1 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7 * lra-assigns.c (assign_by_spills): Return a flag of reload
8 assignment failure. Do not process the reload assignment
9 failures. Do not spill other reload pseudos if they has the same
11 (lra_assign): Add a return arg. Set up from the result of
12 assign_by_spills call.
13 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15 usage_insns if it is not NULL.
16 (spill_hard_reg_in_range): New function.
17 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20 (lra_assign): Change prototype.
21 * lra.c (lra): Add code to deal with fails by splitting hard reg
24 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
27 * config/i386/i386.opt: Replace Enforcment with Enforcement.
29 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
32 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
33 (dw_val_node): Add val_symbolic_view.
34 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
35 (symview_upper_bound): New.
36 (new_line_info_table): Initialize symviews_since_reset.
37 (dwarf2out_source_line): Count symviews_since_reset and set
39 (dw_val_equal_p): Handle symview.
40 (add_AT_symview): New.
41 (print_dw_val): Handle symview.
42 (attr_checksum, attr_checksum_ordered): Likewise.
43 (same_dw_val_p, size_of_die): Likewise.
44 (value_format, output_die): Likewise.
45 (add_high_low_attributes): Use add_AT_symview for entry_view.
46 (dwarf2out_finish): Reset symview_upper_bound, clear
49 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
52 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
53 Add strict argument and use it.
54 (rs6000_split_multireg_move): Update for new strict argument.
55 (mem_operand_gpr): Disallow all non-offsettable addresses.
56 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
58 2018-03-09 Jakub Jelinek <jakub@redhat.com>
61 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
62 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
63 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
66 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
67 decl, use remap_type if we want to use the type.
69 2018-03-09 Martin Sebor <msebor@redhat.com>
71 PR tree-optimization/84526
72 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
74 (builtin_access::generic_overlap): Be prepared to handle non-array
77 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
79 PR rtl-optimization/84682
80 * lra-constraints.c (process_address_1): Check is_address flag
81 for address constraints.
82 (process_alt_operands): Likewise.
83 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
84 preprocess_constraints.
85 * recog.h (preprocess_constraints): Add oploc parameter.
87 * recog.c (preprocess_constraints): Test address_operand for
88 CT_ADDRESS constraints.
90 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
93 * lra-assigns.c (assign_by_spills): Return a flag of reload
94 assignment failure. Do not process the reload assignment
95 failures. Do not spill other reload pseudos if they has the same
97 (lra_assign): Add a return arg. Set up from the result of
98 assign_by_spills call.
99 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
100 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
101 usage_insns if it is not NULL.
102 (spill_hard_reg_in_range): New function.
103 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
104 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
106 (lra_assign): Change prototype.
107 * lra.c (lra): Add code to deal with fails by splitting hard reg
110 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
113 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
114 Accept complain bool parameter. Only emit errors if it is true.
115 (arm_parse_cpu_option_name): Likewise.
116 (arm_target_thumb_only): Adjust callers of the above.
117 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
118 prototype to take a default true bool parameter.
119 (arm_parse_arch_option_name): Likewise.
121 2018-03-09 David Malcolm <dmalcolm@redhat.com>
122 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
126 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
127 * configure: Regenerate.
128 * configure.ac ("linker --version-script option"): New.
129 ("linker soname option"): New.
131 2018-03-09 Richard Biener <rguenther@suse.de>
133 PR tree-optimization/84775
134 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
135 immediate uses of predicate stmts and mark them modified.
138 PR tree-optimization/84178
139 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
141 (version_loop_for_if_conversion): Delay update_ssa call.
142 (tree_if_conversion): Delay update_ssa until after predicate
145 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
148 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
149 when the function accesses prior frames.
151 2018-03-08 Jakub Jelinek <jakub@redhat.com>
154 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
155 gen_llsym, otherwise call maybe_gen_llsym.
158 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
159 has ',' character inside of it.
161 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
164 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
165 as clobbering CC_REGNUM.
167 2018-03-08 Richard Biener <rguenther@suse.de>
170 * tree-scalar-evolution.c: Include tree-into-ssa.h.
171 (follow_copies_to_constant): Do not follow SSA names registered
174 2018-03-08 Richard Biener <rguenther@suse.de>
176 PR tree-optimization/84178
177 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
179 (version_loop_for_if_conversion): Delay update_ssa call.
180 (tree_if_conversion): Delay update_ssa until after predicate
183 2018-03-08 David Malcolm <dmalcolm@redhat.com>
185 PR tree-optimization/84178
186 * tree-if-conv.c (release_bb_predicate): Remove the
187 the assertion that the stmts have NULL use_ops.
188 Discard the statements, asserting that they haven't
189 yet been added to a BB.
191 2018-03-08 Richard Biener <rguenther@suse.de>
193 PR tree-optimization/84746
194 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
195 (phi_translate): Pass in destination ANTIC_OUT set.
196 (phi_translate_1): Likewise. For a simplified result lookup
197 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
198 (phi_translate_set): Adjust.
199 (do_pre_regular_insertion): Likewise.
200 (do_pre_partial_partial_insertion): Likewise.
202 2018-03-08 Martin Liska <mliska@suse.cz>
204 PR gcov-profile/84735
205 * doc/gcov.texi: Document usage of profile files.
206 * gcov-io.h: Document changes in the format.
208 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
212 * dwarf2out.c (struct dw_line_info_table): Update comments for
214 (FORCE_RESET_NEXT_VIEW): New.
215 (FORCE_RESETTING_VIEW_P): New.
216 (RESETTING_VIEW_P): Check for -1 too.
217 (ZERO_VIEW_P): Likewise.
218 (new_line_info_table): Force-reset next view.
219 (dwarf2out_begin_function): Likewise.
220 (dwarf2out_source_line): Simplify zero_view_p initialization.
221 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
222 view directly. Omit view when omitting .loc at line 0.
224 2018-03-08 Jakub Jelinek <jakub@redhat.com>
226 PR tree-optimization/84740
227 * tree-switch-conversion.c (process_switch): Call build_constructors
228 only if info.phi_count is non-zero.
230 PR tree-optimization/84739
231 * tree-tailcall.c (find_tail_calls): Check call arguments against
232 DECL_ARGUMENTS (current_function_decl) rather than
233 DECL_ARGUMENTS (func) when checking for tail recursion.
235 2018-03-07 Jakub Jelinek <jakub@redhat.com>
237 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
238 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
239 Volker Reichelt's entry and add entries for people that perform
240 GCC fuzzy testing and report numerous bugs.
242 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
245 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
246 readonly data in sdata, if that is disabled.
247 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
248 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
249 -mreadonly-in-sdata option.
251 2018-03-07 Martin Sebor <msebor@redhat.com>
253 PR tree-optimization/84468
254 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
255 basic block when looking for nul assignment.
257 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
260 * except.h (output_function_exception_table): Adjust prototype.
261 * except.c (output_function_exception_table): Remove FNNAME parameter
262 and add SECTION parameter. Ouput one part of the table at a time.
263 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
264 the first part of the exception table and emit unwind directives.
265 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
266 (i386_pe_seh_cold_init): Likewise.
267 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
268 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
269 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
270 (ix86_output_call_insn): Emit a nop in one more case for SEH.
271 * config/i386/winnt.c: Include except.h.
272 (struct seh_frame_state): Add reg_offset, after_prologue and
273 in_cold_section fields.
274 (i386_pe_seh_end_prologue): Set seh->after_prologue.
275 (i386_pe_seh_cold_init): New function.
276 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
277 to seh->in_cold_section.
278 (seh_emit_push): Record the offset of the push.
279 (seh_emit_save): Record the offet of the save.
280 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
281 Test seh->after_prologue to disregard the epilogue.
282 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
283 (i386_pe_end_cold_function): New function.
285 2018-03-07 Jakub Jelinek <jakub@redhat.com>
288 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
289 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
292 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
294 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
295 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
298 * multiple_target.c: Include tree-inline.h and intl.h.
299 (expand_target_clones): Diagnose and fail if node->definition and
300 !tree_versionable_function_p (node->decl).
302 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
304 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
306 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
307 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
308 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
310 2018-03-06 Jakub Jelinek <jakub@redhat.com>
313 * combine.c (try_combine): Use reg_or_subregno instead of handling
314 just paradoxical SUBREGs and REGs.
316 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
318 * config/arc/arc.c (arc_finalize_pic): Remove function.
319 (arc_must_save_register): We use single base PIC register, remove
320 checks to save/restore the PIC register.
321 (arc_expand_prologue): Likewise.
322 * config/arc/arc-protos.h (arc_set_default_type_attributes):
324 (arc_verify_short): Likewise.
325 (arc_attr_type): Likewise.
326 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
327 (walk_stores): Likewise.
328 (arc_address_cost): Make it static.
329 (arc_verify_short): Likewise.
330 (branch_dest): Likewise.
331 (arc_attr_type): Likewise.
332 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
333 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
334 (arc_final_prescan_insn): Remove inserting the nops due to
335 hardware hazards. It is done in reorg step.
336 (insn_length_variant_t): Remove.
337 (insn_length_parameters_t): Likewise.
338 (arc_insn_length_parameters): Likewise.
339 (arc_get_insn_variants): Likewise.
340 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
342 2018-03-06 Jakub Jelinek <jakub@redhat.com>
345 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
348 PR tree-optimization/84687
349 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
351 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
353 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
355 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
356 Rename to ppc_speculation_barrier.
357 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
358 __builtin_ppc_speculation_barrier.
360 2018-03-05 Jakub Jelinek <jakub@redhat.com>
363 * combine.c (combine_simplify_rtx): Don't try to simplify if
364 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
367 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
369 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
370 to 32 bytes when compiling for POWER9.
372 2018-03-05 Jakub Jelinek <jakub@redhat.com>
375 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
376 regparm >= 3 with no arg reg available also for calls with
377 flag_force_indirect_call. Pass decl to ix86_function_regparm.
380 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
382 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
384 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
387 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
389 2018-03-05 Richard Biener <rguenther@suse.de>
391 PR tree-optimization/84486
392 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
393 When inserting a __builtin_assume_aligned call set the LHS
394 SSA name alignment info accordingly.
396 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
398 PR tree-optimization/84114
399 * config/aarch64/aarch64.c (aarch64_reassociation_width)
400 Avoid reassociation of FLOAT_MODE addition.
402 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
404 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
405 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
406 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
407 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
408 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
409 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
410 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
412 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
413 __builtin_ia32_wbinvd): New builtins.
414 (SPECIAL_ARGS2): New.
415 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
416 (SPECIAL_ARGS2): New.
417 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
418 (ix86_valid_target_attribute_inner_p): Ditto.
419 (ix86_init_mmx_sse_builtins): Add special_args2.
420 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
421 TARGET_WBNOINVD_P): New.
422 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
423 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
424 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
425 * config/i386/immintrin.h (_wbinvd): New intrinsic.
426 * config/i386/pconfigintrin.h: New file.
427 * config/i386/wbnoinvdintrin.h: Ditto.
428 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
429 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
431 2018-03-05 Richard Biener <rguenther@suse.de>
433 PR tree-optimization/84670
434 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
436 (BB_VISITED_WITH_VISITED_SUCCS): New define.
437 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
438 (compute_antic_aux): Only assert the number of values in ANTIC_IN
439 doesn't grow if all successors (recursively) were visited at least
442 2018-03-05 Richard Biener <rguenther@suse.de>
444 PR tree-optimization/84650
445 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
446 if executed in the loop pipeline.
448 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
450 * doc/configfiles.texi (Configuration Files): Move info about
451 conditionalizing $target-protos.h to...
452 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
453 differs from $target-protos.h.
455 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
456 Chung-Ju Wu <jasonwucj@gmail.com>
458 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
459 * config/nds32/nds32-multiple.md (setmemsi): Define.
460 * config/nds32/nds32-memory-manipulation.c
461 (nds32_gen_dup_4_byte_to_word_value): New.
462 (emit_setmem_word_loop): New.
463 (emit_setmem_byte_loop): New.
464 (nds32_expand_setmem_loop): New.
465 (nds32_expand_setmem_loop_v3m): New.
466 (nds32_expand_setmem_unroll): New.
467 (nds32_expand_setmem): New.
469 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
470 Chung-Ju Wu <jasonwucj@gmail.com>
472 * config/nds32/nds32-memory-manipulation.c
473 (nds32_emit_load_store): New.
474 (nds32_emit_post_inc_load_store): New.
475 (nds32_emit_mem_move): New.
476 (nds32_emit_mem_move_block): New.
477 (nds32_expand_movmemsi_loop_unknown_size): New.
478 (nds32_expand_movmemsi_loop_known_size): New.
479 (nds32_expand_movmemsi_loop): New.
480 (nds32_expand_movmemsi_unroll): New.
481 (nds32_expand_movmemqi): Rename ...
482 (nds32_expand_movmemsi): ... to this.
483 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
484 (movmemsi): ... to this.
485 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
486 (nds32_expand_movmemsi): ... to this.
488 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
489 Monk Chiang <sh.chiang04@gmail.com>
490 Chung-Ju Wu <jasonwucj@gmail.com>
492 * config/nds32/nds32-protos.h
493 (nds32_expand_load_multiple): New arguments.
494 (nds32_expand_store_multiple): Ditto.
495 (nds32_valid_multiple_load_store): Rename ...
496 (nds32_valid_multiple_load_store_p): ... to this.
497 * config/nds32/nds32-memory-manipulation.c
498 (nds32_expand_load_multiple): Refine implementation.
499 (nds32_expand_store_multiple): Ditto.
500 * config/nds32/nds32-multiple.md
501 (load_multiple): Update nds32_expand_load_multiple interface.
502 (store_multiple): Update nds32_expand_store_multiple interface.
503 * config/nds32/nds32-predicates.c
504 (nds32_valid_multiple_load_store): Rename ...
505 (nds32_valid_multiple_load_store_p): ... to this and refine
507 * config/nds32/predicates.md
508 (nds32_load_multiple_and_update_address_operation): New predicate.
509 (nds32_store_multiple_and_update_address_operation): New predicate.
511 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
512 Chung-Ju Wu <jasonwucj@gmail.com>
514 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
515 (combo): New attribute.
516 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
518 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
520 * config/nds32/nds32.opt: Change -mcmodel= default value.
522 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
523 Monk Chiang <sh.chiang04@gmail.com>
524 Chung-Ju Wu <jasonwucj@gmail.com>
526 * config/nds32/constants.md (unspec_element): New enum.
527 * config/nds32/constraints.md (Umw): New constraint.
528 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
529 * config/nds32/nds32-intrinsic.md: Likewise.
530 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
531 (nds32_valid_smw_lwm_base_p): New.
532 (nds32_output_smw_single_word): New.
533 (nds32_output_lmw_single_word): New.
534 (nds32_expand_unaligned_load): New.
535 (nds32_expand_unaligned_store): New.
536 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
537 (nds32_output_smw_single_word): Declare.
538 (nds32_output_lmw_single_word): Declare.
539 (nds32_expand_unaligned_load): Declare.
540 (nds32_expand_unaligned_store): Declare.
