal_laneq_high_u32): Ditto.
8038 (vfmlsl_laneq_high_u32): Ditto.
8039 (vfmlalq_lane_high_u32): Ditto.
8040 (vfmlslq_lane_high_u32): Ditto.
8041 (vfmlalq_laneq_high_u32): Ditto.
8042 (vfmlslq_laneq_high_u32): Ditto.
8043 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8044 (AARCH64_FL_FOR_ARCH8_4): New.
8045 (AARCH64_ISA_F16FML): New ISA flag.
8046 (TARGET_F16FML): New feature flag for fp16fml.
8047 (doc/invoke.texi): Document new fp16fml option.
8049 2018-01-10 Michael Collison <michael.collison@arm.com>
8051 * config/aarch64/aarch64-builtins.c:
8052 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8053 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8054 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8055 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8056 (AARCH64_ISA_SHA3): New ISA flag.
8057 (TARGET_SHA3): New feature flag for sha3.
8058 * config/aarch64/iterators.md (sha512_op): New int attribute.
8059 (CRYPTO_SHA512): New int iterator.
8060 (UNSPEC_SHA512H): New unspec.
8061 (UNSPEC_SHA512H2): Ditto.
8062 (UNSPEC_SHA512SU0): Ditto.
8063 (UNSPEC_SHA512SU1): Ditto.
8064 * config/aarch64/aarch64-simd-builtins.def
8065 (aarch64_crypto_sha512hqv2di): New builtin.
8066 (aarch64_crypto_sha512h2qv2di): Ditto.
8067 (aarch64_crypto_sha512su0qv2di): Ditto.
8068 (aarch64_crypto_sha512su1qv2di): Ditto.
8069 (aarch64_eor3qv8hi): Ditto.
8070 (aarch64_rax1qv2di): Ditto.
8071 (aarch64_xarqv2di): Ditto.
8072 (aarch64_bcaxqv8hi): Ditto.
8073 * config/aarch64/aarch64-simd.md:
8074 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8075 (aarch64_crypto_sha512su0qv2di): Ditto.
8076 (aarch64_crypto_sha512su1qv2di): Ditto.
8077 (aarch64_eor3qv8hi): Ditto.
8078 (aarch64_rax1qv2di): Ditto.
8079 (aarch64_xarqv2di): Ditto.
8080 (aarch64_bcaxqv8hi): Ditto.
8081 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8082 (vsha512h2q_u64): Ditto.
8083 (vsha512su0q_u64): Ditto.
8084 (vsha512su1q_u64): Ditto.
8085 (veor3q_u16): Ditto.
8086 (vrax1q_u64): Ditto.
8088 (vbcaxq_u16): Ditto.
8089 * config/arm/types.md (crypto_sha512): New type attribute.
8090 (crypto_sha3): Ditto.
8091 (doc/invoke.texi): Document new sha3 option.
8093 2018-01-10 Michael Collison <michael.collison@arm.com>
8095 * config/aarch64/aarch64-builtins.c:
8096 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8097 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8098 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8099 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8100 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8101 (AARCH64_ISA_SM4): New ISA flag.
8102 (TARGET_SM4): New feature flag for sm4.
8103 * config/aarch64/aarch64-simd-builtins.def
8104 (aarch64_sm3ss1qv4si): Ditto.
8105 (aarch64_sm3tt1aq4si): Ditto.
8106 (aarch64_sm3tt1bq4si): Ditto.
8107 (aarch64_sm3tt2aq4si): Ditto.
8108 (aarch64_sm3tt2bq4si): Ditto.
8109 (aarch64_sm3partw1qv4si): Ditto.
8110 (aarch64_sm3partw2qv4si): Ditto.
8111 (aarch64_sm4eqv4si): Ditto.
8112 (aarch64_sm4ekeyqv4si): Ditto.
8113 * config/aarch64/aarch64-simd.md:
8114 (aarch64_sm3ss1qv4si): Ditto.
8115 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8116 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8117 (aarch64_sm4eqv4si): Ditto.
8118 (aarch64_sm4ekeyqv4si): Ditto.
8119 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8120 (sm3part_op): Ditto.
8121 (CRYPTO_SM3TT): Ditto.
8122 (CRYPTO_SM3PART): Ditto.
8123 (UNSPEC_SM3SS1): New unspec.
8124 (UNSPEC_SM3TT1A): Ditto.
8125 (UNSPEC_SM3TT1B): Ditto.
8126 (UNSPEC_SM3TT2A): Ditto.
8127 (UNSPEC_SM3TT2B): Ditto.
8128 (UNSPEC_SM3PARTW1): Ditto.
8129 (UNSPEC_SM3PARTW2): Ditto.
8130 (UNSPEC_SM4E): Ditto.
8131 (UNSPEC_SM4EKEY): Ditto.
8132 * config/aarch64/constraints.md (Ui2): New constraint.
8133 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8134 * config/arm/types.md (crypto_sm3): New type attribute.
8135 (crypto_sm4): Ditto.
8136 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8137 (vsm3tt1aq_u32): Ditto.
8138 (vsm3tt1bq_u32): Ditto.
8139 (vsm3tt2aq_u32): Ditto.
8140 (vsm3tt2bq_u32): Ditto.
8141 (vsm3partw1q_u32): Ditto.
8142 (vsm3partw2q_u32): Ditto.
8143 (vsm4eq_u32): Ditto.
8144 (vsm4ekeyq_u32): Ditto.
8145 (doc/invoke.texi): Document new sm4 option.
8147 2018-01-10 Michael Collison <michael.collison@arm.com>
8149 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8150 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8151 (AARCH64_FL_FOR_ARCH8_4): New.
8152 (AARCH64_FL_V8_4): New flag.
8153 (doc/invoke.texi): Document new armv8.4-a option.
8155 2018-01-10 Michael Collison <michael.collison@arm.com>
8157 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8158 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8159 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8160 * config/aarch64/aarch64-option-extension.def: Add
8161 AARCH64_OPT_EXTENSION of 'sha2'.
8162 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8163 (crypto): Disable sha2 and aes if crypto disabled.
8164 (crypto): Enable aes and sha2 if enabled.
8165 (simd): Disable sha2 and aes if simd disabled.
8166 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8168 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8169 (TARGET_SHA2): New feature flag for sha2.
8170 (TARGET_AES): New feature flag for aes.
8171 * config/aarch64/aarch64-simd.md:
8172 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8173 conditional on TARGET_AES.
8174 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8175 (aarch64_crypto_sha1hsi): Make pattern conditional
8177 (aarch64_crypto_sha1hv4si): Ditto.
8178 (aarch64_be_crypto_sha1hv4si): Ditto.
8179 (aarch64_crypto_sha1su1v4si): Ditto.
8180 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8181 (aarch64_crypto_sha1su0v4si): Ditto.
8182 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8183 (aarch64_crypto_sha256su0v4si): Ditto.
8184 (aarch64_crypto_sha256su1v4si): Ditto.
8185 (doc/invoke.texi): Document new aes and sha2 options.
8187 2018-01-10 Martin Sebor <msebor@redhat.com>
8189 PR tree-optimization/83781
8190 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8193 2018-01-11 Martin Sebor <msebor@gmail.com>
8194 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8196 PR tree-optimization/83501
8197 PR tree-optimization/81703
8199 * tree-ssa-strlen.c (get_string_cst): Rename...
8200 (get_string_len): ...to this. Handle global constants.
8201 (handle_char_store): Adjust.
8203 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8204 Jim Wilson <jimw@sifive.com>
8206 * config/riscv/riscv-protos.h (riscv_output_return): New.
8207 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8208 (riscv_attribute_table, riscv_output_return),
8209 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8210 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8211 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8213 (riscv_expand_prologue): Add early return for naked function.
8214 (riscv_expand_epilogue): Likewise.
8215 (riscv_function_ok_for_sibcall): Return false for naked function.
8216 (riscv_set_current_function): New.
8217 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8218 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8219 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8220 * doc/extend.texi (RISC-V Function Attributes): New.
8222 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8224 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8225 check for 128-bit long double before checking TCmode.
8226 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8227 128-bit long doubles before checking TFmode or TCmode.
8228 (FLOAT128_IBM_P): Likewise.
8230 2018-01-10 Martin Sebor <msebor@redhat.com>
8232 PR tree-optimization/83671
8233 * builtins.c (c_strlen): Unconditionally return zero for the empty
8235 Use -Warray-bounds for warnings.
8236 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8237 for non-constant array indices with COMPONENT_REF, arrays of
8238 arrays, and pointers to arrays.
8239 (gimple_fold_builtin_strlen): Determine and set length range for
8240 non-constant character arrays.
8242 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8245 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8248 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8250 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8252 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8255 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8256 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8257 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8258 indexed_or_indirect_operand predicate.
8259 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8260 (*vsx_le_perm_load_v8hi): Likewise.
8261 (*vsx_le_perm_load_v16qi): Likewise.
8262 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8263 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8264 (*vsx_le_perm_store_v8hi): Likewise.
8265 (*vsx_le_perm_store_v16qi): Likewise.
8266 (eight unnamed splitters): Likewise.