541 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
542 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
543 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
544 NDS32_BUILTIN_UASTORE_DW.
545 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
548 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
549 Kito Cheng <kito.cheng@gmail.com>
550 Chung-Ju Wu <jasonwucj@gmail.com>
552 * config/nds32/nds32-intrinsic.c
553 (nds32_expand_builtin_null_ftype_reg): Delete.
554 (nds32_expand_builtin_reg_ftype_imm): Ditto.
555 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
556 (nds32_read_argument): New.
557 (nds32_legitimize_target): Ditto.
558 (nds32_legitimize_argument): Ditto.
559 (nds32_check_constant_argument): Ditto.
560 (nds32_expand_unop_builtin): Ditto.
561 (nds32_expand_unopimm_builtin): Ditto.
562 (nds32_expand_binop_builtin): Ditto.
563 (nds32_builtin_decl_impl): Ditto.
564 (builtin_description): Ditto.
565 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
566 (nds32_init_builtins_impl): Ditto.
567 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
568 (nds32_builtin_decl): New.
569 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
570 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
572 2018-03-02 Jeff Law <law@redhat.com>
574 * reorg.c (stop_search_p): Handle DEBUG_INSN.
575 (redundant_insn, fill_simple_delay_slots): Likewise.
576 (fill_slots_from_thread): Likewise.
577 * resource.c (mark_referenced_resources): Likewise.
578 (mark_set_resources, find_dead_or_set_registers): Likewise.
580 2018-03-02 Jakub Jelinek <jakub@redhat.com>
582 * substring-locations.h (format_warning_va): Formatting fix for
584 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
586 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
587 * substring-locations.c: Include intl.h.
588 (format_warning_va): Turned into small wrapper around
589 format_warning_n_va, renamed to ...
590 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
591 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
593 (format_warning_at_substring_n): New function.
594 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
595 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
596 format_warning_at_substring with just a shorter name instead of
597 const function pointer.
598 (fmtwarn_n): New function.
599 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
600 appropriate, get rid of all the fmtstr temporaries, move conditionals
601 with G_() wrapped string literals directly into fmtwarn arguments,
602 cast dir.len to (int), formatting fixes.
604 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
606 * doc/invoke.texi: Remove "Cilk Plus" references.
608 2018-03-02 Jakub Jelinek <jakub@redhat.com>
609 Richard Biener <rguenther@suse.de>
612 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
613 for error or warning attributes if CALL_FROM_THUNK_P is set.
616 2018-03-02 Jakub Jelinek <jakub@redhat.com>
619 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
620 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
623 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
624 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
626 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
627 instead of -1U in last predictors element's probability member.
629 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
632 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
633 arguments if they are comparable.
635 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
637 PR tree-optimization/84634
638 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
639 masks and masked_loop_p with a single loop_masks, making sure it's
640 null for bb vectorization.
642 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
644 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
645 (vect_analyze_data_ref_access): Use loop->safe_len rather than
646 loop->force_vectorize to check whether there is no alias.
648 2018-03-02 Jakub Jelinek <jakub@redhat.com>
651 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
653 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
655 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
656 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
657 instead of a loop around prev_real_insn.
658 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
662 * config/i386/i386.c (ix86_print_operand): Use conditional
663 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
666 2018-03-02 Richard Biener <rguenther@suse.de>
668 PR tree-optimization/84427
669 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
670 (bitmap_set_subtract_values): Rewrite to handle multiple
673 (prune_clobbered_mems): Likewise.
674 (phi_translate): Take edge instead of pred/phiblock.
675 (phi_translate_1): Likewise.
676 (phi_translate_set): Likewise. Insert all translated
677 exprs for a value into the set, keeping possibly multiple
678 expressions per value.
679 (compute_antic_aux): Adjust for phi_translate changes.
680 When intersecting union the expressions and prune those
681 not in the final value set, keeping possibly multiple
682 expressions per value. Do not use value-insertion
683 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
684 all expressions. Add verification that the value-sets
685 only shrink during iteration.
686 (compute_partial_antic_aux): Adjust for the phi_translate changes.
687 (do_pre_regular_insertion): Likewise.
688 (do_pre_partial_partial_insertion): Likewise.
690 2018-03-02 Richard Biener <rguenther@suse.de>
693 * config/darwin.c (saved_debug_info_level): New static global.
694 (darwin_asm_lto_start): Disable debug info generation for LTO out.
695 (darwin_asm_lto_end): Restore debug info generation settings.
697 2018-03-01 Martin Liska <mliska@suse.cz>
700 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
703 2018-03-01 Richard Biener <rguenther@suse.de>
706 * dwarf2out.c (gen_variable_die): Properly handle late VLA
707 type annotation with LTO when debug was disabled at compile-time.
709 2018-03-01 Matthew Fortune <mfortune@gmail.com>
711 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
713 (mips_final_postscan_insn): Likewise.
715 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
717 PR rtl-optimization/84528
718 * alias.c (init_alias_target): Add commentary.
719 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
720 a unique base value if the frame pointer is not eliminated
721 to the stack pointer.
723 2018-03-01 Tom de Vries <tom@codesourcery.com>
725 PR rtl-optimization/83327
726 * lra-int.h (hard_regs_spilled_into): Declare.
727 * lra.c (hard_regs_spilled_into): Define.
728 (init_reg_info): Init hard_regs_spilled_into.
729 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
730 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
731 (process_bb_lives): Handle hard_regs_spilled_into.
732 (lra_create_live_ranges_1): Before doing liveness propagation, clear
733 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
735 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
737 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
738 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
739 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
740 * config/rs6000/aix72.h: New file.
742 2018-02-28 Jakub Jelinek <jakub@redhat.com>
744 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
745 instead of warning_at with conditional singular and plural messages
749 * stor-layout.c (update_alignment_for_field): For
750 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
751 && !DECL_PACKED (field), do the alignment update, just use
752 only desired_align instead of MAX (type_align, desired_align)
754 (place_field): Don't do known_align < desired_align handling
755 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
756 is non-NULL, instead do it after rli->prev_field handling and
757 only if not within a bitfield word. For DECL_PACKED (field)
758 use type_align of BITS_PER_UNIT.
760 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
762 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
763 superfluous parentheses and trailing spaces.
765 2018-02-28 Richard Biener <rguenther@suse.de>
767 PR tree-optimization/84584
768 * graphite-scop-detection.c (scop_detection::add_scop): Discard
769 SCoPs with fake exit edge.
771 2018-02-28 Martin Liska <mliska@suse.cz>
774 * timevar.c (timer::print): Fix format to properly print 100%
777 2018-02-28 Richard Biener <rguenther@suse.de>
780 * genmatch.c (capture_info::walk_match): Do not mark
781 captured expressions without operands as expr_p given
782 they act more like predicates and should be subject to
783 "lost tail" side-effect preserving.
785 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
787 PR rtl-optimization/81611
788 * auto-inc-dec.c (attempt_change): Move dead note from
789 mem_insn if it's the next use of regno
790 (find_address): Take address use of reg holding
791 non-incremented value. Add parm to limit search to the named
793 (merge_in_block): Attempt to use a mem insn that is the next
794 use of the original regno.
796 2018-02-27 Martin Sebor <msebor@redhat.com>
799 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
800 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
802 2018-02-27 Martin Sebor <msebor@redhat.com>
805 * diagnostic-core.h (warning_n, error_n, inform_n): Change
806 n argument to unsigned HOST_WIDE_INT.
807 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
808 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
809 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
810 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
812 2018-02-27 Richard Biener <rguenther@suse.de>
814 PR tree-optimization/84512
815 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
816 Do not use the estimate returned from record_stmt_cost for
817 the scalar iteration cost but sum properly using add_stmt_cost.
819 2018-02-27 Richard Biener <rguenther@suse.de>
821 PR tree-optimization/84466
822 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
823 Adjust last change to less strictly validate use operands.
825 2018-02-27 Martin Liska <mliska@suse.cz>
827 PR gcov-profile/84548
828 * gcov.c (process_file): Allow partial overlap and consider it
829 also as group functions.
830 (output_lines): Properly calculate range of lines for a group.
832 2018-02-27 Martin Liska <mliska@suse.cz>
834 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
835 'ggc' suffixes. Change first column width.
836 (timer::print): Fix formatting of the column.
838 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
840 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
841 preserve inline entry blocks for the sake of debug inline
842 entry point markers alone.
843 (remove_unused_locals): Suggest in comments a better place to
844 force the preservation of inline entry blocks that are
845 otherwise unused, but do not preserve them.
847 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
849 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
851 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
854 * config/i386/constraints.md (Bs): Replace
855 ix86_indirect_branch_register with
856 TARGET_INDIRECT_BRANCH_REGISTER.
858 * config/i386/i386.md (indirect_jump): Likewise.
859 (tablejump): Likewise.
860 (*sibcall_memory): Likewise.
861 (*sibcall_value_memory): Likewise.
862 Peepholes of indirect call and jump via memory: Likewise.
863 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
864 (*sibcall_value_GOT_32): Likewise.
865 * config/i386/predicates.md (indirect_branch_operand): Likewise.
866 (GOT_memory_operand): Likewise.
867 (call_insn_operand): Likewise.
868 (sibcall_insn_operand): Likewise.
869 (GOT32_symbol_operand): Likewise.
870 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
872 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
874 PR rtl-optimization/83496
875 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
876 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
877 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
878 redundant insn, if any.
879 (relax_delay_slots): Likewise.
880 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
882 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
884 PR tree-optimization/83965
885 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
886 that grouped statements are part of a reduction chain. Return
887 true if the statement is not marked as a reduction itself but
889 (vect_recog_dot_prod_pattern): Don't check whether the statement
890 is part of a group here.
891 (vect_recog_sad_pattern): Likewise.
892 (vect_recog_widen_sum_pattern): Likewise.
894 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
897 * final.c (rest_of_clean_state): Also look for calls inside sequences.
899 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
902 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
904 (ix86_output_indirect_function_return): New prototype.
905 (ix86_split_simple_return_pop_internal): Likewise.
906 * config/i386/i386.c (indirect_return_via_cx): New.
907 (indirect_return_via_cx_bnd): Likewise.
908 (indirect_thunk_name): Handle return va CX_REG.
909 (output_indirect_thunk_function): Create alias for
910 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
911 (ix86_output_indirect_jmp): Remove the bool argument.
912 (ix86_output_indirect_function_return): New function.
913 (ix86_split_simple_return_pop_internal): Likewise.
914 * config/i386/i386.md (*indirect_jump): Don't pass false
915 to ix86_output_indirect_jmp.
916 (*tablejump_1): Likewise.
917 (simple_return_pop_internal): Change it to define_insn_and_split.
918 Call ix86_split_simple_return_pop_internal to split it for
920 (simple_return_indirect_internal): Call
921 ix86_output_indirect_function_return instead of
922 ix86_output_indirect_jmp.
924 2018-02-26 Jakub Jelinek <jakub@redhat.com>
927 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
928 memset and value initialization afterwards.
930 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
932 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
934 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
937 * common/config/aarch64/aarch64-common.c
938 (aarch_option_optimization_table[]): Switch
939 off fomit-frame-pointer
941 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
942 Chung-Ju Wu <jasonwucj@gmail.com>
944 * config/nds32/nds32-multiple.md (load_multiple): Disallow
946 (store_multiple): Ditto.
948 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
950 * config.gcc: Add --with-cpu support for nds32 target.
951 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
952 * config/nds32/nds32.opt: Add -mcpu= option.
954 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
956 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
957 isel=yes): Warn for these deprecated options.
959 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
961 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
962 ISA_2_5_MASKS_EMBEDDED.
964 2018-02-23 Jakub Jelinek <jakub@redhat.com>
966 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
967 p->max as pointers rather than using iterative_hash_expr.
969 2018-02-23 Carl Love <cel@us.ibm.com>
971 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
972 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
974 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
975 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
976 P8V_BUILTIN_VEC_VUNSIGNED2.
978 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
981 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
982 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
984 * lra-constraints.c (curr_insn_transform): Set up
985 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
987 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
988 * lra-eliminations.c (spill_pseudos): Ditto.
989 (process_insn_for_elimination): Ditto.
990 * lra-lives.c (reg_early_clobber_p): Use the new macros.
991 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
992 LRA_NON_CLOBBERED_ALT.
994 2018-02-22 Martin Sebor <msebor@redhat.com>
996 PR tree-optimization/84480
997 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
998 to maybe_diag_stxncpy_trunc. Call it.
999 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1000 from gimple_fold_builtin_strcpy. Print inlining stack.
1001 (handle_builtin_stxncpy): Print inlining stack.
1002 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1004 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1007 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1008 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1009 and -fcheck-pointer-bounds are used together.
1010 (indirect_thunk_prefix): New enum.
1011 (indirect_thunk_need_prefix): New function.
1012 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1013 "_nt" instead of "_bnd" for NOTRACK prefix.
1014 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1015 (output_indirect_thunk_function): Likewise.
1017 (ix86_code_end): Update output_indirect_thunk_function calls.
1018 (ix86_output_indirect_branch_via_reg): Replace
1019 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1020 (ix86_output_indirect_branch_via_push): Likewise.
1021 (ix86_output_function_return): Likewise.
1022 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1023 incompatible with -fcf-protection=branch and
1024 -fcheck-pointer-bounds.
1026 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1029 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1030 Change gcc_assert call to output_operand_lossage.
1032 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1034 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1036 2018-02-22 DJ Delorie <dj@redhat.com>
1037 Sebastian Perta <sebastian.perta@renesas.com>
1038 Oleg Endo <olegendo@gcc.gnu.org>
1040 * config/rx/rx.c (rx_rtx_costs): New function.
1041 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1043 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1045 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1047 2018-02-22 Martin Liska <mliska@suse.cz>
1050 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1051 Add "native" as a possible value.
1053 2018-02-22 Martin Liska <mliska@suse.cz>
1056 * config/i386/i386.c (ix86_option_override_internal):
1057 Add "native" as a possible value for -march and -mtune.
1059 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1062 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1063 to all type variants.
1065 PR tree-optimization/84503
1066 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1067 width as info->bitpos + info->bitsize - start.
1068 (merged_store_group::merge_overlapping): Simplify width computation.
1069 (check_no_overlap): New function.
1070 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1071 start + width and last_order of the group, fail if check_no_overlap
1073 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1074 to group if check_no_overlap fails.
1076 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1078 * config/rs6000/altivec.md: Delete contraint arguments to
1079 define_expand, define_split, and define_peephole2, and in
1080 define_insn_and_split if always unused.
1081 * config/rs6000/darwin.md: Ditto.
1082 * config/rs6000/dfp.md: Ditto.
1083 * config/rs6000/rs6000.md: Ditto.
1084 * config/rs6000/sync.md: Ditto.
1085 * config/rs6000/vector.md: Ditto.
1086 * config/rs6000/vsx.md: Ditto.
1088 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1090 * config/rs6000/altivec.md: Write output control strings as braced
1091 blocks instead of double-quoted strings.
1092 * config/rs6000/darwin.md: Ditto.
1093 * config/rs6000/rs6000.md: Ditto.
1094 * config/rs6000/vector.md: Ditto.
1095 * config/rs6000/vsx.md: Ditto.
1097 2018-02-21 Jason Merrill <jason@redhat.com>
1099 PR c++/84314 - ICE with templates and fastcall attribute.
1100 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1102 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1104 * ipa-cp.c (determine_versionability): Fix comment typos.
1106 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1109 * ipa-cp.c (determine_versionability): Do not version functions caling
1112 2018-02-21 Martin Liska <mliska@suse.cz>
1115 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1116 Add "native" as a possible value.
1117 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1118 the macro when native cpu detection is available.
1120 2018-02-21 Martin Liska <mliska@suse.cz>
1123 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1124 Add "native" as a possible value.
1125 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1126 when native cpu detection is available.
1128 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1129 Martin Sebor <msebor@redhat.com>
1131 PR tree-optimization/84478
1132 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1134 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1135 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1136 support which is conservatively correct, for 2 only stay conservative
1137 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1138 argument to the 2 argument get_range_strlen, adjust 6 arg
1139 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1141 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1142 (gimple_fold_builtin_strlen): Pass true as last argument to
1145 2018-02-20 Martin Sebor <msebor@redhat.com>
1148 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1149 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1150 (builtin_memref::builtin_memref): Factor out parts into
1151 set_base_and_offset and call it.
1153 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1156 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1157 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1158 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1159 search at the associated MODE_INT.
1161 2018-02-20 Jeff Law <law@redhat.com>
1164 PR tree-optimization/81592
1166 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1167 for range data rather than using global data.
1168 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1169 range data rather than using global data.
1170 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1171 pass it to children as needed.
1172 (struct directive::fmtresult): Similarly.
1173 (struct directive::set_width): Similarly.
1174 (struct directive::set_precision): Similarly.
1175 (format_integer, format_directive, parse_directive): Similarly.
1176 (format_none): Accept unnamed vr_values parameter.
1177 (format_percent, format_floating, format_character): Similarly.
1178 (format_string, format_plain): Similarly.
1179 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1180 the EVRP range analyzer for range data rather than using global data.
1181 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1182 gimple-ssa-evrp-analyze.h
1183 (class sprintf_dom_walker): Add after_dom_children member function.
1184 Add evrp_range_analyzer member.
1185 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1186 range analyzer as needed.
1187 (sprintf_dom_walker::after_dom_children): New member function.
1188 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1190 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1191 (evrp_range_analyzer::pop_to_marker): Likewise.
1193 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1195 PR tree-optimization/84419
1196 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1197 with the required type if its current type is compatible but
1200 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1203 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1204 after vectorization.
1206 2018-02-20 Martin Liska <mliska@suse.cz>
1209 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1210 possible values if we don't have a hint.
1212 2018-02-20 Martin Liska <mliska@suse.cz>
1216 * final.c (shorten_branches): Build align_tab array with one
1218 * opts.c (finish_options): Add alignment option limit check.
1219 (MAX_CODE_ALIGN): Likewise.
1220 (MAX_CODE_ALIGN_VALUE): Likewise.
1221 * doc/invoke.texi: Document maximum allowed option value for
1222 all -falign-* options.
1224 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1227 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1228 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1229 * var-tracking.c (emit_note_insn_var_location): Remove all references
1230 to NOTE_INSN_CALL_ARG_LOCATION.
1231 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1232 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1233 Use copy_rtx_if_shared.
1234 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1235 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1236 (dwarf2out_var_location): Remove handling of
1237 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1239 * final.c (final_scan_insn): Remove all references to
1240 NOTE_INSN_CALL_ARG_LOCATION.
1241 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1242 before dumping final insns.
1243 * except.c (emit_note_eh_region_end): Remove all references to
1244 NOTE_INSN_CALL_ARG_LOCATION.
1245 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1246 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1247 * config/arc/arc.c (hwloop_optimize): Likewise.
1248 * config/arm/arm.c (create_fix_barrier): Likewise.
1249 * config/s390/s390.c (s390_chunkify_start): Likewise.
1250 * config/sh/sh.c (find_barrier): Likewise.
1251 * config/i386/i386.c (rest_of_insert_endbranch,
1252 ix86_seh_fixup_eh_fallthru): Likewise.
1253 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1254 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1255 * config/frv/frv.c (frv_function_prologue): Likewise.
1256 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1258 (note_outside_basic_block_p): Remove all references to
1259 NOTE_INSN_CALL_ARG_LOCATION.
1260 * gengtype.c (adjust_field_rtx_def): Likewise.
1261 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1263 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1264 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1267 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1270 PR tree-optimization/84452
1271 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1272 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1275 2018-02-19 Martin Liska <mliska@suse.cz>
1278 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1280 2018-02-19 Martin Liska <mliska@suse.cz>
1281 Richard Sandiford <richard.sandiford@linaro.org>
1283 PR tree-optimization/82491
1284 * gimple-fold.c (get_base_constructor): Make earlier bail out
1287 2018-02-19 Carl Love <cel@us.ibm.com>
1289 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1290 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1292 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1293 P8V_BUILTIN_VEC_NEG.
1295 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1297 * config/rl78/rl78.md (movdf): New define expand.
1299 2018-02-19 Martin Liska <mliska@suse.cz>
1302 * doc/invoke.texi: Fix typo.
1303 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1305 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1307 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1308 handle rs6000_single_float and rs6000_double_float specially for
1311 2018-02-16 Jeff Law <law@redhat.com>
1313 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1316 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1319 * ipa-inline.c (inline_small_functions): Fix a typo.
1321 2018-02-16 Nathan Sidwell <nathan@acm.org>
1323 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1325 2018-02-16 Carl Love <cel@us.ibm.com>
1327 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1328 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1329 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1330 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1331 expansion to P8V_BUILTIN_VEC_FLOAT2.
1333 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1335 PR rtl-optimization/70023
1336 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1337 src_regno into account.
1339 2018-02-16 Carl Love <cel@us.ibm.com>
1341 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1342 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1343 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1344 * config/rs6000/rs6000.c: Remove case statements for
1345 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1346 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1347 and P9V_BUILTIN_VEC_VINSERT4B.
1348 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1349 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1350 * config/rs6000/vsx.md:
1351 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1354 2018-02-16 Carl Love <cel@us.ibm.com>
1356 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1358 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1360 * config/rs6000/rs6000-c.c: Add the definitions for
1361 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1362 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1363 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1364 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1365 definition for insert4b and define insn *insert3b_internal.
1366 * doc/extend.texi: Add documentation for vec_extract4b.
1368 2018-02-16 Nathan Sidwell <nathan@acm.org>
1370 * doc/extend.texi (Backwards Compatibility): Mention friend
1371 injection. Note for-scope is deprecated.
1372 * doc/invoke.texi (-ffriend-injection): Deprecate.
1374 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1376 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1377 that moved to I2, also allow destinations that are a paradoxical
1378 subreg (instead of a normal reg).
1380 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1383 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1386 2018-02-16 Richard Biener <rguenther@suse.de>
1388 PR tree-optimization/84037
1389 PR tree-optimization/84016
1391 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1392 Adjust vec_construct for the fact we need additional higher latency
1393 128bit inserts for AVX256 and AVX512 vector builds.
1394 (ix86_add_stmt_cost): Scale vector construction cost for
1397 2018-02-16 Richard Biener <rguenther@suse.de>
1399 PR tree-optimization/84417
1400 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1401 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1402 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1404 2018-02-16 Martin Liska <mliska@suse.cz>
1407 * internal-fn.def (ASAN_CHECK): Set proper flags.
1408 (ASAN_MARK): Likewise.
1410 2018-02-16 Julia Koval <julia.koval@intel.com>
1412 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1413 from PTA_CANNONLAKE.
1415 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1418 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1419 Use ++iter rather than iter++ for std::list iterators.
1420 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1421 defer deleting them until all nodes in the forest are processed. Do
1422 free even leaf nodes. Change to_process into auto_vec.
1425 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1426 * vec.h (vec_default_construct): Use memset instead of placement new
1427 if BROKEN_VALUE_INITIALIZATION is defined.
1428 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1429 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1432 PR rtl-optimization/83723
1433 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1434 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1435 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1437 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1439 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1441 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1443 PR rtl-optimization/81443
1444 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1445 from inner REGs to paradoxical SUBREGs.
1447 2018-02-16 Richard Biener <rguenther@suse.de>
1449 PR tree-optimization/84399
1450 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1451 For operands we can analyze at their definition make sure we can
1452 analyze them at each use as well.
1454 2018-02-16 Richard Biener <rguenther@suse.de>
1456 PR tree-optimization/84190
1457 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1458 volatile accesses if the decl isn't volatile.
1460 2018-02-15 Jason Merrill <jason@redhat.com>
1462 PR c++/84314 - ICE with templates and fastcall attribute.
1463 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1464 TYPE_CANONICAL on an existing type.
1466 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1468 PR tree-optimization/84383
1469 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1470 dstoff nor call operand_equal_p if dstbase is NULL.
1472 PR tree-optimization/84334
1473 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1474 also a CONSTANT_CLASS_P, punt.
1476 2018-02-14 Jim Wilson <jimw@sifive.com>
1478 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1479 first SMALL_OPERAND check. New local min_second_step. Move assert
1480 to where locals are set. Add TARGET_RVC support.
1481 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1483 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1485 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1487 2018-02-14 Martin Sebor <msebor@redhat.com>
1489 PR tree-optimization/83698
1490 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1491 arrays constrain the offset range to their bounds.
1492 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1493 (builtin_access::overlap): Avoid setting the size of overlap if it's
1495 (maybe_diag_overlap): Also consider arrays when deciding what values
1496 of offsets to include in diagnostics.
1498 2018-02-14 Martin Sebor <msebor@redhat.com>
1501 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1502 that correspond to the kind of a declaration.
1504 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1507 * config/pa/pa.md: Load address of PIC label using the linkage table
1508 if the label is nonlocal.
1510 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1512 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1513 warning message if user requests -maltivec=be.
1514 * doc/invoke.texi: Document deprecation of -maltivec=be.
1516 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1519 * config/rs6000/rs6000-c.c: Update definitions for
1520 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1521 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1523 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1526 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1527 add _get_ssp intrinsics. Remove argument from
1528 __builtin_ia32_rdssp[d|q].
1529 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1530 * config/i386/i386-builtin.def: Remove argument from
1531 __builtin_ia32_rdssp[d|q].
1532 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1533 ix86_expand_special_args_builtin for _rdssp[d|q].
1534 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1535 Clear register before usage.
1536 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1537 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1539 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1541 PR tree-optimization/84357
1542 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1543 operand 1 of an ARRAY_REF too.
1545 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1548 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1549 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1551 (set_of_reg): New struct.
1552 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1553 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1554 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1556 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1557 Split into bitclr, bitset, bitinvert patterns if appropriate.
1558 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1559 use rx_fuse_in_memory_bitop.
1560 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1561 to named insn, correct maximum insn length.
1563 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1566 * machmode.def: Define a complex mode for PARTIAL_INT.
1567 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1569 * doc/rtl.texi: Document CSPImode.
1570 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1572 (msp430_hard_regno_nregs_with_padding): Likewise.
1574 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1577 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1579 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1581 PR rtl-optimization/84169
1582 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1583 we generated a parallel as new i3 and we split that to new i2 and i3
1584 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1585 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1586 those to i2, not i1. Partially rewrite this scan code.
1588 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1591 * stor-layout.c (place_field): For variable length fields, adjust
1592 offset_align afterwards not just based on the field's alignment,
1593 but also on the size.
1596 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1597 of exps and logs in the use_exp2 case.
1599 2018-02-13 Jeff Law <law@redhat.com>
1601 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1604 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1607 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1611 * common.opt (gas-loc-support, gas-locview-support): New.
1612 (ginline-points, ginternal-reset-location-views): New.
1613 * doc/invoke.texi: Document them. Use @itemx where intended.
1614 (gvariable-location-views): Adjust.
1615 * target.def (reset_location_view): New.
1616 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1617 (TARGET_RESET_LOCATION_VIEW): New.
1618 * doc/tm.texi: Rebuilt.
1619 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1620 (dwarf2out_default_as_locview_support): New.
1621 (output_asm_line_debug_info): Use option variables.
1622 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1623 (output_loc_list): Likewise.
1624 (add_high_low_attributes): Check option variables.
1625 Don't output entry view attribute in strict mode.
1626 (gen_inlined_subroutine_die): Check option variables.
1627 (dwarf2out_inline_entry): Likewise.
1628 (init_sections_and_labels): Likewise.
1629 (dwarf2out_early_finish): Likewise.
1630 (maybe_reset_location_view): New, from...
1631 (dwarf2out_var_location): ... here. Call it.
1632 * debug.h (dwarf2out_default_as_loc_support): Declare.
1633 (dwarf2out_default_as_locview_support): Declare.
1634 * hooks.c (hook_int_rtx_insn_0): New.
1635 * hooks.h (hook_int_rtx_insn_0): Declare.
1636 * toplev.c (process_options): Take -gas-loc-support and
1637 -gas-locview-support from dwarf2out. Enable
1638 -gvariable-location-views by default only with locview
1639 assembler support. Enable -ginternal-reset-location-views by
1640 default only if the target defines the corresponding hook.
1641 Enable -ginline-points by default if location views are
1642 enabled; force it disabled if statement frontiers are
1644 * tree-inline.c (expand_call_inline): Check option variables.
1645 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1647 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1649 PR tree-optimization/84321
1650 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1651 handling. Also check whether the anti-range contains any values
1652 that satisfy the mask; switch to a VR_RANGE if not.
1654 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1657 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1659 2018-02-13 Martin Jambor <mjambor@suse.cz>
1662 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1663 of call statements, also set location of a load to a temporary.
1665 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1667 * config/rl78/rl78.c (add_vector_labels): New function.
1668 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1669 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1670 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1671 which checks that no arguments are passed.
1672 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1673 * doc/extend.texi: Documentation for the new attribute.
1675 2018-02-13 Andreas Schwab <schwab@suse.de>
1677 * config/riscv/linux.h (CPP_SPEC): Define.
1679 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1682 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1683 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1684 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1685 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1686 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1687 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1688 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1689 temporarily for AES and PCLMUL builtins.
1691 PR tree-optimization/84339
1692 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1693 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1697 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1698 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1699 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1701 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1703 * omp-simd-clone.h: New file.
1704 * omp-simd-clone.c: Include omp-simd-clone.h.
1705 (expand_simd_clones): No longer static.
1706 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1707 cgraph.h and omp-simd-clone.h.
1708 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1709 (vect_recog_widen_shift_pattern): Formatting fix.
1710 (vect_pattern_recog_1): Don't check optab for calls.
1713 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1714 operands[2] into a REG before using gen_lowpart on it.
1716 2018-02-12 Jeff Law <law@redhat.com>
1719 * config/sh/sh.c (find_barrier): Consider a sibling call
1722 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1723 successfully back substituting a reg.
1725 2018-02-12 Richard Biener <rguenther@suse.de>
1727 PR tree-optimization/84037
1728 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1729 parameter, move visited init to caller.
1730 (vect_slp_analyze_operations): Separate cost from validity
1731 check, initialize visited once for all instances.
1732 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1734 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1736 (vect_model_promotion_demotion_cost): Likewise.
1737 (vectorizable_bswap): Guard cost modeling with !slp_node
1738 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1740 (vectorizable_call): Likewise.
1741 (vectorizable_conversion): Likewise.
1742 (vectorizable_assignment): Likewise.