8268 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8270 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8271 * config/rs6000/emmintrin.h: Likewise.
8272 * config/rs6000/mmintrin.h: Likewise.
8273 * config/rs6000/xmmintrin.h: Likewise.
8275 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8278 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8280 * tree.c (tree_nop_conversion): Return true for location wrapper
8282 (maybe_wrap_with_location): New function.
8283 (selftest::check_strip_nops): New function.
8284 (selftest::test_location_wrappers): New function.
8285 (selftest::tree_c_tests): Call it.
8286 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8287 (maybe_wrap_with_location): New decl.
8288 (EXPR_LOCATION_WRAPPER_P): New macro.
8289 (location_wrapper_p): New inline function.
8290 (tree_strip_any_location_wrapper): New inline function.
8292 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8295 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8296 stack_realign_offset for the largest alignment of stack slot
8298 (ix86_find_max_used_stack_alignment): New function.
8299 (ix86_finalize_stack_frame_flags): Use it. Set
8300 max_used_stack_alignment if we don't realign stack.
8301 * config/i386/i386.h (machine_function): Add
8302 max_used_stack_alignment.
8304 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8306 * config/arm/arm.opt (-mbranch-cost): New option.
8307 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8310 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8313 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8314 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8316 2018-01-10 Richard Biener <rguenther@suse.de>
8319 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8320 early out so it also covers the case where we have a non-NULL
8323 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8325 PR tree-optimization/83753
8326 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8327 for non-strided grouped accesses if the number of elements is 1.
8329 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8332 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8333 * i386.h (TARGET_USE_GATHER): Define.
8334 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8336 2018-01-10 Martin Liska <mliska@suse.cz>
8339 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8340 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8342 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8343 CLEANUP_NO_PARTITIONING is not set.
8345 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8347 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8348 for vectors, as a partial revert of r254296.
8349 * rtl.h (const_vec_p): Delete.
8350 (const_vec_duplicate_p): Don't test for vector CONSTs.
8351 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8352 * expmed.c (make_tree): Likewise.
8355 * common.md (E, F): Use CONSTANT_P instead of checking for
8357 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8358 checking for CONST_VECTOR.
8360 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8363 * predict.c (force_edge_cold): Handle in more sane way edges
8366 2018-01-09 Carl Love <cel@us.ibm.com>
8368 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8370 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8371 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8372 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8373 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8374 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8375 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8376 * config/rs6000/rs6000-protos.h: Add extern defition for
8377 rs6000_generate_float2_double_code.
8378 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8380 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8381 (float2_v2df): Add define_expand.
8383 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8386 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8387 op_mode in the force_to_mode call.
8389 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8391 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8392 instead of checking each element individually.
8393 (aarch64_evpc_uzp): Likewise.
8394 (aarch64_evpc_zip): Likewise.
8395 (aarch64_evpc_ext): Likewise.
8396 (aarch64_evpc_rev): Likewise.
8397 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8398 instead of checking each element individually. Return true without
8400 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8401 whether all selected elements come from the same input, instead of
8402 checking each element individually. Remove calls to gen_rtx_REG,
8403 start_sequence and end_sequence and instead assert that no rtl is
8406 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8408 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8409 order of HIGH and CONST checks.
8411 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8413 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8414 if the destination isn't an SSA_NAME.
8416 2018-01-09 Richard Biener <rguenther@suse.de>
8418 PR tree-optimization/83668
8419 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8421 (canonicalize_loop_form): ... here, renamed from ...
8422 (canonicalize_loop_closed_ssa_form): ... this and amended to
8423 swap successor edges for loop exit blocks to make us use
8424 the RPO order we need for initial schedule generation.
8426 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8428 PR tree-optimization/64811
8429 * match.pd: When optimizing comparisons with Inf, avoid
8430 introducing or losing exceptions from comparisons with NaN.
8432 2018-01-09 Martin Liska <mliska@suse.cz>
8435 * asan.c (shadow_mem_size): Add gcc_assert.
8437 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8439 Don't save registers in main().
8442 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8443 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8444 * config/avr/avr.c (avr_set_current_function): Don't error if
8445 naked, OS_task or OS_main are specified at the same time.
8446 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8448 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8450 * common/config/avr/avr-common.c (avr_option_optimization_table):
8451 Switch on -mmain-is-OS_task for optimizing compilations.
8453 2018-01-09 Richard Biener <rguenther@suse.de>
8455 PR tree-optimization/83572
8456 * graphite.c: Include cfganal.h.
8457 (graphite_transform_loops): Connect infinite loops to exit
8458 and remove fake edges at the end.
8460 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8462 * ipa-inline.c (edge_badness): Revert accidental checkin.
8464 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8467 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8468 symbols; not inline clones.
8470 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8473 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8474 hard registers. Formatting fixes.
8476 PR preprocessor/83722
8477 * gcc.c (try_generate_repro): Pass
8478 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8479 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8482 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8483 Kito Cheng <kito.cheng@gmail.com>
8485 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8486 (riscv_leaf_function_p): Delete.
8487 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8489 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8491 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8493 (do_ifelse): New function.
8494 (do_isel): New function.
8495 (do_sub3): New function.
8496 (do_add3): New function.
8497 (do_load_mask_compare): New function.
8498 (do_overlap_load_compare): New function.
8499 (expand_compare_loop): New function.
8500 (expand_block_compare): Call expand_compare_loop() when appropriate.
8501 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8503 (-mblock-compare-inline-loop-limit): New option.
8505 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8508 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8509 Reverse order of second and third operands in first alternative.
8510 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8511 of first and second elements in UNSPEC_VPERMR vector.
8512 (altivec_expand_vec_perm_le): Likewise.
8514 2018-01-08 Jeff Law <law@redhat.com>
8516 PR rtl-optimizatin/81308
8517 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8518 (process_switch): If group_case_labels makes a change, then set
8520 (pass_convert_switch::execute): If a switch is converted, then
8521 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8523 PR rtl-optimization/81308
8524 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8527 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8529 PR target/83663 - Revert r255946
8530 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8531 generation for cases where splatting a value is not useful.
8532 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8533 across a vec_duplicate and a paradoxical subreg forming a vector
8534 mode to a vec_concat.
8536 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8538 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8539 -march=armv8.3-a variants.
8540 * config/arm/t-multilib: Likewise.
8541 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8543 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8545 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8547 (cceq_ior_compare_complement): Give it a name so I can use it, and
8548 change boolean_or_operator predicate to boolean_operator so it can
8549 be used to generate a crand.
8550 (eqne): New code iterator.
8551 (bd/bd_neg): New code_attrs.
8552 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8553 a single define_insn.
8554 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8555 decrement (bdnzt/bdnzf/bdzt/bdzf).
8556 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8557 with the new names of the branch decrement patterns, and added the
8558 names of the branch decrement conditional patterns.
8560 2018-01-08 Richard Biener <rguenther@suse.de>
8562 PR tree-optimization/83563
8563 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8566 2018-01-08 Richard Biener <rguenther@suse.de>
8569 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8571 2018-01-08 Richard Biener <rguenther@suse.de>
8573 PR tree-optimization/83685
8574 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8575 references to abnormals.
8577 2018-01-08 Richard Biener <rguenther@suse.de>
8580 * dwarf2out.c (output_indirect_strings): Handle empty
8581 skeleton_debug_str_hash.
8582 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8584 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8586 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8587 (emit_store_direct): Likewise.
8588 (arc_trampoline_adjust_address): Likewise.
8589 (arc_asm_trampoline_template): New function.
8590 (arc_initialize_trampoline): Use asm_trampoline_template.
8591 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8592 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8593 * config/arc/arc.md (flush_icache): Delete pattern.
8595 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8597 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8598 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8601 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8604 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8605 by not USED_FOR_TARGET.
8606 (make_pass_resolve_sw_modes): Likewise.
8608 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8610 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8613 2018-01-08 Richard Biener <rguenther@suse.de>
8616 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8618 2018-01-08 Richard Biener <rguenther@suse.de>
8621 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8623 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8626 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8627 basic blocks with a small number of successors.
8628 (convert_control_dep_chain_into_preds): Improve handling of
8630 (dump_predicates): Split apart into...
8631 (dump_pred_chain): ...here...
8632 (dump_pred_info): ...and here.
8633 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8634 (can_chain_union_be_invalidated_p): Improve check for invalidation
8636 (uninit_uses_cannot_happen): Avoid unnecessary if
8637 convert_control_dep_chain_into_preds yielded nothing.
8639 2018-01-06 Martin Sebor <msebor@redhat.com>
8641 PR tree-optimization/83640
8642 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8643 subtracting negative offset from size.
8644 (builtin_access::overlap): Adjust offset bounds of the access to fall
8645 within the size of the object if possible.
8647 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8649 PR rtl-optimization/83699
8650 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8651 extract_bit_field_as_subreg to cases in which the extracted
8652 value is also a vector.
8654 * lra-constraints.c (process_alt_operands): Test for the equivalence
8655 substitutions when detecting a possible reload cycle.
8657 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8660 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8661 by default if flag_selective_schedling{,2}. Formatting fixes.