1743 (vectorizable_shift): Likewise.
1744 (vectorizable_operation): Likewise.
1745 (vectorizable_store): Likewise.
1746 (vectorizable_load): Likewise.
1747 (vectorizable_condition): Likewise.
1748 (vectorizable_comparison): Likewise.
1750 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1753 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1754 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1755 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1757 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1760 * params.def (inline-min-speedup): Increase from 8 to 15.
1761 (max-inline-insns-auto): Decrease from 40 to 30.
1762 * ipa-split.c (consider_split): Add some buffer for function to
1763 be considered inlining candidate.
1764 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1767 2018-02-12 Richard Biener <rguenther@suse.de>
1769 PR tree-optimization/84037
1770 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1771 matched stmts if we cannot swap the non-matched ones.
1773 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1775 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1776 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1777 _mm_maskz_scalef_round_ss): New intrinsics.
1778 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1779 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1780 __builtin_ia32_scalefss_round): Remove.
1781 (__builtin_ia32_scalefsd_mask_round,
1782 __builtin_ia32_scalefss_mask_round): New intrinsics.
1783 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1784 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1785 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1786 "<round_constraint>")): Changed to ...
1787 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1788 "<round_scalar_constraint>")): ... this.
1789 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1790 %0, %1, %2<round_op3>}"): Changed to ...
1791 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1792 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1793 %2<round_scalar_mask_op3>}"): ... this.
1794 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1796 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1798 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1799 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1800 (_mm_maskz_sqrt_round_ss): New intrinsics.
1801 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1802 (__builtin_ia32_sqrtsd_mask_round)
1803 (__builtin_ia32_sqrtss_mask_round): New builtins.
1804 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1805 (__builtin_ia32_sqrtss_round): Remove.
1806 (__builtin_ia32_sqrtsd_mask_round)
1807 (__builtin_ia32_sqrtss_mask_round): New builtins.
1808 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1809 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1810 ((match_operand:VF_128 1 "vector_operand"
1811 "xBm,<round_constraint>")): Changed to ...
1812 ((match_operand:VF_128 1 "vector_operand"
1813 "xBm,<round_scalar_constraint>")): ... this.
1814 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1815 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1816 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1817 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1818 %<iptr>1<round_scalar_mask_op3>}): ... this.
1819 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1820 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1822 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1825 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1826 Cast vec_cmpeq result to correct type.
1827 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1828 Cast vec_cmpgt result to correct type.
1830 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1832 * final.c (final_scan_insn_1): Renamed from...
1833 (final_scan_insn): ... this. New wrapper, to recover
1834 seen from the outermost call in recursive ones.
1835 * config/sparc/sparc.c (output_return): Drop seen from call.
1836 (output_sibcall): Likewise.
1837 * config/visium/visium.c (output_branch): Likewise.
1839 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1841 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1844 2018-02-10 Alan Modra <amodra@gmail.com>
1847 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1848 Specify LR as an input.
1850 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1853 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1854 remove_member_access_dummy_vars): New functions.
1855 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1856 lower_omp_1, execute_lower_omp): Use them.
1858 PR rtl-optimization/84308
1859 * shrink-wrap.c (spread_components): Release todo vector.
1861 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1863 PR rtl-optimization/57193
1864 * ira-color.c (struct allocno_color_data): Add member
1865 conflict_allocno_hard_prefs.
1866 (update_conflict_allocno_hard_prefs): New.
1867 (bucket_allocno_compare_func): Add a preference based on
1868 conflict_allocno_hard_prefs.
1869 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1870 (color_allocnos): Remove a dead code. Initiate
1871 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1873 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1876 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1877 constraint from =wa to wa. Avoid a subreg on the output operand,
1878 instead use a pseudo and subreg it in a move.
1879 (p9_xxbrd_<mode>): Changed to ...
1880 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1881 (p9_xxbrd_v2df): New expander.
1882 (p9_xxbrw_<mode>): Changed to ...
1883 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1884 (p9_xxbrw_v4sf): New expander.
1886 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1888 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1890 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1893 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1894 multiply in 32-bit mode.
1895 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1896 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1899 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1901 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1902 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1903 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1904 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1906 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1909 * dwarf2out.c (is_trivial_indirect_ref): New function.
1910 (dwarf2out_late_global_decl): Do not generate a location
1911 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1912 and that are not defined in the current unit.
1914 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1916 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1917 instead of a libcall for UNORDERED.
1919 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1922 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1923 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1925 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1928 * config/s390/s390.c (s390_set_current_function): Invoke
1929 s390_indirect_branch_settings also if fndecl didn't change.
1931 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1933 * config/rs6000/rs6000.md (blockage): Set length to zero.
1935 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1937 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1939 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1942 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1943 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1947 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1948 PARALLEL incoming that failed vt_get_decl_and_offset check.
1951 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1952 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1954 (get_variable_section): For decls in named .bss* sections pass true as
1955 second argument to bss_initializer_p.
1957 2018-02-09 Marek Polacek <polacek@redhat.com>
1958 Jakub Jelinek <jakub@redhat.com>
1961 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1962 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1963 Sync some changes from cxx_fold_indirect_ref.
1965 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1967 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1969 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1970 (BLOCK_INLINE_ENTRY_LABEL): New.
1971 (dwarf2out_var_location): Disregard inline entry markers.
1972 (inline_entry_data): New struct.
1973 (inline_entry_data_hasher): New hashtable type.
1974 (inline_entry_data_hasher::hash): New.
1975 (inline_entry_data_hasher::equal): New.
1976 (inline_entry_data_table): New variable.
1977 (add_high_low_attributes): Add DW_AT_entry_pc and
1978 DW_AT_GNU_entry_view attributes if a pending entry is found
1979 in inline_entry_data_table. Add old entry_pc attribute only
1980 if debug nonbinding markers are disabled.
1981 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1982 markers are enabled.
1983 (block_within_block_p, dwarf2out_inline_entry): New.
1984 (dwarf2out_finish): Check that no entries remained in
1985 inline_entry_data_table.
1986 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1987 (final_scan_insn, notice_source_line): Likewise.
1988 (rest_of_clean_state): Skip inline entry markers.
1989 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1991 * gimple.c (gimple_build_debug_inline_entry): New.
1992 * gimple.h (enum gimple_debug_subcode): Add
1993 GIMPLE_DEBUG_INLINE_ENTRY.
1994 (gimple_build_debug_inline_entry): Declare.
1995 (gimple_debug_inline_entry_p): New.
1996 (gimple_debug_nonbind_marker_p): Adjust.
1997 * insn-notes.def (INLINE_ENTRY): New.
1998 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1999 inline entry marker notes.
2000 (print_insn): Likewise.
2001 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2002 (INSN_DEBUG_MARKER_KIND): Likewise.
2003 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2004 * tree-inline.c (expand_call_inline): Build and insert
2005 debug_inline_entry stmt.
2006 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2007 inline entry blocks early, if nonbind markers are enabled.
2008 (dump_scope_block): Dump fragment info.
2009 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2010 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2011 (gimple_build_debug_inline_entry): New.
2012 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2013 Enable/disable inline entry points too.
2014 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2015 (DEBUG_INSN): Describe inline entry markers.
2017 * common.opt (gvariable-location-views): New.
2018 (gvariable-location-views=incompat5): New.
2019 * config.in: Rebuilt.
2020 * configure: Rebuilt.
2021 * configure.ac: Test assembler for view support.
2022 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2023 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2024 * dwarf2out.c (var_loc_view): New typedef.
2025 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2026 (dwarf2out_locviews_in_attribute): New.
2027 (dwarf2out_locviews_in_loclist): New.
2028 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2029 (enum dw_line_info_opcode): Add LI_adv_address.
2030 (struct dw_line_info_table): Add view.
2031 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2032 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2033 (zero_view_p): New variable.
2034 (ZERO_VIEW_P): New macro.
2035 (output_asm_line_debug_info): New.
2036 (struct var_loc_node): Add view.
2037 (add_AT_view_list, AT_loc_list): New.
2038 (add_var_loc_to_decl): Add view param. Test it against last.
2039 (new_loc_list): Add view params. Record them.
2040 (AT_loc_list_ptr): Handle loc and view lists.
2041 (view_list_to_loc_list_val_node): New.
2042 (print_dw_val): Handle dw_val_class_view_list.
2043 (size_of_die): Likewise.
2044 (value_format): Likewise.
2045 (loc_list_has_views): New.
2046 (gen_llsym): Set vl_symbol too.
2047 (maybe_gen_llsym, skip_loc_list_entry): New.
2048 (dwarf2out_maybe_output_loclist_view_pair): New.
2049 (output_loc_list): Output view list or entries too.
2050 (output_view_list_offset): New.
2051 (output_die): Handle dw_val_class_view_list.
2052 (output_dwarf_version): New.
2053 (output_compilation_unit_header): Use it.
2054 (output_skeleton_debug_sections): Likewise.
2055 (output_rnglists, output_line_info): Likewise.
2056 (output_pubnames, output_aranges): Update version comments.
2057 (output_one_line_info_table): Output view numbers in asm comments.
2058 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2059 Call maybe_gen_llsym.
2060 (loc_list_from_tree_1): Adjust.
2061 (add_AT_location_description): Create view list attribute if
2062 needed, check it's absent otherwise.
2063 (convert_cfa_to_fb_loc_list): Adjust.
2064 (maybe_emit_file): Call output_asm_line_debug_info for test.
2065 (dwarf2out_var_location): Reset views as needed. Precompute
2066 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2067 attribute. Set view.
2068 (new_line_info_table): Reset next view.
2069 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2070 (dwarf2out_source_line): Likewise. Output view resets and labels to
2071 the assembler, or select appropriate line info opcodes.
2072 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2073 (optimize_string_length): Catch it. Adjust.
2074 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2075 dw_val_class_view_list, and remove it if no longer needed.
2076 (hash_loc_list): Hash view numbers.
2077 (loc_list_hasher::equal): Compare them.
2078 (optimize_location_lists): Check whether a view list symbol is
2079 needed, and whether the locview attribute is present, and
2080 whether they match. Remove the locview attribute if no longer
2082 (index_location_lists): Call skip_loc_list_entry for test.
2083 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2084 Use output_dwarf_version.
2085 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2086 (struct dw_val_node): Add val_view_list.
2087 * final.c (SEEN_NEXT_VIEW): New.
2088 (set_next_view_needed): New.
2089 (clear_next_view_needed): New.
2090 (maybe_output_next_view): New.
2091 (final_start_function): Rename to...
2092 (final_start_function_1): ... this. Take pointer to FIRST,
2093 add SEEN parameter. Emit param bindings in the initial view.
2094 (final_start_function): Reintroduce SEEN-less interface.
2095 (final): Rename to...
2096 (final_1): ... this. Take SEEN parameter. Output final pending
2097 next view at the end.
2098 (final): Reintroduce seen-less interface.
2099 (final_scan_insn): Output pending next view before switching
2100 sections or ending a block. Mark the next view as needed when
2101 outputting variable locations. Notify debug backend of section
2102 changes, and of location view changes.
2103 (rest_of_handle_final): Adjust.
2104 * toplev.c (process_options): Autodetect value for debug variable
2105 location views option. Warn on incompat5 without -gdwarf-5.
2106 * doc/invoke.texi (gvariable-location-views): New.
2107 (gvariable-location-views=incompat5): New.
2108 (gno-variable-location-views): New.
2110 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2112 PR tree-optimization/84136
2113 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2114 that the result of find_edge is non-NULL.
2116 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2119 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2120 storing integer register in SImode. Fix cost of 256 and 512
2121 byte aligned SSE register store.
2123 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2125 * config/i386/i386.c (ix86_multiplication_cost): Fix
2126 multiplication cost for TARGET_AVX512DQ.
2128 2018-02-08 Marek Polacek <polacek@redhat.com>
2130 PR tree-optimization/84238
2131 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2134 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2136 PR tree-optimization/84265
2137 * tree-vect-stmts.c (vectorizable_store): Don't treat
2138 VMAT_CONTIGUOUS accesses as grouped.
2139 (vectorizable_load): Likewise.
2141 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2143 PR tree-optimization/81635
2144 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2145 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2146 (test_round_for_mask): New functions.
2147 (wide_int_cc_tests): Call test_round_for_mask.
2148 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2149 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2150 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2151 range returned by get_range_info.
2153 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2156 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2157 * symtab.c: Include builtins.h
2158 (symtab_node::output_to_lto_symbol_table_p): Move here
2159 from lto-streamer-out.c:output_symbol_p.
2160 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2161 (output_symbol_p): Move all logic to symtab.c
2162 (produce_symtab): Update.
2164 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2166 * config/s390/s390-opts.h (enum indirect_branch): Define.
2167 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2168 (s390_indirect_branch_via_thunk)
2169 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2170 (enum s390_indirect_branch_type): Define.
2171 * config/s390/s390.c (struct s390_frame_layout, struct
2172 machine_function): Remove.
2173 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2174 (indirect_branch_table_label_no, indirect_branch_table_name):
2176 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2177 (enum s390_indirect_branch_option): Define.
2178 (s390_return_addr_from_memory): New function.
2179 (s390_handle_string_attribute): New function.
2180 (s390_attribute_table): Add new attribute handler.
2181 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2182 (s390_indirect_branch_via_thunk): New function.
2183 (s390_indirect_branch_via_inline_thunk): New function.
2184 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2185 sibling call optimization for non z10 compiles.
2186 (s390_emit_call): Force indirect branch target to be a single
2187 register. Add r1 clobber for non-z10 compiles.
2188 (s390_emit_epilogue): Emit return jump via return_use expander.
2189 (s390_reorg): Handle JUMP_INSNs as execute targets.
2190 (s390_option_override_internal): Perform validity checks for the
2191 new command line options.
2192 (s390_indirect_branch_attrvalue): New function.
2193 (s390_indirect_branch_settings): New function.
2194 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2195 (s390_output_indirect_thunk_function): New function.
2196 (s390_code_end): Implement target hook.
2197 (s390_case_values_threshold): Implement target hook.
2198 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2200 * config/s390/s390.h (struct s390_frame_layout)
2201 (struct machine_function): Move here from s390.c.
2202 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2203 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2204 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2205 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2206 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2207 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2208 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2209 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2210 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2211 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2212 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2213 (mnemonic attribute): Add values which aren't recognized
2215 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2216 pattern for branch conversion. Fix mnemonic attribute.
2217 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2218 indirect branch via thunk if requested.
2219 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2220 ("*indirect_jump"): Disable for branch conversion using out of
2222 ("indirect_jump_via_thunk<mode>_z10")
2223 ("indirect_jump_via_thunk<mode>")
2224 ("indirect_jump_via_inlinethunk<mode>_z10")
2225 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2226 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2227 ("casesi_jump_via_inlinethunk<mode>_z10")
2228 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2229 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2230 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2231 ("*indirect2_jump"): Disable for branch conversion.
2232 ("casesi_jump"): Turn into expander and expand patterns for branch
2234 ("return_use"): New expander.
2235 ("*return"): Emit return via thunk and rename it to ...
2236 ("*return<mode>"): ... this one.
2237 * config/s390/s390.opt: Add new options and and enum for the
2240 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2242 * lra-constraints.c (match_reload): Unconditionally use
2243 gen_lowpart_SUBREG, rather than selecting between that
2244 and equivalent gen_rtx_SUBREG code.