8663 PR rtl-optimization/83682
8664 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8665 if it has non-VECTOR_MODE element mode.
8666 (vec_duplicate_p): Likewise.
8669 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8670 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8672 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8675 * config/i386/i386-builtin.def
8676 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8677 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8678 Require also OPTION_MASK_ISA_AVX512F in addition to
8679 OPTION_MASK_ISA_GFNI.
8680 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8681 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8682 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8683 to OPTION_MASK_ISA_GFNI.
8684 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8685 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8686 OPTION_MASK_ISA_AVX512BW.
8687 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8688 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8689 addition to OPTION_MASK_ISA_GFNI.
8690 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8691 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8692 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8693 to OPTION_MASK_ISA_GFNI.
8694 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8695 a requirement for all ISAs rather than any of them with a few
8697 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8699 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8700 bitmasks to be enabled with 3 exceptions, instead of requiring any
8701 enabled ISA with lots of exceptions.
8702 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8703 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8704 Change avx512bw in isa attribute to avx512f.
8705 * config/i386/sgxintrin.h: Add license boilerplate.
8706 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8707 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8708 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8709 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8711 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8712 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8713 temporarily sse2 rather than sse if not enabled already.
8716 * config/i386/sse.md (VI248_VLBW): Rename to ...
8717 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8718 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8719 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8720 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8721 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8722 mode iterator instead of VI248_VLBW.
8724 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8726 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8727 (record_modified): Skip clobbers; add debug output.
8728 (param_change_prob): Use sreal frequencies.
8730 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8732 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8733 punt for user-aligned variables.
8735 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8737 * tree-chrec.c (chrec_contains_symbols): Return true for
8740 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8743 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8744 of (x|y) == x for BICS pattern.
8746 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8748 PR tree-optimization/83605
8749 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8750 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8753 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8755 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8756 * config/epiphany/rtems.h: New file.
8758 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8759 Uros Bizjak <ubizjak@gmail.com>
8762 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8763 QIreg_operand instead of register_operand predicate.
8764 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8765 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8766 comments instead of -fmitigate[-_]rop.
8768 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8771 * cgraphunit.c (symbol_table::compile): Switch to text_section
8772 before calling assembly_start debug hook.
8773 * run-rtl-passes.c (run_rtl_passes): Likewise.
8776 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8778 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8779 range_int_cst_p rather than !symbolic_range_p before calling
8780 extract_range_from_multiplicative_op_1.
8782 2018-01-04 Jeff Law <law@redhat.com>
8784 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8785 redundant test in assertion.
8787 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8789 * doc/rtl.texi: Document machine_mode wrapper classes.
8791 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8793 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8796 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8798 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8799 the VEC_PERM_EXPR fold to fail.
8801 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8804 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8805 to switched_sections.
8807 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8810 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8813 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8816 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8817 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8819 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8822 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8823 is BLKmode and bitpos not zero or mode change is needed.
8825 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8828 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8831 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8834 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8835 instead of MULT rtx. Update all corresponding splitters.
8837 (*ssub<modesuffix>): Ditto.
8839 (*cmp_sadd_di): Update split patterns.
8840 (*cmp_sadd_si): Ditto.
8841 (*cmp_sadd_sidi): Ditto.
8842 (*cmp_ssub_di): Ditto.
8843 (*cmp_ssub_si): Ditto.
8844 (*cmp_ssub_sidi): Ditto.
8845 * config/alpha/predicates.md (const23_operand): New predicate.
8846 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8847 Look for ASHIFT, not MULT inner operand.
8848 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8850 2018-01-04 Martin Liska <mliska@suse.cz>
8852 PR gcov-profile/83669
8853 * gcov.c (output_intermediate_file): Add version to intermediate
8855 * doc/gcov.texi: Document new field 'version' in intermediate
8856 file format. Fix location of '-k' option of gcov command.
8858 2018-01-04 Martin Liska <mliska@suse.cz>
8861 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8863 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8865 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8867 2018-01-03 Martin Sebor <msebor@redhat.com>
8869 PR tree-optimization/83655
8870 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8871 checking calls with invalid arguments.
8873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8875 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8876 (vectorizable_mask_load_store): Delete.
8877 (vectorizable_call): Return false for masked loads and stores.
8878 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8879 instead of gimple_assign_rhs1.
8880 (vectorizable_load): Handle IFN_MASK_LOAD.
8881 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8885 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8887 (vectorizable_mask_load_store): ...here.
8888 (vectorizable_load): ...and here.
8890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8892 * tree-vect-stmts.c (vect_build_all_ones_mask)
8893 (vect_build_zero_merge_argument): New functions, split out from...
8894 (vectorizable_load): ...here.
8896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8898 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8900 (vectorizable_mask_load_store): ...here.
8901 (vectorizable_store): ...and here.
8903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8905 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8907 (vectorizable_mask_load_store): ...here.
8909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8911 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8912 (vect_model_store_cost): Take a vec_load_store_type instead of a
8914 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8915 (vect_model_store_cost): Take a vec_load_store_type instead of a
8917 (vectorizable_mask_load_store): Update accordingly.
8918 (vectorizable_store): Likewise.
8919 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8923 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8924 IFN_MASK_LOAD calls here rather than...
8925 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8928 Alan Hayward <alan.hayward@arm.com>
8929 David Sherwood <david.sherwood@arm.com>
8931 * expmed.c (extract_bit_field_1): For vector extracts,
8932 fall back to extract_bit_field_as_subreg if vec_extract
8935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8936 Alan Hayward <alan.hayward@arm.com>
8937 David Sherwood <david.sherwood@arm.com>
8939 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8940 they are variable or constant sized.
8941 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8942 slots for constant-sized data.
8944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8945 Alan Hayward <alan.hayward@arm.com>
8946 David Sherwood <david.sherwood@arm.com>
8948 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8949 handling COND_EXPRs with boolean comparisons, try to find a better
8950 basis for the mask type than the boolean itself.
8952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8954 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8955 is calculated and how it can be overridden.
8956 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8957 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8959 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8963 Alan Hayward <alan.hayward@arm.com>
8964 David Sherwood <david.sherwood@arm.com>
8966 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8967 Remove the mode argument.
8968 (aarch64_simd_valid_immediate): Remove the mode and inverse
8970 * config/aarch64/iterators.md (bitsize): New iterator.
8971 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8972 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8973 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8974 aarch64_simd_valid_immediate.
8975 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8976 (aarch64_reg_or_bic_imm): Likewise.
8977 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8978 with an insn_type enum and msl with a modifier_type enum.
8979 Replace element_width with a scalar_mode. Change the shift
8980 to unsigned int. Add constructors for scalar_float_mode and
8981 scalar_int_mode elements.
8982 (aarch64_vect_float_const_representable_p): Delete.
8983 (aarch64_can_const_movi_rtx_p)
8984 (aarch64_simd_scalar_immediate_valid_for_move)
8985 (aarch64_simd_make_constant): Update call to
8986 aarch64_simd_valid_immediate.
8987 (aarch64_advsimd_valid_immediate_hs): New function.
8988 (aarch64_advsimd_valid_immediate): Likewise.
8989 (aarch64_simd_valid_immediate): Remove mode and inverse
8990 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8991 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8992 and aarch64_float_const_representable_p on the result.
8993 (aarch64_output_simd_mov_immediate): Remove mode argument.
8994 Update call to aarch64_simd_valid_immediate and use of
8995 simd_immediate_info.
8996 (aarch64_output_scalar_simd_mov_immediate): Update call
8999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9000 Alan Hayward <alan.hayward@arm.com>
9001 David Sherwood <david.sherwood@arm.com>
9003 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9004 (mode_nunits): Likewise CONST_MODE_NUNITS.
9005 * machmode.def (ADJUST_NUNITS): Document.
9006 * genmodes.c (mode_data::need_nunits_adj): New field.
9007 (blank_mode): Update accordingly.
9008 (adj_nunits): New variable.
9009 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9011 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9012 listed in adj_nunits.
9013 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9014 listed in adj_nunits. Don't emit case statements for such modes.
9015 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9016 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9017 nothing if adj_nunits is nonnull.
9018 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9019 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9020 (emit_mode_fbit): Update use of print_maybe_const_decl.
9021 (emit_move_size): Likewise. Treat the array as non-const
9023 (emit_mode_adjustments): Handle adj_nunits.
9025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9027 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9028 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9029 (VECTOR_MODES): Use it.
9030 (make_vector_modes): Take the prefix as an argument.
9032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9033 Alan Hayward <alan.hayward@arm.com>
9034 David Sherwood <david.sherwood@arm.com>
9036 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9037 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9038 for MODE_VECTOR_BOOL.
9039 * machmode.def (VECTOR_BOOL_MODE): Document.
9040 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9041 (make_vector_bool_mode): New function.
9042 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9044 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9045 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9047 * stor-layout.c (int_mode_for_mode): Likewise.
9048 * tree.c (build_vector_type_for_mode): Likewise.
9049 * varasm.c (output_constant_pool_2): Likewise.
9050 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9051 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9052 for MODE_VECTOR_BOOL.
9053 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9054 of mode class checks.
9055 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9056 instead of a list of mode class checks.
9057 (expand_vector_scalar_condition): Likewise.
9058 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9061 Alan Hayward <alan.hayward@arm.com>
9062 David Sherwood <david.sherwood@arm.com>
9064 * machmode.h (mode_size): Change from unsigned short to
9066 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9067 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9068 or if measurement_type is not polynomial.
9069 (fixed_size_mode::includes_p): Check for constant-sized modes.
9070 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9071 return a poly_uint16 rather than an unsigned short.
9072 (emit_mode_size): Change the type of mode_size from unsigned short
9073 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9074 (emit_mode_adjustments): Cope with polynomial vector sizes.
9075 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9077 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9079 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9080 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9081 * caller-save.c (setup_save_areas): Likewise.
9082 (replace_reg_with_saved_mem): Likewise.
9083 * calls.c (emit_library_call_value_1): Likewise.
9084 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9085 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9086 (gen_lowpart_for_combine): Likewise.
9087 * convert.c (convert_to_integer_1): Likewise.
9088 * cse.c (equiv_constant, cse_insn): Likewise.
9089 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9090 (cselib_subst_to_values): Likewise.
9091 * dce.c (word_dce_process_block): Likewise.
9092 * df-problems.c (df_word_lr_mark_ref): Likewise.
9093 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9094 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9095 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9096 (rtl_for_decl_location): Likewise.
9097 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9098 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9099 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9100 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9101 (expand_expr_real_1): Likewise.
9102 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9103 (pad_below): Likewise.
9104 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9105 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9106 * ira.c (get_subreg_tracking_sizes): Likewise.
9107 * ira-build.c (ira_create_allocno_objects): Likewise.
9108 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9109 (ira_sort_regnos_for_alter_reg): Likewise.
9110 * ira-costs.c (record_operand_costs): Likewise.
9111 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9112 (resolve_simple_move): Likewise.
9113 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9114 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9115 (lra_constraints): Likewise.
9116 (CONST_POOL_OK_P): Reject variable-sized modes.
9117 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9118 (add_pseudo_to_slot, lra_spill): Likewise.
9119 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9120 * optabs-query.c (get_best_extraction_insn): Likewise.
9121 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9122 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9123 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9124 * recog.c (offsettable_address_addr_space_p): Likewise.
9125 * regcprop.c (maybe_mode_change): Likewise.
9126 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9127 * regrename.c (build_def_use): Likewise.
9128 * regstat.c (dump_reg_info): Likewise.
9129 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9130 (find_reloads, find_reloads_subreg_address): Likewise.
9131 * reload1.c (eliminate_regs_1): Likewise.
9132 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9133 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9134 (simplify_binary_operation_1, simplify_subreg): Likewise.
9135 * targhooks.c (default_function_arg_padding): Likewise.
9136 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9137 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9138 (verify_gimple_assign_ternary): Likewise.
9139 * tree-inline.c (estimate_move_cost): Likewise.
9140 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9141 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9142 (get_address_cost_ainc): Likewise.
9143 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9144 (vect_supportable_dr_alignment): Likewise.
9145 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9146 (vectorizable_reduction): Likewise.
9147 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9148 (vectorizable_operation, vectorizable_load): Likewise.
9149 * tree.c (build_same_sized_truth_vector_type): Likewise.
9150 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9151 * var-tracking.c (emit_note_insn_var_location): Likewise.
9152 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9153 (ADDR_VEC_ALIGN): Likewise.
9155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9156 Alan Hayward <alan.hayward@arm.com>
9157 David Sherwood <david.sherwood@arm.com>
9159 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9161 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9162 or if measurement_type is polynomial.
9163 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9164 * combine.c (make_extraction): Likewise.
9165 * dse.c (find_shift_sequence): Likewise.
9166 * dwarf2out.c (mem_loc_descriptor): Likewise.
9167 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9168 (extract_bit_field, extract_low_bits): Likewise.
9169 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9170 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9171 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9172 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9173 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9174 * reload.c (find_reloads): Likewise.
9175 * reload1.c (alter_reg): Likewise.
9176 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9177 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9178 * tree-if-conv.c (predicate_mem_writes): Likewise.
9179 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9180 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9181 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9182 * valtrack.c (dead_debug_insert_temp): Likewise.
9183 * varasm.c (mergeable_constant_section): Likewise.
9184 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9187 Alan Hayward <alan.hayward@arm.com>
9188 David Sherwood <david.sherwood@arm.com>
9190 * expr.c (expand_assignment): Cope with polynomial mode sizes
9191 when assigning to a CONCAT.
9193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9194 Alan Hayward <alan.hayward@arm.com>
9195 David Sherwood <david.sherwood@arm.com>
9197 * machmode.h (mode_precision): Change from unsigned short to
9199 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9201 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9202 or if measurement_type is not polynomial.
9203 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9204 in which the mode is already known to be a scalar_int_mode.
9205 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9206 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9208 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9209 for GET_MODE_PRECISION.
9210 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9211 for GET_MODE_PRECISION.
9212 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9214 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9215 (expand_field_assignment, make_extraction): Likewise.
9216 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9217 (get_last_value): Likewise.
9218 * convert.c (convert_to_integer_1): Likewise.
9219 * cse.c (cse_insn): Likewise.
9220 * expr.c (expand_expr_real_1): Likewise.
9221 * lra-constraints.c (simplify_operand_subreg): Likewise.
9222 * optabs-query.c (can_atomic_load_p): Likewise.
9223 * optabs.c (expand_atomic_load): Likewise.
9224 (expand_atomic_store): Likewise.
9225 * ree.c (combine_reaching_defs): Likewise.
9226 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9227 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9228 * tree.h (type_has_mode_precision_p): Likewise.
9229 * ubsan.c (instrument_si_overflow): Likewise.
9231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9232 Alan Hayward <alan.hayward@arm.com>
9233 David Sherwood <david.sherwood@arm.com>
9235 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9236 polynomial numbers of units.
9237 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9238 (valid_vector_subparts_p): New function.
9239 (build_vector_type): Remove temporary shim and take the number
9240 of units as a poly_uint64 rather than an int.
9241 (build_opaque_vector_type): Take the number of units as a
9242 poly_uint64 rather than an int.
9243 * tree.c (build_vector_from_ctor): Handle polynomial
9244 TYPE_VECTOR_SUBPARTS.
9245 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9246 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9247 (build_vector_from_val): If the number of units is variable,
9248 use build_vec_duplicate_cst for constant operands and
9249 VEC_DUPLICATE_EXPR otherwise.
9250 (make_vector_type): Remove temporary is_constant ().
9251 (build_vector_type, build_opaque_vector_type): Take the number of
9252 units as a poly_uint64 rather than an int.
9253 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9255 * cfgexpand.c (expand_debug_expr): Likewise.
9256 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9257 (store_constructor, expand_expr_real_1): Likewise.
9258 (const_scalar_mask_from_tree): Likewise.
9259 * fold-const-call.c (fold_const_reduction): Likewise.
9260 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9261 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9262 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9263 (fold_relational_const): Likewise.
9264 (native_interpret_vector): Likewise. Change the size from an
9265 int to an unsigned int.
9266 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9267 TYPE_VECTOR_SUBPARTS.
9268 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9269 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9270 duplicating a non-constant operand into a variable-length vector.
9271 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9272 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9273 * ipa-icf.c (sem_variable::equals): Likewise.
9274 * match.pd: Likewise.
9275 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9276 * print-tree.c (print_node): Likewise.
9277 * stor-layout.c (layout_type): Likewise.
9278 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9279 * tree-cfg.c (verify_gimple_comparison): Likewise.
9280 (verify_gimple_assign_binary): Likewise.
9281 (verify_gimple_assign_ternary): Likewise.
9282 (verify_gimple_assign_single): Likewise.
9283 * tree-pretty-print.c (dump_generic_node): Likewise.
9284 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9285 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9286 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9287 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9288 (vect_shift_permute_load_chain): Likewise.
9289 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9290 (expand_vector_condition, optimize_vector_constructor): Likewise.
9291 (lower_vec_perm, get_compute_type): Likewise.
9292 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9293 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9294 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9295 (vect_recog_mask_conversion_pattern): Likewise.
9296 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9297 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9298 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9299 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9300 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9301 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9302 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9303 (supportable_widening_operation): Likewise.
9304 (supportable_narrowing_operation): Likewise.
9305 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9307 * varasm.c (output_constant): Likewise.
9309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9310 Alan Hayward <alan.hayward@arm.com>
9311 David Sherwood <david.sherwood@arm.com>
9313 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9314 so that both the length == 3 and length != 3 cases set up their
9315 own permute vectors. Add comments explaining why we know the
9316 number of elements is constant.
9317 (vect_permute_load_chain): Likewise.
9319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9320 Alan Hayward <alan.hayward@arm.com>
9321 David Sherwood <david.sherwood@arm.com>
9323 * machmode.h (mode_nunits): Change from unsigned char to
9325 (ONLY_FIXED_SIZE_MODES): New macro.
9326 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9327 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9328 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9330 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9331 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9332 or if measurement_type is not polynomial.
9333 * genmodes.c (ZERO_COEFFS): New macro.