2246 2018-02-08 Richard Biener <rguenther@suse.de>
2248 PR tree-optimization/84233
2249 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2250 changed flag instead of boguously re-using phi_inserted.
2252 2018-02-08 Martin Jambor <mjambor@suse.cz>
2254 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2255 static local variables.
2257 2018-02-08 Richard Biener <rguenther@suse.de>
2259 PR tree-optimization/84278
2260 * tree-vect-stmts.c (vectorizable_store): When looking for
2261 smaller vector types to perform grouped strided loads/stores
2262 make sure the mode is supported by the target.
2263 (vectorizable_load): Likewise.
2265 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2267 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2268 Increase LDP/STP opportunities by adding adjacent callee-saves.
2270 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2272 PR rtl-optimization/84068
2273 PR rtl-optimization/83459
2274 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2276 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2278 PR tree-optimization/84224
2279 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2280 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2283 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2286 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2287 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2288 count, remove one USE.
2290 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2292 * doc/install.texi (Configuration): Document the
2293 --with-long-double-format={ibm,ieee} PowerPC configuration
2297 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2298 Convert from define_expand to be define_insn_and_split. Rework
2299 float/double/_Float128 conversions to QI/HI/SImode to work with
2300 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2301 conversions to QI/HImode types did a store and then a load to
2302 truncate the value. For conversions to VSX registers, don't split
2303 the insn, instead emit the code directly. Use the code iterator
2304 any_fix to combine signed and unsigned conversions.
2305 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2306 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2307 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2308 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2309 (fix_<mode>di2_hw): Likewise.
2310 (fixuns_<mode>di2_hw): Likewise.
2311 (fix_<mode>si2_hw): Likewise.
2312 (fixuns_<mode>si2_hw): Likewise.
2313 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2314 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2315 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2316 fix<uns>_trunc<SFDF:mode>si2_p8.
2317 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2319 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2320 (fix<uns>_<mode>_mem): Likewise.
2321 (fctiw<u>z_<mode>_mem): Likewise.
2322 (fix<uns>_<mode>_mem): Likewise.
2323 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2324 the register allocator from doing a direct move to the GPRs to do
2325 a store, and instead use the ISA 3.0 store byte/half-word from
2326 vector register instruction. For IEEE 128-bit floating point,
2327 also optimize stores of 32-bit ints.
2328 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2330 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2332 * genextract.c (push_pathstr_operand): New function to support
2334 (walk_rtx): Call push_pathstr_operand.
2335 (print_path): Support [a-zA-Z].
2337 2018-02-07 Richard Biener <rguenther@suse.de>
2339 PR tree-optimization/84037
2340 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2341 (cse_and_gimplify_to_preheader): Declare.
2342 (vect_get_place_in_interleaving_chain): Likewise.
2343 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2345 (_loop_vec_info::~_loop_vec_info): Delete it.
2346 (cse_and_gimplify_to_preheader): New function.
2347 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2348 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2349 (vectorizable_load): Likewise. For grouped stores always base
2350 the IV on the first element.
2351 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2352 condition before gimplifying.
2354 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2356 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2357 *DIV_EXPR and *MOD_EXPR.
2359 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2362 * config/i386/i386.c (ix86_option_override_internal): Mask out
2363 the CF_SET bit when checking -fcf-protection.
2365 2018-02-07 Tom de Vries <tom@codesourcery.com>
2368 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2371 2018-02-07 Richard Biener <rguenther@suse.de>
2373 PR tree-optimization/84204
2374 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2377 PR tree-optimization/84205
2378 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2379 special-case isl_ast_op_zdiv_r.
2381 PR tree-optimization/84223
2382 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2383 Only add conditions from within the region.
2384 (gather_bbs::after_dom_children): Adjust.
2386 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2389 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2390 * config/avr/avr.md: Only post-reload split REG-REG moves if
2391 either register is GENERAL_REG_P.
2393 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2395 PR tree-optimization/84235
2396 * tree-ssa-scopedtables.c
2397 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2398 if the subtraction is performed in floating point type where NaNs are
2399 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2400 build 1. Formatting fix.
2402 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2405 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2406 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2407 and skip it regardless of bb boundaries. Use CALL_P macro,
2408 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2409 unnecessarily, formatting fix.
2411 2018-02-06 Michael Collison <michael.collison@arm.com>
2413 * config/arm/thumb2.md:
2414 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2415 (*thumb_mov_notscc): Ditto.
2417 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2420 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2421 unsigned_fix, not "s".
2423 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2425 * configure.ac (gcc_fn_eh_frame_ro): New function.
2426 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2427 correct .eh_frame permissions.
2428 * configure: Regenerate.
2430 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2432 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2435 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2437 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2438 Display warning message for -mno-speculate-indirect-jumps.
2440 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2442 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2444 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2446 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2448 PR tree-optimization/84225
2449 * tree-eh.c (find_trapping_overflow): Only call
2450 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2452 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2455 * config/i386/i386.c: Reimplement the check of possible options
2456 -mibt/-mshstk conbination. Change error messages.
2457 * doc/invoke.texi: Fix a typo: remove extra '='.
2459 2018-02-06 Marek Polacek <polacek@redhat.com>
2461 PR tree-optimization/84228
2462 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2464 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2467 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2468 emitted arch directives.
2469 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2470 __ARM_FEATURE_COPROC before changing architectures.
2472 2018-02-06 Richard Biener <rguenther@suse.de>
2474 * config/i386/i386.c (print_reg): Fix typo.
2475 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2477 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2479 * configure: Regenerate.
2481 2018-02-05 Martin Sebor <msebor@redhat.com>
2483 PR tree-optimization/83369
2484 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2487 2018-02-05 Martin Liska <mliska@suse.cz>
2489 * doc/invoke.texi: Cherry-pick upstream r323995.
2491 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2493 * ira.c (ira_init_register_move_cost): Adjust comment.
2495 2018-02-05 Martin Liska <mliska@suse.cz>
2497 PR gcov-profile/84137
2498 * doc/gcov.texi: Fix typo in documentation.
2500 2018-02-05 Martin Liska <mliska@suse.cz>
2502 PR gcov-profile/83879
2503 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2504 using dlopen functionality.
2506 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2508 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2509 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2510 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2511 _mm_maskz_range_round_ss): New intrinsics.
2512 (__builtin_ia32_rangesd128_round)
2513 (__builtin_ia32_rangess128_round): Remove.
2514 (__builtin_ia32_rangesd128_mask_round,
2515 __builtin_ia32_rangess128_mask_round): New builtins.
2516 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2517 __builtin_ia32_rangess128_round): Remove.
2518 (__builtin_ia32_rangesd128_mask_round,
2519 __builtin_ia32_rangess128_mask_round): New builtins.
2520 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2521 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2522 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2523 "<round_saeonly_constraint>")): Changed to ...
2524 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2525 "<round_saeonly_scalar_constraint>")): ... this.
2526 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2527 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2528 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2529 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2530 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2532 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2534 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2536 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2537 Remove all values except native, 8540 and 8548.
2539 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2541 * config/i386/i386.c (ix86_output_function_return): Pass
2542 INVALID_REGNUM, instead of -1, as invalid register number to
2543 indirect_thunk_name and output_indirect_thunk.
2545 2018-02-02 Julia Koval <julia.koval@intel.com>
2547 * config.gcc: Add -march=icelake.
2548 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2549 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2550 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2551 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2552 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2553 (processor_target_table): Add icelake.
2554 (ix86_option_override_internal): Handle new PTAs.
2555 (get_builtin_code_for_version): Handle icelake.
2556 (M_INTEL_COREI7_ICELAKE): New.
2557 (fold_builtin_cpu): Handle icelake.
2558 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2559 * doc/invoke.texi: Add -march=icelake.
2561 2018-02-02 Julia Koval <julia.koval@intel.com>
2563 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2564 to wide_int_bitmask.
2565 * wide-int-bitmask.h: New.
2567 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2570 * config/i386/i386.md: Replace Pmode with word_mode in
2571 builtin_setjmp_setup and builtin_longjmp to support x32.
2573 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2577 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2579 (rs6000_supported_cpu_names): New static variable.
2580 (linux_cpu_translation_table): Likewise.
2581 (elf_platform) <cpu>: Define new static variable and use it.
2582 Translate kernel AT_PLATFORM name to canonical name if needed.
2583 Error if platform name is unknown.
2585 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2588 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2590 2018-02-01 Jeff Law <law@redhat.com>
2593 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2594 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2595 the scratch if RELEASE_VIA_POP is false.
2596 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2597 If we have to save a temporary register, decrement SIZE appropriately.
2598 Pass new arguments to release_scratch_register_on_entry.
2599 (ix86_adjust_stack_and_probe): Likewise.
2600 (ix86_emit_probe_stack_range): Pass new arguments to
2601 release_scratch_register_on_entry.
2603 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2605 PR rtl-optimization/84157
2606 * combine.c (change_zero_ext): Use REG_P predicate in
2607 front of HARD_REGISTER_P predicate.
2609 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2611 * config/avr/avr.c (avr_option_override): Move disabling of
2612 -fdelete-null-pointer-checks to...
2613 * common/config/avr/avr-common.c (avr_option_optimization_table):
2616 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2618 PR tree-optimization/81635
2619 * tree-data-ref.c (split_constant_offset_1): For types that
2620 wrap on overflow, try to use range info to prove that wrapping
2623 2018-02-01 Renlin Li <renlin.li@arm.com>
2626 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2628 (aarch64_register_move_cost): Likewise.
2629 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2631 (REG_CLASS_NAMES): Likewise.
2632 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2633 TAILCALL_ADDR_REGS. Remove IP registers.
2634 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2636 2018-02-01 Richard Biener <rguenther@suse.de>
2638 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2639 for specifying RPO order and allow NULL for that.
2640 * domwalk.c (dom_walker::dom_walker): Likewise.
2641 (dom_walker::walk): Handle NULL RPO order.
2642 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2644 (rewrite_update_dom_walker): Likewise.
2645 (mark_def_dom_walker): Likewise.
2647 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2649 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2650 (aarch64_maybe_expand_sve_subreg_move): Declare.
2651 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2652 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2654 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2655 that are semantically a reverse operation.
2656 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2657 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2658 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2660 (aarch64_can_change_mode_class): For big-endian, forbid changes
2661 between two SVE modes if they have different element sizes.
2663 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2665 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2666 the TImode handling for big-endian targets.
2668 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2670 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2671 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2673 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2674 Remove BSWAP handing for big-endian targets and use the form of
2675 LD1RQ appropariate for the mode.
2677 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2679 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2680 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2683 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2686 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2687 check for operands that need to go through aarch64_sve_reload_be.
2689 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2691 PR tree-optimization/81661
2692 PR tree-optimization/84117
2693 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2694 * tree-eh.c: Include gimplify.h.
2695 (find_trapping_overflow, replace_trapping_overflow,
2696 rewrite_to_non_trapping_overflow): New functions.
2697 * tree-vect-loop.c: Include tree-eh.h.
2698 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2699 * tree-data-ref.c: Include tree-eh.h.
2700 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2702 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2704 PR rtl-optimization/84123
2705 * combine.c (change_zero_ext): Check if hard register satisfies
2706 can_change_dest_mode before calling gen_lowpart_SUBREG.
2708 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2711 * ira.c (ira_init_register_move_cost): Remove assert.
2713 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2715 PR rtl-optimization/84071
2716 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2717 * doc/tm.texi: Regenerate.
2719 2018-01-31 Richard Biener <rguenther@suse.de>
2721 PR tree-optimization/84132
2722 * tree-data-ref.c (analyze_miv_subscript): Properly
2723 check whether evolution_function_is_affine_multivariate_p
2724 before calling gcd_of_steps_may_divide_p.
2726 2018-01-31 Julia Koval <julia.koval@intel.com>
2729 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2730 * config/i386/i386.md (rdpid_rex64) New.
2731 (rdpid): Make 32bit only.
2733 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2736 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2737 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2739 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2742 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2744 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2746 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2748 PR rtl-optimization/84071
2749 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2750 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2752 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2754 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2755 (arc_attribute_table): Add 'aux' attribute.
2756 (arc_in_small_data_p): Consider aux like variables.
2757 (arc_is_aux_reg_p): New function.
2758 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2759 (arc_get_aux_arg): New function.
2760 (prepare_move_operands): Handle aux-register access.
2761 (arc_handle_aux_attribute): New function.
2762 * doc/extend.texi (ARC Variable attributes): Add subsection.
2764 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2766 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2767 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2768 (arc_attribute_table): Add 'uncached' attribute.
2769 (arc_print_operand): Print '.di' flag for uncached memory
2771 (arc_in_small_data_p): Do not consider for small data the uncached
2773 (arc_is_uncached_mem_p): New function.
2774 * config/arc/predicates.md (compact_store_memory_operand): Check
2775 for uncached memory accesses.
2776 (nonvol_nonimm_operand): Likewise.
2777 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2779 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2782 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2783 falign-loops=): Add Optimization flag.
2785 2018-01-30 Jeff Law <law@redhat.com>
2788 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2789 INT_REGISTERS_SAVED. Check it prior to calling
2790 get_scratch_register_on_entry.
2791 (ix86_adjust_stack_and_probe): Similarly.
2792 (ix86_emit_probe_stack_range): Similarly.
2793 (ix86_expand_prologue): Corresponding changes.
2795 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2798 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2799 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2801 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2804 * lra-constraints.c (curr_insn_transform): Process AND in the
2807 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2809 PR rtl-optimization/83986
2810 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2811 dependence against last_pending_memory_flush in addition to
2814 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2816 PR tree-optimization/81611
2817 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2820 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2823 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2826 2018-01-30 Richard Biener <rguenther@suse.de>
2827 Jakub Jelinek <jakub@redhat.com>
2829 PR tree-optimization/84111
2830 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2831 inner loops added during recursion, as they don't have up-to-date
2834 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2837 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2838 (can_inline_edge_by_limits_p): ... here.
2839 (can_early_inline_edge_p, check_callers,
2840 update_caller_keys, update_callee_keys, recursive_inlining,
2841 add_new_edges_to_heap, speculation_useful_p,
2842 inline_small_functions,
2843 inline_small_functions, flatten_function,
2844 inline_to_all_callers_1): Update.
2846 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2848 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2851 2018-01-30 Richard Biener <rguenther@suse.de>
2853 PR tree-optimization/83008
2854 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2855 invariant and constant vector uses in stmts when they need
2858 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2861 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2862 * configure: Regenerate.
2864 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2866 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2868 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2869 Use gen_rtx_REG rather than gen_lowpart.
2871 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2873 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2874 rather than 0 when creating partial subregs.
2876 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2878 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2881 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2884 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2885 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2886 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2887 flags. This restores the settings used before the 2017-07-24.
2888 Turning off pre increment/decrement/modify allows IVOPTS to
2889 optimize DF/SF loops where the index is an int.