9334 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9336 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9337 Use ZERO_COEFFS when emitting initializers.
9338 * data-streamer.h (bp_pack_poly_value): New function.
9339 (bp_unpack_poly_value): Likewise.
9340 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9341 for GET_MODE_NUNITS.
9342 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9343 for GET_MODE_NUNITS.
9344 * tree.c (make_vector_type): Remove temporary shim and make
9345 the real function take the number of units as a poly_uint64
9347 (build_vector_type_for_mode): Handle polynomial nunits.
9348 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9349 * emit-rtl.c (const_vec_series_p_1): Likewise.
9350 (gen_rtx_CONST_VECTOR): Likewise.
9351 * fold-const.c (test_vec_duplicate_folding): Likewise.
9352 * genrecog.c (validate_pattern): Likewise.
9353 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9354 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9355 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9356 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9357 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9358 * rtlanal.c (subreg_get_info): Likewise.
9359 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9360 (vect_grouped_load_supported): Likewise.
9361 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9362 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9363 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9364 (simplify_const_unary_operation, simplify_binary_operation_1)
9365 (simplify_const_binary_operation, simplify_ternary_operation)
9366 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9367 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9368 instead of CONST_VECTOR_NUNITS.
9369 * varasm.c (output_constant_pool_2): Likewise.
9370 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9371 explicit-encoded elements in the XVEC for variable-length vectors.
9373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9375 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9378 Alan Hayward <alan.hayward@arm.com>
9379 David Sherwood <david.sherwood@arm.com>
9381 * coretypes.h (fixed_size_mode): Declare.
9382 (fixed_size_mode_pod): New typedef.
9383 * builtins.h (target_builtins::x_apply_args_mode)
9384 (target_builtins::x_apply_result_mode): Change type to
9385 fixed_size_mode_pod.
9386 * builtins.c (apply_args_size, apply_result_size, result_vector)
9387 (expand_builtin_apply_args_1, expand_builtin_apply)
9388 (expand_builtin_return): Update accordingly.
9390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9392 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9393 * cselib.c (cselib_hash_rtx): Likewise.
9394 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9395 CONST_VECTOR encoding.
9397 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9398 Jeff Law <law@redhat.com>
9401 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9402 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9403 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9404 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9407 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9408 explicitly probe *sp in a noreturn function if there were any callee
9409 register saves or frame pointer is needed.
9411 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9414 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9415 BLKmode for ternary, binary or unary expressions.
9418 * var-tracking.c (delete_vta_debug_insn): New inline function.
9419 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9420 insns from get_insns () to NULL instead of each bb separately.
9421 Use delete_vta_debug_insn. No longer static.
9422 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9423 delete_vta_debug_insns callers.
9424 * rtl.h (delete_vta_debug_insns): Declare.
9425 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9426 instead of variable_tracking_main.
9428 2018-01-03 Martin Sebor <msebor@redhat.com>
9430 PR tree-optimization/83603
9431 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9432 arguments past the endof the argument list in functions declared
9433 without a prototype.
9434 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9435 Avoid checking when arguments are null.
9437 2018-01-03 Martin Sebor <msebor@redhat.com>
9440 * doc/extend.texi (attribute const): Fix a typo.
9441 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9442 issuing -Wsuggest-attribute for void functions.
9444 2018-01-03 Martin Sebor <msebor@redhat.com>
9446 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9447 offset_int::from instead of wide_int::to_shwi.
9448 (maybe_diag_overlap): Remove assertion.
9449 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9450 * gimple-ssa-sprintf.c (format_directive): Same.
9451 (parse_directive): Same.
9452 (sprintf_dom_walker::compute_format_length): Same.
9453 (try_substitute_return_value): Same.
9455 2018-01-03 Jeff Law <law@redhat.com>
9458 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9459 non-constant residual for zero at runtime and avoid probing in
9460 that case. Reorganize code for trailing problem to mirror handling
9463 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9465 PR tree-optimization/83501
9466 * tree-ssa-strlen.c (get_string_cst): New.
9467 (handle_char_store): Call get_string_cst.
9469 2018-01-03 Martin Liska <mliska@suse.cz>
9471 PR tree-optimization/83593
9472 * tree-ssa-strlen.c: Include tree-cfg.h.
9473 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9474 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9475 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9477 (strlen_dom_walker::before_dom_children): Call
9478 gimple_purge_dead_eh_edges. Dump tranformation with details
9480 (strlen_dom_walker::before_dom_children): Update call by adding
9481 new argument cleanup_eh.
9482 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9484 2018-01-03 Martin Liska <mliska@suse.cz>
9487 * cif-code.def (VARIADIC_THUNK): New enum value.
9488 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9491 2018-01-03 Jan Beulich <jbeulich@suse.com>
9493 * sse.md (mov<mode>_internal): Tighten condition for when to use
9494 vmovdqu<ssescalarsize> for TI and OI modes.
9496 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9498 Update copyright years.
9500 2018-01-03 Martin Liska <mliska@suse.cz>
9503 * ipa-visibility.c (function_and_variable_visibility): Skip
9504 functions with noipa attribure.
9506 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9508 * gcc.c (process_command): Update copyright notice dates.
9509 * gcov-dump.c (print_version): Ditto.
9510 * gcov.c (print_version): Ditto.
9511 * gcov-tool.c (print_version): Ditto.
9512 * gengtype.c (create_file): Ditto.
9513 * doc/cpp.texi: Bump @copying's copyright year.
9514 * doc/cppinternals.texi: Ditto.
9515 * doc/gcc.texi: Ditto.
9516 * doc/gccint.texi: Ditto.
9517 * doc/gcov.texi: Ditto.
9518 * doc/install.texi: Ditto.
9519 * doc/invoke.texi: Ditto.
9521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9523 * vector-builder.h (vector_builder::m_full_nelts): Change from
9524 unsigned int to poly_uint64.
9525 (vector_builder::full_nelts): Update prototype accordingly.
9526 (vector_builder::new_vector): Likewise.
9527 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9528 (vector_builder::operator ==): Likewise.
9529 (vector_builder::finalize): Likewise.
9530 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9531 Take the number of elements as a poly_uint64 rather than an
9533 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9534 from unsigned int to poly_uint64.
9535 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9536 (vec_perm_indices::new_vector): Likewise.
9537 (vec_perm_indices::length): Likewise.
9538 (vec_perm_indices::nelts_per_input): Likewise.
9539 (vec_perm_indices::input_nelts): Likewise.
9540 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9541 number of elements per input as a poly_uint64 rather than an
9542 unsigned int. Use the original encoding for variable-length
9543 vectors, rather than clamping each individual element.
9544 For the second and subsequent elements in each pattern,
9545 clamp the step and base before clamping their sum.
9546 (vec_perm_indices::series_p): Handle polynomial element counts.
9547 (vec_perm_indices::all_in_range_p): Likewise.
9548 (vec_perm_indices_to_tree): Likewise.
9549 (vec_perm_indices_to_rtx): Likewise.
9550 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9551 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9552 (tree_vector_builder::new_binary_operation): Handle polynomial
9553 element counts. Return false if we need to know the number
9554 of elements at compile time.
9555 * fold-const.c (fold_vec_perm): Punt if the number of elements
9556 isn't known at compile time.
9558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9560 * vec-perm-indices.h (vec_perm_builder): Change element type
9561 from HOST_WIDE_INT to poly_int64.
9562 (vec_perm_indices::element_type): Update accordingly.
9563 (vec_perm_indices::clamp): Handle polynomial element_types.
9564 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9565 (vec_perm_indices::all_in_range_p): Likewise.
9566 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9568 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9569 polynomial vec_perm_indices element types.
9570 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9571 * fold-const.c (fold_vec_perm): Likewise.
9572 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9573 * tree-vect-generic.c (lower_vec_perm): Likewise.
9574 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9575 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9576 element type to HOST_WIDE_INT.
9578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9579 Alan Hayward <alan.hayward@arm.com>
9580 David Sherwood <david.sherwood@arm.com>
9582 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9583 rather than an int. Use plus_constant.
9584 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9585 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9588 Alan Hayward <alan.hayward@arm.com>
9589 David Sherwood <david.sherwood@arm.com>
9591 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9592 a HOST_WIDE_INT to a poly_int64.
9594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9595 Alan Hayward <alan.hayward@arm.com>
9596 David Sherwood <david.sherwood@arm.com>
9598 * calls.c (load_register_parameters): Cope with polynomial
9599 mode sizes. Require a constant size for BLKmode parameters
9600 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9601 forces a parameter to be padded at the lsb end in order to
9602 fill a complete number of words, require the parameter size
9603 to be ordered wrt UNITS_PER_WORD.
9605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9606 Alan Hayward <alan.hayward@arm.com>
9607 David Sherwood <david.sherwood@arm.com>
9609 * reload1.c (spill_stack_slot_width): Change element type
9610 from unsigned int to poly_uint64_pod.
9611 (alter_reg): Treat mode sizes as polynomial.
9613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9614 Alan Hayward <alan.hayward@arm.com>
9615 David Sherwood <david.sherwood@arm.com>
9617 * reload.c (complex_word_subreg_p): New function.