2891 2018-01-29 Richard Biener <rguenther@suse.de>
2892 Kelvin Nilsen <kelvin@gcc.gnu.org>
2895 * tree-vect-stmts.c (vectorizable_call): Don't call
2896 targetm.vectorize_builtin_md_vectorized_function if callee is
2899 2018-01-22 Carl Love <cel@us.ibm.com>
2901 * doc/extend.tex: Fix typo in second arg in
2902 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2904 2018-01-29 Richard Biener <rguenther@suse.de>
2906 PR tree-optimization/84086
2907 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2908 (flush_ssaname_freelist): When SSA names were released reset
2909 the SCEV hash table.
2911 2018-01-29 Richard Biener <rguenther@suse.de>
2913 PR tree-optimization/84057
2914 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2915 removed paths when removing edges.
2917 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2919 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2920 -mfunction-return=@var{choice}.
2922 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2925 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2927 (layout::print_source_line): Likewise.
2928 (test_get_line_width_without_trailing_whitespace): Add test cases.
2930 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2933 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2936 2018-01-26 Jim Wilson <jimw@sifive.com>
2938 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2940 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2943 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2945 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2946 and CMP + SUB-immediate -> SUBS.
2948 2018-01-26 Martin Sebor <msebor@redhat.com>
2950 PR tree-optimization/83896
2951 * tree-ssa-strlen.c (get_string_len): Rename...
2952 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2953 Avoid assuming length is constant.
2954 (handle_char_store): Use HOST_WIDE_INT for string length.
2956 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2959 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2960 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2962 2018-01-26 Richard Biener <rguenther@suse.de>
2964 PR rtl-optimization/84003
2965 * dse.c (record_store): Only record redundant stores when
2966 the earlier store aliases at least all accesses the later one does.
2968 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2970 PR rtl-optimization/83985
2971 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2972 REG_CFA_RESTORE insns.
2973 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2974 REG_CFA_RESTORE insns here.
2977 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2978 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2980 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2982 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2983 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2984 (arc_init): Likewise.
2985 (arc_override_options): Likewise.
2986 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2988 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2990 * config/arc/arc.h (TARGET_DBNZ): Define.
2991 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2992 properly set the tune attribute.
2993 (dbnz): Use TARGET_DBNZ guard.
2994 * config/arc/arc.opt (mtune): Add core3 option.
2996 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2998 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2999 recognize new pic like addresses.
3000 (arc_delegitimize_address): Clean up.
3002 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3004 * config/arc/arc-arches.def: Option mrf16 valid for all
3006 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3007 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3008 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3009 * config/arc/arc-tables.opt: Regenerate.
3010 * config/arc/arc.c (arc_conditional_register_usage): Handle
3011 reduced register file case.
3012 (arc_file_start): Set must have build attributes.
3013 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3015 * config/arc/arc.opt (mrf16): Add new option.
3016 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3017 * config/arc/genmultilib.awk: Handle new mrf16 option.
3018 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3019 * config/arc/t-multilib: Regenerate.
3020 * doc/invoke.texi (ARC Options): Document mrf16 option.
3022 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3024 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3025 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3026 (arc_attribute_table): Add 'secure_call' attribute.
3027 (arc_print_operand): Print secure call operand.
3028 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3030 (arc_is_secure_call_p): New function. * config/arc/arc.md
3031 (call_i): Add support for sjli instruction.
3032 (call_value_i): Likewise.
3033 * config/arc/constraints.md (Csc): New constraint.
3035 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3036 John Eric Martin <John.Martin@emmicro-us.com>
3038 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3039 * config/arc/arc.c (_arc_jli_section): New struct.
3040 (arc_jli_section): New type.
3041 (rc_jli_sections): New static variable.
3042 (arc_handle_jli_attribute): New function.
3043 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3044 (arc_file_end): New function.
3045 (TARGET_ASM_FILE_END): Define.
3046 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3047 (arc_add_jli_section): New function.
3048 (jli_call_scan): Likewise.
3049 (arc_reorg): Call jli_call_scan.
3050 (arc_output_addsi): Remove 'S' from printing asm operand.
3051 (arc_is_jli_call_p): New function.
3052 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3054 (movhi_insn): Likewise.
3055 (movsi_insn): Likewise.
3056 (movsi_set_cc_insn): Likewise.
3057 (loadqi_update): Likewise.
3058 (load_zeroextendqisi_update): Likewise.
3059 (load_signextendqisi_update): Likewise.
3060 (loadhi_update): Likewise.
3061 (load_zeroextendhisi_update): Likewise.
3062 (load_signextendhisi_update): Likewise.
3063 (loadsi_update): Likewise.
3064 (loadsf_update): Likewise.
3065 (movsicc_insn): Likewise.
3066 (bset_insn): Likewise.
3067 (bxor_insn): Likewise.
3068 (bclr_insn): Likewise.
3069 (bmsk_insn): Likewise.
3070 (bicsi3_insn): Likewise.
3071 (cmpsi_cc_c_insn): Likewise.
3072 (movsi_ne): Likewise.
3073 (movsi_cond_exec): Likewise.
3074 (clrsbsi2): Likewise.
3085 (call_i): Remove 'S' asm letter, add jli instruction.
3086 (call_value_i): Likewise.
3087 * config/arc/arc.op (mjli-always): New option.
3088 * config/arc/constraints.md (Cji): New constraint.
3089 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3091 (subsf3_fpx): Likewise.
3092 (mulsf3_fpx): Likewise.
3093 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3095 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3097 * doc/invoke.texi (ARC): Document mjli-always option.
3099 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3101 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3102 avoid addition with 0 and use incw and decw where possible.
3104 2018-01-26 Richard Biener <rguenther@suse.de>
3106 PR tree-optimization/81082
3107 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3108 association if it requires casting to unsigned.
3109 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3110 from fold_plusminus_mult_expr to catch important cases late when
3111 range info is available.
3113 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3115 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3116 * configure.ac (hidden_linkonce): New test.
3117 * configure: Regenerate.
3118 * config.in: Regenerate.
3120 2018-01-26 Julia Koval <julia.koval@intel.com>
3122 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3123 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3124 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3125 _mm_mask_bitshuffle_epi64_mask): Fix type.
3126 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3127 USI_FTYPE_V4DI_V4DI_USI): Remove.
3128 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3129 __builtin_ia32_vpshufbitqmb256_mask,
3130 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3131 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3132 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3134 2018-01-26 Alan Modra <amodra@gmail.com>
3137 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3138 UNSPEC_VBPERMQ. Sort other unspecs.
3140 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3142 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3144 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3147 * predict.c (drop_profile): Do not push/pop cfun; update also
3149 (handle_missing_profiles): Fix logic looking for zero profiles.
3151 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3154 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3155 on functions with #pragma omp declare simd or functions with simd
3157 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3158 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3159 Remove trailing \n from warning_at calls.
3161 2018-01-25 Tom de Vries <tom@codesourcery.com>
3164 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3165 for neutered workers.
3167 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3170 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3171 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3173 2018-01-24 Jeff Law <law@redhat.com>
3176 * i386.c (get_probe_interval): Move to earlier point.
3177 (ix86_compute_frame_layout): If -fstack-clash-protection and
3178 the frame is larger than the probe interval, then use pushes
3179 to save registers rather than reg->mem moves.
3180 (ix86_expand_prologue): Remove conditional for int_registers_saved
3183 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3186 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3187 min/max for never referenced object.
3189 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3192 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3194 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3195 attributes from DECL_ATTRIBUTES (decl) without affecting
3196 DECL_ATTRIBUTES (current_function_decl).
3197 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3198 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3200 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3202 PR tree-optimization/83979
3203 * fold-const.c (fold_comparison): Use constant_boolean_node
3204 instead of boolean_{true,false}_node.
3206 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3208 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3211 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3213 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3214 Simplify the clause that sets the length attribute.
3215 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3216 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3217 clause that sets the length attribute.
3218 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3220 2018-01-24 Tom de Vries <tom@codesourcery.com>
3223 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3224 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3225 Add strict parameter.
3226 (prevent_branch_around_nothing): Insert dummy insn between branch to
3227 label and label with no ptx insn inbetween.
3228 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3230 2018-01-24 Tom de Vries <tom@codesourcery.com>
3233 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3234 for neutered threads in warp.
3235 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3237 2018-01-24 Richard Biener <rguenther@suse.de>
3239 PR tree-optimization/83176
3240 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3243 2018-01-24 Richard Biener <rguenther@suse.de>
3245 PR tree-optimization/82819
3246 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3247 code generating pluses that are no-ops in the target precision.
3249 2018-01-24 Richard Biener <rguenther@suse.de>
3252 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3254 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3256 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3257 to merge probabilities.
3258 * predict.c (probably_never_executed): Also mark as cold functions
3259 with global 0 profile and guessed local profile.
3260 * profile-count.c (profile_probability::combine_with_count): New
3262 * profile-count.h (profile_probability::operator*,
3263 profile_probability::operator*=, profile_probability::operator/,
3264 profile_probability::operator/=): Reduce precision to adjusted
3265 and set value to guessed on contradictory divisions.
3266 (profile_probability::combine_with_freq): Remove.
3267 (profile_probability::combine_wiht_count): Declare.
3268 (profile_count::force_nonzero):: Set to adjusted.
3269 (profile_count::probability_in):: Set quality to adjusted.
3270 * tree-ssa-tail-merge.c (replace_block_by): Use
3273 2018-01-23 Andrew Waterman <andrew@sifive.com>
3274 Jim Wilson <jimw@sifive.com>
3276 * config/riscv/riscv.c (riscv_stack_boundary): New.
3277 (riscv_option_override): Set riscv_stack_boundary. Handle
3278 riscv_preferred_stack_boundary_arg.
3279 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3280 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3281 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3282 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3283 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3284 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3286 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3289 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3290 of struct ix86_frame.
3291 (ix86_expand_epilogue): Likewise. Add a local variable for
3292 the reg_save_offset field in struct ix86_frame.
3294 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3296 PR tree-optimization/82604
3297 * tree-loop-distribution.c (enum partition_kind): New enum item
3298 PKIND_PARTIAL_MEMSET.
3299 (partition_builtin_p): Support above new enum item.
3300 (generate_code_for_partition): Ditto.
3301 (compute_access_range): Differentiate cases that equality can be
3302 proven at all loops, the innermost loops or no loops.
3303 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3304 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3305 (finalize_partitions, distribute_loop): Don't fuse partition of
3306 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3307 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3310 2018-01-23 Martin Liska <mliska@suse.cz>
3312 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3313 order to ignore the predictor.
3314 (PRED_POLYMORPHIC_CALL): Likewise.
3315 (PRED_RECURSIVE_CALL): Likewise.
3317 2018-01-23 Martin Liska <mliska@suse.cz>
3319 * tree-profile.c (tree_profiling): Print function header to
3320 aware reader which function we are working on.
3321 * value-prof.c (gimple_find_values_to_profile): Do not print
3322 not interesting value histograms.
3324 2018-01-23 Martin Liska <mliska@suse.cz>
3326 * profile-count.h (enum profile_quality): Add
3327 profile_uninitialized as the first value. Do not number values
3328 as they are zero based.
3329 (profile_count::verify): Update sanity check.
3330 (profile_probability::verify): Likewise.
3332 2018-01-23 Nathan Sidwell <nathan@acm.org>
3334 * doc/invoke.texi (ffor-scope): Deprecate.
3336 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3338 PR tree-optimization/83510
3339 * domwalk.c (set_all_edges_as_executable): New function.
3340 (dom_walker::dom_walker): Convert bool param
3341 "skip_unreachable_blocks" to enum reachability. Move setup of
3342 edge flags to set_all_edges_as_executable and only do it when
3343 reachability is REACHABLE_BLOCKS.
3344 * domwalk.h (enum dom_walker::reachability): New enum.
3345 (dom_walker::dom_walker): Convert bool param
3346 "skip_unreachable_blocks" to enum reachability.
3347 (set_all_edges_as_executable): New decl.
3348 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3349 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3351 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3352 but converting true to REACHABLE_BLOCKS.
3353 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3355 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3356 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3357 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3359 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3360 if check_all_array_refs will be called.
3362 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3364 * tree.c (selftest::test_location_wrappers): Add more test
3367 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3369 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3370 (selftest::test_bit_in_range): Likewise.
3372 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3375 * doc/sourcebuild.texi (vect_float): Say that the selector
3376 only describes the situation when -funsafe-math-optimizations is on.
3377 (vect_float_strict): Document.
3379 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3381 PR tree-optimization/83965
3382 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3383 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3384 instead of checking only for a reduction.
3385 (vect_recog_widen_sum_pattern): Likewise.
3387 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3389 * predict.c (probably_never_executed): Only use precise profile info.
3390 (compute_function_frequency): Skip after inlining hack since we now
3391 have quality checking.
3393 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3395 * profile-count.h (profile_probability::very_unlikely,
3396 profile_probability::unlikely, profile_probability::even): Set
3397 precision to guessed.
3399 2018-01-23 Richard Biener <rguenther@suse.de>
3401 PR tree-optimization/83963
3402 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3403 Properly terminate dominator walk when crossing the exit edge not
3404 when visiting its source block.
3406 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3409 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3410 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3412 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3414 PR tree-optimization/83957
3415 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3416 semicolon after for body surrounded by braces.
3418 PR tree-optimization/83081
3419 * profile-count.h (profile_probability::split): New method.
3420 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3421 Use profile_probability::split.
3422 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3423 when splitting a single conditional jump into 2.
3425 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3427 PR tree-optimization/69452
3428 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3431 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3433 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3434 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3435 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3437 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3439 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3440 * config/rl78/rl78.md (movdi): New define_expand.
3441 * config/rl78/rl78.c (rl78_split_movdi): New function.
3443 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3446 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3448 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3449 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3450 128-bit to produce an UNSPEC move to get the double word with the
3451 signbit and then a shift directly to do signbit.
3452 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3453 implementation with a new version that just does either a direct
3454 move or a regular move. Move memory interface to separate insns.
3455 Move insns so they are next to the expander.
3456 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3457 with signbit move. Split big and little endian case.
3458 (signbit<mode>2_dm_mem_le): Likewise.
3459 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3460 (signbit<mode>2_dm2): Likewise.
3462 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3464 * config/rl78/rl78.md (anddi3): New define_expand.
3466 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3468 * config/rl78/rl78.md (umindi3): New define_expand.
3470 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3472 * config/rl78/rl78.md (smindi3): New define_expand.
3474 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3476 * config/rl78/rl78.md (smaxdi3): New define_expand.
3478 2018-01-22 Carl Love <cel@us.ibm.com>
3480 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3481 LVX_V1TI): Add macro expansion.
3482 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3483 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3484 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3485 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3486 Change check to determine if the instruction is a byte reversing
3487 entry. Fix typo in comment.
3488 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3489 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3490 Add def_builtin calls for new builtins.
3491 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3492 Add define_insn expansion.
3494 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3496 * config/rl78/rl78.md (umaxdi3): New define_expand.
3498 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3500 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3501 for non-QImode registers.
3503 2018-01-22 Richard Biener <rguenther@suse.de>
3505 PR tree-optimization/83963
3506 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3507 including the loop exit block.
3508 (scop_detection::merge_sese): Likewise.
3509 (scop_detection::add_scop): Do it here instead.
3511 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3513 * doc/sourcebuild.texi (arm_softfloat): Document.
3515 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3518 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3519 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3520 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3522 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3523 David Edelsohn <dje.gcc@gmail.com>
3526 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3527 Change "crset eq" to "crset 2".
3528 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3529 (*call_indirect_aix<mode>_nospec): Likewise.
3530 (*call_value_indirect_aix<mode>_nospec): Likewise.
3531 (*call_indirect_elfv2<mode>_nospec): Likewise.
3532 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3533 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3534 change assembly output from . to $.
3535 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3536 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3537 (*tablejump<mode>_internal1_nospec): Likewise.
3539 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3542 * config/sh/sh_optimize_sett_clrt.cc:
3543 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3545 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3547 PR tree-optimization/83940
3548 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3549 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3550 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3551 use it to pass back the definition type.
3552 (vect_check_store_rhs): Likewise rhs_dt_out.
3553 (vect_build_gather_load_calls): Add a mask_dt argument and use
3554 it instead of a call to vect_is_simple_use.
3555 (vectorizable_store): Update calls to vect_check_load_store_mask
3556 and vect_check_store_rhs. Use the dt returned by the latter instead
3557 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3558 instead of calls to vect_is_simple_use. Pass the scalar rather
3559 than the vector operand to vect_is_simple_use when handling
3560 second and subsequent copies of an rhs value.
3561 (vectorizable_load): Update calls to vect_check_load_store_mask
3562 and vect_build_gather_load_calls. Use the cached mask_dt and
3563 gs_info.offset_dt instead of calls to vect_is_simple_use.
3565 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3568 * tree-emutls.c: Include gimplify.h.
3569 (lower_emutls_2): New function.
3570 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3571 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3572 it before further processing.
3575 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3576 UINTVAL (trueop1) instead of INTVAL (op1).
3578 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3582 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3583 INCOMING_FRAME_SP_OFFSET if not defined.
3584 (scan_trace): Add ENTRY argument. If true and
3585 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3586 emit a note to adjust the CFA offset.
3587 (create_cfi_notes): Adjust scan_trace callers.
3588 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3589 INCOMING_FRAME_SP_OFFSET in the CIE.
3590 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3591 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3593 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3594 * doc/tm.texi: Regenerated.
3596 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3598 PR rtl-optimization/83147
3599 * lra-constraints.c (remove_inheritance_pseudos): Use
3600 lra_substitute_pseudo_within_insn.
3602 2018-01-19 Tom de Vries <tom@codesourcery.com>
3603 Cesar Philippidis <cesar@codesourcery.com>
3606 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3608 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3611 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3612 spaces for function labels.
3614 2018-01-19 Martin Liska <mliska@suse.cz>
3616 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3617 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3618 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3619 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3620 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3621 (PRED_CONST_RETURN): Change from 69 to 65.
3622 (PRED_NULL_RETURN): Change from 91 to 71.
3623 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3624 (PRED_LOOP_GUARD): Change from 66 to 73.
3626 2018-01-19 Martin Liska <mliska@suse.cz>
3628 * predict.c (predict_insn_def): Add new assert.
3629 (struct branch_predictor): Change type to signed integer.
3630 (test_prediction_value_range): Amend test to cover
3632 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3633 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3634 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3635 (PRED_LOOP_IV_COMPARE): Likewise.
3636 * predict.h (PROB_UNINITIALIZED): Define new constant.
3638 2018-01-19 Martin Liska <mliska@suse.cz>
3640 * predict.c (dump_prediction): Add new format for
3641 analyze_brprob.py script which is enabled with -details
3643 * profile-count.h (precise_p): New function.
3645 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3647 PR tree-optimization/83922
3648 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3649 there are no statements that need masking.
3650 (vect_active_double_reduction_p): New function.
3651 (vect_analyze_loop_operations): Use it when handling phis that
3652 are not in the loop header.
3654 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3656 PR tree-optimization/83914
3657 * tree-vect-loop.c (vectorizable_induction): Don't convert
3658 init_expr or apply the peeling adjustment for inductions
3659 that are nested within the vectorized loop.
3661 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3663 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3666 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3670 * function.h (gimplify_parameters): Add gimple_seq * argument.
3671 * function.c: Include gimple.h and options.h.
3672 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3673 for the added local temporaries if needed.
3674 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3675 if there are any parameter cleanups, wrap whole body into a
3676 try/finally with the cleanups.
3678 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3681 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3682 Use GET_MODE_CLASS for scalar floating point.
3684 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3688 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3689 Fix call of call_cgraph_insertion_hooks.
3691 2018-01-18 Martin Sebor <msebor@redhat.com>
3693 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3695 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3698 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3701 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3704 * common.opt: (-ffile-prefix-map): New option.
3705 * opts.c (common_handle_option): Defer it.
3706 * opts-global.c (handle_common_deferred_options): Handle it.
3707 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3708 * file-prefix-map.h: New file.
3709 (remap_debug_filename, add_debug_prefix_map): ...here.
3710 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3711 * final.c (debug_prefix_map, add_debug_prefix_map
3712 remap_debug_filename): Move to...
3713 * file-prefix-map.c: New file.
3714 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3715 generalize, get rid of alloca(), use strrchr() instead of strchr().
3716 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3717 Implement in terms of add_prefix_map().
3718 (remap_macro_filename, remap_debug_filename): Implement in term of
3720 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3721 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3722 * dbxout.c: Include file-prefix-map.h.
3723 * varasm.c: Likewise.
3724 * vmsdbgout.c: Likewise.
3725 * xcoffout.c: Likewise.
3726 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3727 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3728 * doc/invoke.texi (-ffile-prefix-map): Document.
3729 (-fdebug-prefix-map): Update description.
3731 2018-01-18 Martin Liska <mliska@suse.cz>
3733 * config/i386/i386.c (indirect_thunk_name): Document that also
3735 (output_indirect_thunk): Document why both instructions
3736 (pause and lfence) are generated.
3738 2018-01-18 Richard Biener <rguenther@suse.de>
3740 PR tree-optimization/83887
3741 * graphite-scop-detection.c
3742 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3743 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3744 (scop_detection::merge_sese): Re-implement with a flood-fill
3745 algorithm that properly finds a SESE region if it exists.
3747 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3750 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3751 pointer_diff optimizations use view_convert instead of convert.
3753 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3755 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3756 Generate different code for -mno-speculate-indirect-jumps.
3757 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3758 (*call_indirect_aix<mode>): Disable for
3759 -mno-speculate-indirect-jumps.
3760 (*call_indirect_aix<mode>_nospec): New define_insn.
3761 (*call_value_indirect_aix<mode>): Disable for
3762 -mno-speculate-indirect-jumps.
3763 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3764 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3765 -mno-speculate-indirect-jumps.
3766 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3768 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3770 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3771 long double type, set the flags for noting the default long double
3772 type, even if we don't pass or return a long double type.
3774 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3777 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3780 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3782 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3783 support for merge[hl].
3784 (fold_mergehl_helper): New helper function.
3785 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3786 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3787 (altivec_vmrglw_direct): Add xxmrglw insn.
3789 2018-01-17 Andrew Waterman <andrew@sifive.com>
3791 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3792 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3794 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3797 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3798 call the lto_location_cache before reading the
3799 DECL_SOURCE_LOCATION of the types.
3801 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3802 Richard Sandiford <richard.sandiford@linaro.org>
3804 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3805 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3806 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3807 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3808 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3810 * config/aarch64/constraints.md (aarch64_movti_operand):
3812 * config/aarch64/predicates.md (Uti): Add new constraint.
3814 2018-01-17 Carl Love <cel@us.ibm.com>
3816 * config/rs6000/vsx.md (define_expand xl_len_r,
3817 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3818 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3820 (define_expand, define_insn): Move the shift left from the
3821 define_insn to the define_expand for lxvl and stxvl instructions.
3822 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3823 and XL_LEN_R definitions to PURE.
3825 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3827 * config/i386/i386.c (indirect_thunk_name): Declare regno
3828 as unsigned int. Compare regno with INVALID_REGNUM.
3829 (output_indirect_thunk): Ditto.
3830 (output_indirect_thunk_function): Ditto.
3831 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3832 in the call to output_indirect_thunk_function.
3834 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3837 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3838 rather than the size of inner_type to determine the stack slot size
3839 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3841 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3844 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3847 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3849 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3850 endian Linux systems to optionally enable multilibs for selecting
3851 the long double type if the user configured an explicit type.
3852 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3853 have no long double multilibs if not defined.
3854 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3855 warn if the user used -mabi={ieee,ibm}longdouble and we built
3856 multilibs for long double.
3857 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3858 appropriate multilib option.
3859 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3861 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3862 for building long double multilibs.
3863 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3865 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3867 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3870 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3872 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3875 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3878 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3881 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3883 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3884 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3886 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3888 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3889 different rtl trees depending on TARGET_64BIT.
3890 (rs6000_gen_lvx): Likewise.
3892 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3894 * config/visium/visium.md (nop): Tweak comment.
3895 (hazard_nop): Likewise.
3897 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3899 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3900 -mspeculate-indirect-jumps.
3901 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3902 for -mno-speculate-indirect-jumps.
3903 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3904 (*call_value_indirect_elfv2<mode>): Disable for
3905 -mno-speculate-indirect-jumps.
3906 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3907 (indirect_jump): Emit different RTL for
3908 -mno-speculate-indirect-jumps.
3909 (*indirect_jump<mode>): Disable for
3910 -mno-speculate-indirect-jumps.
3911 (*indirect_jump<mode>_nospec): New define_insn.
3912 (tablejump): Emit different RTL for
3913 -mno-speculate-indirect-jumps.
3914 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3915 (tablejumpsi_nospec): New define_expand.
3916 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3917 (tablejumpdi_nospec): New define_expand.
3918 (*tablejump<mode>_internal1): Disable for
3919 -mno-speculate-indirect-jumps.
3920 (*tablejump<mode>_internal1_nospec): New define_insn.
3921 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3924 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3926 * caller-save.c (insert_save): Drop unnecessary parameter. All
3929 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3930 Richard Biener <rguenth@suse.de>
3933 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3934 return early, inline manually is_gimple_sizepos. Make sure if we
3935 call gimplify_expr we don't end up with a gimple constant.
3936 * tree.c (variably_modified_type_p): Don't return true for
3937 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3938 * gimplify.h (is_gimple_sizepos): Remove.
3940 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3942 PR tree-optimization/83857
3943 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3944 vectorizable_live_operation for pure SLP statements.
3945 (vectorizable_live_operation): Handle PHIs.
3947 2018-01-16 Richard Biener <rguenther@suse.de>
3949 PR tree-optimization/83867
3950 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3951 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3953 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3956 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3957 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3958 If off is not INTEGER_CST, issue a may not be aligned warning
3959 rather than isn't aligned. Use isn%'t rather than isn't.
3960 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3962 <case MULT_EXPR>: Improve the case when bottom and one of the
3963 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3964 operand, in that case check if the other operand is multiple of
3965 bottom divided by the INTEGER_CST operand.
3967 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3970 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3971 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3972 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3973 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3974 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3975 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3976 (pa_function_arg_size): New function.
3978 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3980 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3981 in a separate statement.
3983 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3985 PR tree-optimization/83847
3986 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3987 group gathers and scatters.
3989 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3991 PR rtl-optimization/86620
3992 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3994 PR rtl-optimization/83213
3995 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3996 to last if both are JUMP_INSNs.
3998 PR tree-optimization/83843
3999 * gimple-ssa-store-merging.c
4000 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4001 store_immediate_info for bswap/nop orig_stores.
4003 2018-01-15 Andrew Waterman <andrew@sifive.com>
4005 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4007 <UDIV>: Increase cost if !TARGET_DIV.
4009 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4011 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4012 (define_attr "cr_logical_3op"): New.
4013 (cceq_ior_compare): Adjust.
4014 (cceq_ior_compare_complement): Adjust.
4015 (*cceq_rev_compare): Adjust.
4016 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4017 (is_cracked_insn): Adjust.
4018 (insn_must_be_first_in_group): Adjust.
4019 * config/rs6000/40x.md: Adjust.
4020 * config/rs6000/440.md: Adjust.
4021 * config/rs6000/476.md: Adjust.
4022 * config/rs6000/601.md: Adjust.
4023 * config/rs6000/603.md: Adjust.
4024 * config/rs6000/6xx.md: Adjust.
4025 * config/rs6000/7450.md: Adjust.
4026 * config/rs6000/7xx.md: Adjust.
4027 * config/rs6000/8540.md: Adjust.
4028 * config/rs6000/cell.md: Adjust.
4029 * config/rs6000/e300c2c3.md: Adjust.
4030 * config/rs6000/e500mc.md: Adjust.
4031 * config/rs6000/e500mc64.md: Adjust.
4032 * config/rs6000/e5500.md: Adjust.
4033 * config/rs6000/e6500.md: Adjust.
4034 * config/rs6000/mpc.md: Adjust.
4035 * config/rs6000/power4.md: Adjust.
4036 * config/rs6000/power5.md: Adjust.
4037 * config/rs6000/power6.md: Adjust.
4038 * config/rs6000/power7.md: Adjust.
4039 * config/rs6000/power8.md: Adjust.
4040 * config/rs6000/power9.md: Adjust.
4041 * config/rs6000/rs64.md: Adjust.
4042 * config/rs6000/titan.md: Adjust.
4044 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4046 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4047 ix86_indirect_branch_register logic.
4049 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4051 * config/i386/constraints.md (Bs): Update
4052 ix86_indirect_branch_register check. Don't check
4053 ix86_indirect_branch_register with GOT_memory_operand.
4055 * config/i386/predicates.md (GOT_memory_operand): Don't check
4056 ix86_indirect_branch_register here.
4057 (GOT32_symbol_operand): Likewise.
4059 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4061 * config/i386/predicates.md (constant_call_address_operand):
4062 Rewrite ix86_indirect_branch_register logic.
4063 (sibcall_insn_operand): Likewise.
4065 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4067 * config/i386/constraints.md (Bs): Replace
4068 ix86_indirect_branch_thunk_register with
4069 ix86_indirect_branch_register.
4071 * config/i386/i386.md (indirect_jump): Likewise.
4072 (tablejump): Likewise.
4073 (*sibcall_memory): Likewise.
4074 (*sibcall_value_memory): Likewise.
4075 Peepholes of indirect call and jump via memory: Likewise.
4076 * config/i386/i386.opt: Likewise.
4077 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4078 (GOT_memory_operand): Likewise.
4079 (call_insn_operand): Likewise.
4080 (sibcall_insn_operand): Likewise.
4081 (GOT32_symbol_operand): Likewise.
4083 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4086 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4087 type rather than type addr's type points to.
4088 (expand_omp_atomic_mutex): Likewise.
4089 (expand_omp_atomic): Likewise.
4091 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4094 * config/i386/i386.c (output_indirect_thunk_function): Use
4095 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4096 for __x86_return_thunk.
4098 2018-01-15 Richard Biener <rguenther@suse.de>
4101 * expmed.c (extract_bit_field_1): Fix typo.
4103 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4106 * config/arm/iterators.md (VF): New mode iterator.
4107 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4108 Remove integer-related logic from pattern.
4109 (neon_vabd<mode>_3): Likewise.
4111 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4114 * common.opt (fstrict-overflow): No longer an alias.
4115 (fwrapv-pointer): New option.