9618 (reload_inner_reg_of_subreg, push_reload): Use it.
9620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9621 Alan Hayward <alan.hayward@arm.com>
9622 David Sherwood <david.sherwood@arm.com>
9624 * lra-constraints.c (process_alt_operands): Reject matched
9625 operands whose sizes aren't ordered.
9626 (match_reload): Refer to this check here.
9628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9629 Alan Hayward <alan.hayward@arm.com>
9630 David Sherwood <david.sherwood@arm.com>
9632 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9633 that the mode size is in the set {1, 2, 4, 8, 16}.
9635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9636 Alan Hayward <alan.hayward@arm.com>
9637 David Sherwood <david.sherwood@arm.com>
9639 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9640 Use plus_constant instead of gen_rtx_PLUS.
9642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9643 Alan Hayward <alan.hayward@arm.com>
9644 David Sherwood <david.sherwood@arm.com>
9646 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9647 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9648 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9649 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9650 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9651 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9652 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9653 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9654 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9655 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9657 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9658 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9659 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9660 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9661 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9662 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9663 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9664 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9665 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9666 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9668 * expr.c (emit_move_resolve_push): Treat the input and result
9669 of PUSH_ROUNDING as a poly_int64.
9670 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9671 (emit_push_insn): Likewise.
9672 * lra-eliminations.c (mark_not_eliminable): Likewise.
9673 * recog.c (push_operand): Likewise.
9674 * reload1.c (elimination_effects): Likewise.
9675 * rtlanal.c (nonzero_bits1): Likewise.
9676 * calls.c (store_one_arg): Likewise. Require the padding to be
9677 known at compile time.
9679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9680 Alan Hayward <alan.hayward@arm.com>
9681 David Sherwood <david.sherwood@arm.com>
9683 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9684 Use plus_constant instead of gen_rtx_PLUS.
9686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9687 Alan Hayward <alan.hayward@arm.com>
9688 David Sherwood <david.sherwood@arm.com>
9690 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9694 Alan Hayward <alan.hayward@arm.com>
9695 David Sherwood <david.sherwood@arm.com>
9697 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9698 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9699 via stack temporaries. Treat the mode size as polynomial too.
9701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9702 Alan Hayward <alan.hayward@arm.com>
9703 David Sherwood <david.sherwood@arm.com>
9705 * expr.c (expand_expr_real_2): When handling conversions involving
9706 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9707 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9708 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9717 Alan Hayward <alan.hayward@arm.com>
9718 David Sherwood <david.sherwood@arm.com>
9720 * combine.c (can_change_dest_mode): Handle polynomial
9721 REGMODE_NATURAL_SIZE.
9722 * expmed.c (store_bit_field_1): Likewise.
9723 * expr.c (store_constructor): Likewise.
9724 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9725 and polynomial REGMODE_NATURAL_SIZE.
9726 (gen_lowpart_common): Likewise.
9727 * reginfo.c (record_subregs_of_mode): Likewise.
9728 * rtlanal.c (read_modify_subreg_p): Likewise.
9730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9731 Alan Hayward <alan.hayward@arm.com>
9732 David Sherwood <david.sherwood@arm.com>
9734 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9735 numbers of 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 * match.pd: Cope with polynomial numbers of vector elements.
9743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9744 Alan Hayward <alan.hayward@arm.com>
9745 David Sherwood <david.sherwood@arm.com>
9747 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9748 in a POINTER_PLUS_EXPR.
9750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9751 Alan Hayward <alan.hayward@arm.com>
9752 David Sherwood <david.sherwood@arm.com>
9754 * omp-simd-clone.c (simd_clone_subparts): New function.
9755 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9756 (ipa_simd_modify_function_body): Likewise.
9758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9759 Alan Hayward <alan.hayward@arm.com>
9760 David Sherwood <david.sherwood@arm.com>
9762 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9763 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9764 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9765 (expand_vector_condition, vector_element): Likewise.
9766 (subparts_gt): New function.
9767 (get_compute_type): Use subparts_gt.
9768 (count_type_subparts): Delete.
9769 (expand_vector_operations_1): Use subparts_gt instead of
9770 count_type_subparts.
9772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9773 Alan Hayward <alan.hayward@arm.com>
9774 David Sherwood <david.sherwood@arm.com>
9776 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9777 (vect_compile_time_alias): ...this new function. Do the calculation
9778 on poly_ints rather than trees.
9779 (vect_prune_runtime_alias_test_list): Update call accordingly.
9781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9782 Alan Hayward <alan.hayward@arm.com>
9783 David Sherwood <david.sherwood@arm.com>
9785 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9787 (vect_schedule_slp_instance): Likewise.
9789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9790 Alan Hayward <alan.hayward@arm.com>
9791 David Sherwood <david.sherwood@arm.com>
9793 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9794 constant and extern definitions for variable-length vectors.
9795 (vect_get_constant_vectors): Note that the number of units
9796 is known to be constant.
9798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9799 Alan Hayward <alan.hayward@arm.com>
9800 David Sherwood <david.sherwood@arm.com>
9802 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9803 of units as polynomial. Choose between WIDE and NARROW based
9806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9807 Alan Hayward <alan.hayward@arm.com>
9808 David Sherwood <david.sherwood@arm.com>
9810 * tree-vect-stmts.c (simd_clone_subparts): New function.
9811 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9814 Alan Hayward <alan.hayward@arm.com>
9815 David Sherwood <david.sherwood@arm.com>
9817 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9818 vectors as polynomial. Use build_index_vector for
9821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9822 Alan Hayward <alan.hayward@arm.com>
9823 David Sherwood <david.sherwood@arm.com>
9825 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9826 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9827 for variable-length vectors.
9828 (vectorizable_mask_load_store): Treat the number of units as
9829 polynomial, asserting that it is constant if the condition has
9830 already been enforced.
9831 (vectorizable_store, vectorizable_load): Likewise.
9833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9834 Alan Hayward <alan.hayward@arm.com>
9835 David Sherwood <david.sherwood@arm.com>
9837 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9838 of units as polynomial. Punt if we can't tell at compile time
9839 which vector contains the final result.
9841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9842 Alan Hayward <alan.hayward@arm.com>
9843 David Sherwood <david.sherwood@arm.com>
9845 * tree-vect-loop.c (vectorizable_induction): Treat the number
9846 of units as polynomial. Punt on SLP inductions. Use an integer
9847 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9848 cast of such a series for variable-length floating-point
9851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9852 Alan Hayward <alan.hayward@arm.com>
9853 David Sherwood <david.sherwood@arm.com>
9855 * tree.h (build_index_vector): Declare.
9856 * tree.c (build_index_vector): New function.
9857 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9858 of units as polynomial, forcibly converting it to a constant if
9859 vectorizable_reduction has already enforced the condition.
9860 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9861 to create a {1,2,3,...} vector.
9862 (vectorizable_reduction): Treat the number of units as polynomial.
9863 Choose vectype_in based on the largest scalar element size rather
9864 than the smallest number of units. Enforce the restrictions
9867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9868 Alan Hayward <alan.hayward@arm.com>
9869 David Sherwood <david.sherwood@arm.com>
9871 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9872 number of units as polynomial.
9874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9875 Alan Hayward <alan.hayward@arm.com>
9876 David Sherwood <david.sherwood@arm.com>
9878 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9879 * target.def (autovectorize_vector_sizes): Return the vector sizes
9880 by pointer, using vector_sizes rather than a bitmask.
9881 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9882 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9883 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9885 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9886 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9887 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9888 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9889 * omp-general.c (omp_max_vf): Likewise.
9890 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9891 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9892 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9893 * tree-vect-slp.c (vect_slp_bb): Likewise.
9894 * doc/tm.texi: Regenerate.
9895 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9897 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9898 the vector size as a poly_uint64 rather than an unsigned int.
9899 (current_vector_size): Change from an unsigned int to a poly_uint64.
9900 (get_vectype_for_scalar_type): Update accordingly.
9901 * tree.h (build_truth_vector_type): Take the size and number of
9902 units as a poly_uint64 rather than an unsigned int.
9903 (build_vector_type): Add a temporary overload that takes
9904 the number of units as a poly_uint64 rather than an unsigned int.
9905 * tree.c (make_vector_type): Likewise.
9906 (build_truth_vector_type): Take the number of units as a poly_uint64
9907 rather than an unsigned int.
9909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9910 Alan Hayward <alan.hayward@arm.com>
9911 David Sherwood <david.sherwood@arm.com>
9913 * target.def (get_mask_mode): Take the number of units and length
9914 as poly_uint64s rather than unsigned ints.
9915 * targhooks.h (default_get_mask_mode): Update accordingly.
9916 * targhooks.c (default_get_mask_mode): Likewise.
9917 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9918 * doc/tm.texi: Regenerate.
9920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9921 Alan Hayward <alan.hayward@arm.com>
9922 David Sherwood <david.sherwood@arm.com>
9924 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9925 * omp-general.c (omp_max_vf): Likewise.
9926 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9927 (expand_omp_simd): Handle polynomial safelen.
9928 * omp-low.c (omplow_simd_context): Add a default constructor.
9929 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9930 (lower_rec_simd_input_clauses): Update accordingly.
9931 (lower_rec_input_clauses): Likewise.
9933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9934 Alan Hayward <alan.hayward@arm.com>
9935 David Sherwood <david.sherwood@arm.com>
9937 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9938 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9939 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9940 (vect_analyze_slp_cost): Likewise.
9941 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9942 (vect_model_load_cost): Likewise.
9944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9945 Alan Hayward <alan.hayward@arm.com>
9946 David Sherwood <david.sherwood@arm.com>
9948 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9949 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9950 from an unsigned int * to a poly_uint64_pod *.
9951 (calculate_unrolling_factor): New function.
9952 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9955 Alan Hayward <alan.hayward@arm.com>
9956 David Sherwood <david.sherwood@arm.com>
9958 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9959 from an unsigned int to a poly_uint64.
9960 (_loop_vec_info::slp_unrolling_factor): Likewise.
9961 (_loop_vec_info::vectorization_factor): Change from an int
9963 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9964 (vect_get_num_vectors): New function.
9965 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9966 (vect_get_num_copies): Use vect_get_num_vectors.
9967 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9968 to an unsigned int *.
9969 (vect_analyze_data_refs): Change min_vf from an int * to a
9971 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9972 than an unsigned HOST_WIDE_INT.
9973 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9974 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9975 to an unsigned int *.
9976 (vect_analyze_data_ref_dependences): Likewise.
9977 (vect_compute_data_ref_alignment): Handle polynomial vf.
9978 (vect_enhance_data_refs_alignment): Likewise.
9979 (vect_prune_runtime_alias_test_list): Likewise.
9980 (vect_shift_permute_load_chain): Likewise.
9981 (vect_supportable_dr_alignment): Likewise.
9982 (dependence_distance_ge_vf): Take the vectorization factor as a
9983 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9984 (vect_analyze_data_refs): Change min_vf from an int * to a
9986 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9987 vfm1 as a poly_uint64 rather than an int. Make the same change
9988 for the returned bound_scalar.
9989 (vect_gen_vector_loop_niters): Handle polynomial vf.
9990 (vect_do_peeling): Likewise. Update call to
9991 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9992 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9994 * tree-vect-loop.c (vect_determine_vectorization_factor)
9995 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9996 (vect_get_known_peeling_cost): Likewise.
9997 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9998 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9999 (vect_transform_loop): Likewise. Use the lowest possible VF when
10000 updating the upper bounds of the loop.
10001 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10002 rather than an int.
10003 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10004 polynomial unroll factors.
10005 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10006 (vect_make_slp_decision): Likewise.
10007 (vect_supported_load_permutation_p): Likewise, and polynomial
10009 (vect_analyze_slp_cost): Handle polynomial vf.
10010 (vect_slp_analyze_node_operations): Likewise.
10011 (vect_slp_analyze_bb_1): Likewise.
10012 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10013 than an unsigned HOST_WIDE_INT.
10014 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10015 (vectorizable_load): Handle polynomial vf.
10016 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10018 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10021 Alan Hayward <alan.hayward@arm.com>
10022 David Sherwood <david.sherwood@arm.com>
10024 * match.pd: Handle bit operations involving three constants
10025 and try to fold one pair.
10027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10029 * tree-vect-loop-manip.c: Include gimple-fold.h.
10030 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10031 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10032 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10033 Add a path that uses a step of VF instead of 1, but disable it
10035 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10036 and niters_no_overflow parameters. Update calls to
10037 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10038 Create a new SSA name if the latter choses to use a ste other
10039 than zero, and return it via niters_vector_mult_vf_var.
10040 * tree-vect-loop.c (vect_transform_loop): Update calls to
10041 vect_do_peeling, vect_gen_vector_loop_niters and
10042 slpeel_make_loop_iterate_ntimes.
10043 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10044 (vect_gen_vector_loop_niters): Update declarations after above changes.
10046 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10048 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10049 128-bit round to integer instructions.
10050 (ceil<mode>2): Likewise.
10051 (btrunc<mode>2): Likewise.
10052 (round<mode>2): Likewise.
10054 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10056 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10057 unaligned VSX load/store on P8/P9.
10058 (expand_block_clear): Allow the use of unaligned VSX
10059 load/store on P8/P9.
10061 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10063 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10065 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10066 swap associated with both a load and a store.
10068 2018-01-02 Andrew Waterman <andrew@sifive.com>
10070 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10071 * config/riscv/riscv.md (clear_cache): Use it.
10073 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10075 * web.c: Remove out-of-date comment.
10077 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10079 * expr.c (fixup_args_size_notes): Check that any existing
10080 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10081 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10082 (emit_single_push_insn): ...here.
10084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10086 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10087 (const_vector_encoded_nelts): New function.
10088 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10089 (const_vector_int_elt, const_vector_elt): Declare.
10090 * emit-rtl.c (const_vector_int_elt_1): New function.
10091 (const_vector_elt): Likewise.
10092 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10093 of CONST_VECTOR_ELT.
10095 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10097 * expr.c: Include rtx-vector-builder.h.
10098 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10099 directly on the tree encoding.
10100 (const_vector_from_tree): Likewise.
10101 * optabs.c: Include rtx-vector-builder.h.
10102 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10103 sequence of "u" values.
10104 * vec-perm-indices.c: Include rtx-vector-builder.h.
10105 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10106 directly on the vec_perm_indices encoding.
10108 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10110 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10111 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10112 * rtx-vector-builder.h: New file.
10113 * rtx-vector-builder.c: Likewise.
10114 * rtl.h (rtx_def::u2): Add a const_vector field.
10115 (CONST_VECTOR_NPATTERNS): New macro.
10116 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10117 (CONST_VECTOR_DUPLICATE_P): Likewise.
10118 (CONST_VECTOR_STEPPED_P): Likewise.
10119 (CONST_VECTOR_ENCODED_ELT): Likewise.
10120 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10121 (unwrap_const_vec_duplicate): Likewise.
10122 (const_vec_series_p): Check for a non-duplicated vector encoding.
10123 Say that the function only returns true for integer vectors.
10124 * emit-rtl.c: Include rtx-vector-builder.h.
10125 (gen_const_vec_duplicate_1): Delete.
10126 (gen_const_vector): Call gen_const_vec_duplicate instead of
10127 gen_const_vec_duplicate_1.
10128 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10129 (gen_const_vec_duplicate): Use rtx_vector_builder.
10130 (gen_const_vec_series): Likewise.
10131 (gen_rtx_CONST_VECTOR): Likewise.
10132 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10133 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10134 Build a new vector rather than modifying a CONST_VECTOR in-place.
10135 (handle_special_swappables): Update call accordingly.
10136 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10137 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10138 Build a new vector rather than modifying a CONST_VECTOR in-place.
10139 (handle_special_swappables): Update call accordingly.
10141 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10143 * simplify-rtx.c (simplify_const_binary_operation): Use
10144 CONST_VECTOR_ELT instead of XVECEXP.
10146 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10148 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10149 the selector elements to be different from the data elements
10150 if the selector is a VECTOR_CST.
10151 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10152 ssizetype for the selector.
10154 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10156 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10157 before testing each element individually.
10158 * tree-vect-generic.c (lower_vec_perm): Likewise.
10160 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10162 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10163 * selftest-run-tests.c (selftest::run_tests): Call it.
10164 * vector-builder.h (vector_builder::operator ==): New function.
10165 (vector_builder::operator !=): Likewise.
10166 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10167 (vec_perm_indices::all_from_input_p): New function.
10168 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10169 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10170 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10171 instead of reading the VECTOR_CST directly. Detect whether both
10172 vector inputs are the same before constructing the vec_perm_indices,
10173 and update the number of inputs argument accordingly. Use the
10174 utility functions added above. Only construct sel2 if we need to.
10176 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10178 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10179 the broadcast of the low byte.
10180 (expand_mult_highpart): Use an explicit encoding for the permutes.
10181 * optabs-query.c (can_mult_highpart_p): Likewise.
10182 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10183 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10184 (vectorizable_bswap): Likewise.
10185 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10186 explicit encoding for the power-of-2 permutes.
10187 (vect_permute_store_chain): Likewise.
10188 (vect_grouped_load_supported): Likewise.
10189 (vect_permute_load_chain): Likewise.
10191 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10193 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10194 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10195 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10196 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10197 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10198 (vect_gen_perm_mask_any): Likewise.
10200 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10202 * int-vector-builder.h: New file.
10203 * vec-perm-indices.h: Include int-vector-builder.h.
10204 (vec_perm_indices): Redefine as an int_vector_builder.
10205 (auto_vec_perm_indices): Delete.
10206 (vec_perm_builder): Redefine as a stand-alone class.
10207 (vec_perm_indices::vec_perm_indices): New function.
10208 (vec_perm_indices::clamp): Likewise.
10209 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10210 (vec_perm_indices::new_vector): New function.
10211 (vec_perm_indices::new_expanded_vector): Update for new
10212 vec_perm_indices class.
10213 (vec_perm_indices::rotate_inputs): New function.
10214 (vec_perm_indices::all_in_range_p): Operate directly on the
10215 encoded form, without computing elided elements.
10216 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10217 encoding. Update for new vec_perm_indices class.
10218 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10219 the given vec_perm_builder.
10220 (expand_vec_perm_var): Update vec_perm_builder constructor.
10221 (expand_mult_highpart): Use vec_perm_builder instead of
10222 auto_vec_perm_indices.
10223 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10224 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10225 or double series encoding as appropriate.
10226 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10227 vec_perm_indices instead of auto_vec_perm_indices.
10228 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10229 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10230 (vect_permute_store_chain): Likewise.
10231 (vect_grouped_load_supported): Likewise.
10232 (vect_permute_load_chain): Likewise.
10233 (vect_shift_permute_load_chain): Likewise.
10234 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10235 (vect_transform_slp_perm_load): Likewise.
10236 (vect_schedule_slp_instance): Likewise.
10237 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10238 (vectorizable_mask_load_store): Likewise.
10239 (vectorizable_bswap): Likewise.
10240 (vectorizable_store): Likewise.
10241 (vectorizable_load): Likewise.
10242 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10243 vec_perm_indices instead of auto_vec_perm_indices. Use
10244 tree_to_vec_perm_builder to read the vector from a tree.
10245 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10246 vec_perm_builder instead of a vec_perm_indices.
10247 (have_whole_vector_shift): Use vec_perm_builder and
10248 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10249 truncation to calc_vec_perm_mask_for_shift.
10250 (vect_create_epilog_for_reduction): Likewise.
10251 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10252 from auto_vec_perm_indices to vec_perm_indices.
10253 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10254 instead of changing individual elements.
10255 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10256 the vector in d.perm.
10257 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10258 from auto_vec_perm_indices to vec_perm_indices.
10259 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10260 instead of changing individual elements.
10261 (arm_vectorize_vec_perm_const): Use new_vector to install
10262 the vector in d.perm.
10263 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10264 Update vec_perm_builder constructor.
10265 (rs6000_expand_interleave): Likewise.
10266 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10267 (rs6000_expand_interleave): Likewise.
10269 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10271 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10272 to qimode could truncate the indices.
10273 * optabs.c (expand_vec_perm_var): Likewise.
10275 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10277 * Makefile.in (OBJS): Add vec-perm-indices.o.
10278 * vec-perm-indices.h: New file.
10279 * vec-perm-indices.c: Likewise.
10280 * target.h (vec_perm_indices): Replace with a forward class
10282 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10283 * optabs.h: Include vec-perm-indices.h.
10284 (expand_vec_perm): Delete.
10285 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10286 (expand_vec_perm_const): Declare.
10287 * target.def (vec_perm_const_ok): Replace with...
10288 (vec_perm_const): ...this new hook.
10289 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10290 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10291 * doc/tm.texi: Regenerate.
10292 * optabs.def (vec_perm_const): Delete.
10293 * doc/md.texi (vec_perm_const): Likewise.
10294 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10295 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10296 expand_vec_perm for constant permutation vectors. Assert that
10297 the mode of variable permutation vectors is the integer equivalent
10298 of the mode that is being permuted.
10299 * optabs-query.h (selector_fits_mode_p): Declare.
10300 * optabs-query.c: Include vec-perm-indices.h.
10301 (selector_fits_mode_p): New function.
10302 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10303 is defined, instead of checking whether the vec_perm_const_optab
10304 exists. Use targetm.vectorize.vec_perm_const instead of
10305 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10306 fit in the vector mode before using a variable permute.
10307 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10308 vec_perm_indices instead of an rtx.
10309 (expand_vec_perm): Replace with...
10310 (expand_vec_perm_const): ...this new function. Take the selector
10311 as a vec_perm_indices rather than an rtx. Also take the mode of
10312 the selector. Update call to shift_amt_for_vec_perm_mask.
10313 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10314 Use vec_perm_indices::new_expanded_vector to expand the original
10315 selector into bytes. Check whether the indices fit in the vector
10316 mode before using a variable permute.
10317 (expand_vec_perm_var): Make global.
10318 (expand_mult_highpart): Use expand_vec_perm_const.
10319 * fold-const.c: Includes vec-perm-indices.h.
10320 * tree-ssa-forwprop.c: Likewise.
10321 * tree-vect-data-refs.c: Likewise.
10322 * tree-vect-generic.c: Likewise.
10323 * tree-vect-loop.c: Likewise.
10324 * tree-vect-slp.c: Likewise.
10325 * tree-vect-stmts.c: Likewise.
10326 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10328 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10329 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10330 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10331 (aarch64_vectorize_vec_perm_const): ...this new function.
10332 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10333 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10334 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10335 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10336 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10337 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10338 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10340 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10341 check for NEON modes.
10342 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10343 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10344 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10345 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10347 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10348 the old VEC_PERM_CONST conditions.
10349 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10350 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10351 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10352 (ia64_vectorize_vec_perm_const_ok): Merge into...
10353 (ia64_vectorize_vec_perm_const): ...this new function.
10354 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10355 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10356 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10357 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10358 * config/mips/mips.c (mips_expand_vec_perm_const)
10359 (mips_vectorize_vec_perm_const_ok): Merge into...
10360 (mips_vectorize_vec_perm_const): ...this new function.
10361 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10362 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10363 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10364 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10365 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10366 (rs6000_expand_vec_perm_const): Delete.
10367 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10369 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10370 (altivec_expand_vec_perm_const_le): Take each operand individually.
10371 Operate on constant selectors rather than rtxes.
10372 (altivec_expand_vec_perm_const): Likewise. Update call to
10373 altivec_expand_vec_perm_const_le.
10374 (rs6000_expand_vec_perm_const): Delete.
10375 (rs6000_vectorize_vec_perm_const_ok): Delete.
10376 (rs6000_vectorize_vec_perm_const): New function.
10377 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10378 an element count and rtx array.
10379 (rs6000_expand_extract_even): Update call accordingly.
10380 (rs6000_expand_interleave): Likewise.
10381 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10382 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10383 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10384 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10385 (rs6000_expand_vec_perm_const): Delete.
10386 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10387 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10388 (altivec_expand_vec_perm_const_le): Take each operand individually.
10389 Operate on constant selectors rather than rtxes.
10390 (altivec_expand_vec_perm_const): Likewise. Update call to
10391 altivec_expand_vec_perm_const_le.
10392 (rs6000_expand_vec_perm_const): Delete.
10393 (rs6000_vectorize_vec_perm_const_ok): Delete.
10394 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10395 reference to the SPE evmerge intructions.
10396 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10397 an element count and rtx array.
10398 (rs6000_expand_extract_even): Update call accordingly.
10399 (rs6000_expand_interleave): Likewise.
10400 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10401 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10403 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10405 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10407 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10408 vector mode and that that mode matches the mode of the data
10410 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10411 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10412 directly using expand_vec_perm_1 when forcing selectors into
10414 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10416 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10418 * optabs-query.h (can_vec_perm_p): Delete.
10419 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10420 * optabs-query.c (can_vec_perm_p): Split into...
10421 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10422 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10423 particular selector is valid.
10424 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10425 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10426 (vect_grouped_load_supported): Likewise.
10427 (vect_shift_permute_load_chain): Likewise.
10428 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10429 (vect_transform_slp_perm_load): Likewise.
10430 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10431 (vectorizable_bswap): Likewise.
10432 (vect_gen_perm_mask_checked): Likewise.
10433 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10434 implementations of variable permutation vectors into account
10435 when deciding which selector to use.
10436 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10437 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10438 with a false third argument.
10439 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10440 to test whether the constant selector is valid and can_vec_perm_var_p
10441 to test whether a variable selector is valid.
10443 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10445 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10446 * optabs-query.c (can_vec_perm_p): Likewise.
10447 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10448 instead of vec_perm_indices.
10449 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10450 (vect_gen_perm_mask_checked): Likewise,
10451 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10452 (vect_gen_perm_mask_checked): Likewise,
10454 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10456 * optabs-query.h (qimode_for_vec_perm): Declare.
10457 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10458 (qimode_for_vec_perm): ...this new function.
10459 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10461 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10463 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10464 does not have a conditional at the top.
10466 2018-01-02 Richard Biener <rguenther@suse.de>
10468 * ipa-inline.c (big_speedup_p): Fix expression.
10470 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10473 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10476 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10480 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10481 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10482 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10483 cond_taken_branch_cost 3->4.
10485 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10487 PR tree-optimization/83581
10488 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10489 TODO_cleanup_cfg if any changes have been made.
10491 PR middle-end/83608
10492 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10493 convert_modes if target mode has the right side, but different mode
10496 PR middle-end/83609
10497 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10498 last argument when extracting from CONCAT. If either from_real or
10499 from_imag is NULL, use expansion through memory. If result is not
10500 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10501 the parts directly to inner mode, if even that fails, use expansion
10504 PR middle-end/83623
10505 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10506 check for bswap in mode rather than HImode and use that in expand_unop
10509 Copyright (C) 2018 Free Software Foundation, Inc.
10511 Copying and distribution of this file, with or without modification,
10512 are permitted in any medium without royalty provided the copyright
10513 notice and this notice are preserved.