4116 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4117 also for pointer types based on flag_wrapv_pointer.
4118 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4119 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4120 opts->x_flag_wrapv got set.
4121 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4122 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4123 POINTER_TYPE_OVERFLOW_UNDEFINED.
4124 * match.pd: Likewise in address comparison pattern.
4125 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4127 2018-01-15 Richard Biener <rguenther@suse.de>
4130 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4131 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4132 Reset type names to their identifier if their TYPE_DECL doesn't
4133 have linkage (and thus is used for ODR and devirt).
4134 (save_debug_info_for_decl): Remove.
4135 (save_debug_info_for_type): Likewise.
4136 (add_tree_to_fld_list): Adjust.
4137 * tree-pretty-print.c (dump_generic_node): Make dumping of
4138 type names more robust.
4140 2018-01-15 Richard Biener <rguenther@suse.de>
4142 * BASE-VER: Bump to 8.0.1.
4144 2018-01-14 Martin Sebor <msebor@redhat.com>
4147 * builtins.c (check_access): Avoid warning when the no-warning bit
4150 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4152 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4153 * ira-color (allocno_hard_regs_compare): Likewise.
4155 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4158 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4159 Use .pushsection/.popsection.
4161 2018-01-14 Martin Sebor <msebor@redhat.com>
4164 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4166 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4168 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4169 entry from extra_headers.
4170 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4171 extra_headers, make the list bitwise identical to the i?86-*-* one.
4173 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4175 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4176 -mcmodel=large with -mindirect-branch=thunk,
4177 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4178 -mfunction-return=thunk-extern.
4179 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4180 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4181 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4183 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4185 * config/i386/i386.c (print_reg): Print the name of the full
4186 integer register without '%'.
4187 (ix86_print_operand): Handle 'V'.
4188 * doc/extend.texi: Document 'V' modifier.
4190 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4192 * config/i386/constraints.md (Bs): Disallow memory operand for
4193 -mindirect-branch-register.
4195 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4196 (GOT_memory_operand): Likewise.
4197 (call_insn_operand): Likewise.
4198 (sibcall_insn_operand): Likewise.
4199 (GOT32_symbol_operand): Likewise.
4200 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4201 for -mindirect-branch-register.
4202 (tablejump): Likewise.
4203 (*sibcall_memory): Likewise.
4204 (*sibcall_value_memory): Likewise.
4205 Disallow peepholes of indirect call and jump via memory for
4206 -mindirect-branch-register.
4207 (*call_pop): Replace m with Bw.
4208 (*call_value_pop): Likewise.
4209 (*sibcall_pop_memory): Replace m with Bs.
4210 * config/i386/i386.opt (mindirect-branch-register): New option.
4211 * doc/invoke.texi: Document -mindirect-branch-register option.
4213 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4215 * config/i386/i386-protos.h (ix86_output_function_return): New.
4216 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4217 set function_return_type.
4218 (indirect_thunk_name): Add ret_p to indicate thunk for function
4220 (output_indirect_thunk_function): Pass false to
4221 indirect_thunk_name.
4222 (ix86_output_indirect_branch_via_reg): Likewise.
4223 (ix86_output_indirect_branch_via_push): Likewise.
4224 (output_indirect_thunk_function): Create alias for function
4225 return thunk if regno < 0.
4226 (ix86_output_function_return): New function.
4227 (ix86_handle_fndecl_attribute): Handle function_return.
4228 (ix86_attribute_table): Add function_return.
4229 * config/i386/i386.h (machine_function): Add
4230 function_return_type.
4231 * config/i386/i386.md (simple_return_internal): Use
4232 ix86_output_function_return.
4233 (simple_return_internal_long): Likewise.
4234 * config/i386/i386.opt (mfunction-return=): New option.
4235 (indirect_branch): Mention -mfunction-return=.
4236 * doc/extend.texi: Document function_return function attribute.
4237 * doc/invoke.texi: Document -mfunction-return= option.
4239 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4241 * config/i386/i386-opts.h (indirect_branch): New.
4242 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4243 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4244 with local indirect jump when converting indirect call and jump.
4245 (ix86_set_indirect_branch_type): New.
4246 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4247 (indirectlabelno): New.
4248 (indirect_thunk_needed): Likewise.
4249 (indirect_thunk_bnd_needed): Likewise.
4250 (indirect_thunks_used): Likewise.
4251 (indirect_thunks_bnd_used): Likewise.
4252 (INDIRECT_LABEL): Likewise.
4253 (indirect_thunk_name): Likewise.
4254 (output_indirect_thunk): Likewise.
4255 (output_indirect_thunk_function): Likewise.
4256 (ix86_output_indirect_branch_via_reg): Likewise.
4257 (ix86_output_indirect_branch_via_push): Likewise.
4258 (ix86_output_indirect_branch): Likewise.
4259 (ix86_output_indirect_jmp): Likewise.
4260 (ix86_code_end): Call output_indirect_thunk_function if needed.
4261 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4263 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4264 (ix86_attribute_table): Add indirect_branch.
4265 * config/i386/i386.h (machine_function): Add indirect_branch_type
4266 and has_local_indirect_jump.
4267 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4269 (tablejump): Likewise.
4270 (*indirect_jump): Use ix86_output_indirect_jmp.
4271 (*tablejump_1): Likewise.
4272 (simple_return_indirect_internal): Likewise.
4273 * config/i386/i386.opt (mindirect-branch=): New option.
4274 (indirect_branch): New.
4277 (thunk-inline): Likewise.
4278 (thunk-extern): Likewise.
4279 * doc/extend.texi: Document indirect_branch function attribute.
4280 * doc/invoke.texi: Document -mindirect-branch= option.
4282 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4285 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4287 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4289 * ipa-inline.c (want_inline_small_function_p): Return false if
4290 inlining has already failed with CIF_FINAL_ERROR.
4291 (update_caller_keys): Call want_inline_small_function_p before
4293 (update_callee_keys): Likewise.
4295 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4297 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4299 (rs6000_quadword_masked_address_p): Likewise.
4300 (quad_aligned_load_p): Likewise.
4301 (quad_aligned_store_p): Likewise.
4302 (const_load_sequence_p): Add comment to describe the outer-most loop.
4303 (mimic_memory_attributes_and_flags): New function.
4304 (rs6000_gen_stvx): Likewise.
4305 (replace_swapped_aligned_store): Likewise.
4306 (rs6000_gen_lvx): Likewise.
4307 (replace_swapped_aligned_load): Likewise.
4308 (replace_swapped_load_constant): Capitalize argument name in
4309 comment describing this function.
4310 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4311 and stores that access quad-word aligned addresses and replace
4312 with stvx or lvx instructions when appropriate.
4313 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4314 New function prototype.
4315 (rs6000_quadword_masked_address_p): Likewise.
4316 (rs6000_gen_lvx): Likewise.
4317 (rs6000_gen_stvx): Likewise.
4318 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4319 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4320 when memory address is aligned.
4321 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4322 this split to select lvx instruction when memory address is aligned.
4323 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4324 instruction when memory address is aligned.
4325 (*vsx_le_perm_load_v16qi): Likewise.
4326 (four unnamed splitters): Modify to select the stvx instruction
4327 when memory is aligned.
4329 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4331 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4332 which appears in the queue multiple times.
4334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4335 Alan Hayward <alan.hayward@arm.com>
4336 David Sherwood <david.sherwood@arm.com>
4338 * tree-vectorizer.h (vec_lower_bound): New structure.
4339 (_loop_vec_info): Add check_nonzero and lower_bounds.
4340 (LOOP_VINFO_CHECK_NONZERO): New macro.
4341 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4342 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4343 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4344 fields. Make seg_len the distance travelled, not including the
4346 (dr_direction_indicator): Declare.
4347 (dr_zero_step_indicator): Likewise.
4348 (dr_known_forward_stride_p): Likewise.
4349 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4351 (runtime_alias_check_p): Allow runtime alias checks with
4353 (operator ==): Compare access_size and align.
4354 (prune_runtime_alias_test_list): Rework for new distinction between
4355 the access_size and seg_len.
4356 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4358 (get_segment_min_max): New function.
4359 (create_intersect_range_checks): Use it.
4360 (dr_step_indicator): New function.
4361 (dr_direction_indicator): Likewise.
4362 (dr_zero_step_indicator): Likewise.
4363 (dr_known_forward_stride_p): Likewise.
4364 * tree-loop-distribution.c (data_ref_segment_size): Return
4365 DR_STEP * (niters - 1).
4366 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4368 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4369 (vect_preserves_scalar_order_p): New function, split out from...
4370 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4371 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4372 (vect_vfa_access_size): New function.
4373 (vect_vfa_align): Likewise.
4374 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4375 (dump_lower_bound): New function.
4376 (vect_check_lower_bound): Likewise.
4377 (vect_small_gap_p): Likewise.
4378 (vectorizable_with_step_bound_p): Likewise.
4379 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4380 depencies if the vectorization factor is 1. Convert the checks
4381 for nonzero steps into checks on the bounds of DR_STEP. Try using
4382 a bunds check for variable steps if the minimum required step is
4383 relatively small. Update calls to the dr_with_seg_len
4384 constructor and to vect_compile_time_alias.
4385 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4387 (vect_loop_versioning): Call it.
4388 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4390 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4392 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4393 Alan Hayward <alan.hayward@arm.com>
4394 David Sherwood <david.sherwood@arm.com>
4396 * doc/sourcebuild.texi (vect_scatter_store): Document.
4397 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4399 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4401 * genopinit.c (main): Add supports_vec_scatter_store and
4402 supports_vec_scatter_store_cached to target_optabs.
4403 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4404 IFN_MASK_SCATTER_STORE.
4405 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4407 * internal-fn.h (internal_store_fn_p): Declare.
4408 (internal_fn_stored_value_index): Likewise.
4409 * internal-fn.c (scatter_store_direct): New macro.
4410 (expand_scatter_store_optab_fn): New function.
4411 (direct_scatter_store_optab_supported_p): New macro.
4412 (internal_store_fn_p): New function.
4413 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4414 IFN_MASK_SCATTER_STORE.
4415 (internal_fn_mask_index): Likewise.
4416 (internal_fn_stored_value_index): New function.
4417 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4419 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4420 * optabs-query.c (supports_vec_scatter_store_p): New function.
4421 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4422 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4423 true for scatter stores.
4424 (vect_gather_scatter_fn_p): Handle scatter stores too.
4425 (vect_check_gather_scatter): Consider using scatter stores if
4426 supports_vec_scatter_store_p.
4427 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4429 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4430 internal_fn_stored_value_index.
4431 (check_load_store_masking): Handle scatter stores too.
4432 (vect_get_store_rhs): Make public.
4433 (vectorizable_call): Use internal_store_fn_p.
4434 (vectorizable_store): Handle scatter store internal functions.
4435 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4436 when deciding whether the end of the group has been reached.
4437 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4438 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4439 (mask_scatter_store<mode>): New insns.
4441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4442 Alan Hayward <alan.hayward@arm.com>
4443 David Sherwood <david.sherwood@arm.com>
4445 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4446 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4447 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4449 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4450 Use vect_truncate_gather_scatter_offset if we can't treat the
4451 operation as a normal gather load or scatter store.
4452 (get_group_load_store_type): Take the gather_scatter_info
4453 as argument. Try using a gather load or scatter store for
4454 single-element groups.
4455 (get_load_store_type): Update calls to get_group_load_store_type
4456 and vect_use_strided_gather_scatters_p.
4458 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4459 Alan Hayward <alan.hayward@arm.com>
4460 David Sherwood <david.sherwood@arm.com>
4462 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4463 optional tree argument.
4464 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4466 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4467 but continue to use the current value as a fallback.
4468 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4469 to compare the updates.
4470 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4471 (get_load_store_type): Use it when handling a strided access.
4472 (vect_get_strided_load_store_ops): New function.
4473 (vect_get_data_ptr_increment): Likewise.
4474 (vectorizable_load): Handle strided gather loads. Always pass
4475 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4478 Alan Hayward <alan.hayward@arm.com>
4479 David Sherwood <david.sherwood@arm.com>
4481 * doc/md.texi (gather_load@var{m}): Document.
4482 (mask_gather_load@var{m}): Likewise.
4483 * genopinit.c (main): Add supports_vec_gather_load and
4484 supports_vec_gather_load_cached to target_optabs.
4485 * optabs-tree.c (init_tree_optimization_optabs): Use
4486 ggc_cleared_alloc to allocate target_optabs.
4487 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4488 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4490 * internal-fn.h (internal_load_fn_p): Declare.
4491 (internal_gather_scatter_fn_p): Likewise.
4492 (internal_fn_mask_index): Likewise.
4493 (internal_gather_scatter_fn_supported_p): Likewise.
4494 * internal-fn.c (gather_load_direct): New macro.
4495 (expand_gather_load_optab_fn): New function.
4496 (direct_gather_load_optab_supported_p): New macro.
4497 (direct_internal_fn_optab): New function.
4498 (internal_load_fn_p): Likewise.
4499 (internal_gather_scatter_fn_p): Likewise.
4500 (internal_fn_mask_index): Likewise.
4501 (internal_gather_scatter_fn_supported_p): Likewise.
4502 * optabs-query.c (supports_at_least_one_mode_p): New function.
4503 (supports_vec_gather_load_p): Likewise.
4504 * optabs-query.h (supports_vec_gather_load_p): Declare.
4505 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4506 and memory_type field.
4507 (NUM_PATTERNS): Bump to 15.
4508 * tree-vect-data-refs.c: Include internal-fn.h.
4509 (vect_gather_scatter_fn_p): New function.
4510 (vect_describe_gather_scatter_call): Likewise.
4511 (vect_check_gather_scatter): Try using internal functions for
4512 gather loads. Recognize existing calls to a gather load function.
4513 (vect_analyze_data_refs): Consider using gather loads if
4514 supports_vec_gather_load_p.
4515 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4516 (vect_get_gather_scatter_offset_type): Likewise.
4517 (vect_convert_mask_for_vectype): Likewise.
4518 (vect_add_conversion_to_patterm): Likewise.
4519 (vect_try_gather_scatter_pattern): Likewise.
4520 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4521 (vect_vect_recog_func_ptrs): Add it.
4522 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4523 internal_fn_mask_index and internal_gather_scatter_fn_p.
4524 (check_load_store_masking): Take the gather_scatter_info as an
4525 argument and handle gather loads.
4526 (vect_get_gather_scatter_ops): New function.
4527 (vectorizable_call): Check internal_load_fn_p.
4528 (vectorizable_load): Likewise. Handle gather load internal
4530 (vectorizable_store): Update call to check_load_store_masking.
4531 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4532 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4533 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4534 (aarch64_gather_scale_operand_d): New predicates.
4535 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4536 (mask_gather_load<mode>): New insns.
4538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4539 Alan Hayward <alan.hayward@arm.com>
4540 David Sherwood <david.sherwood@arm.com>
4542 * optabs.def (fold_left_plus_optab): New optab.
4543 * doc/md.texi (fold_left_plus_@var{m}): Document.
4544 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4545 * internal-fn.c (fold_left_direct): Define.
4546 (expand_fold_left_optab_fn): Likewise.
4547 (direct_fold_left_optab_supported_p): Likewise.
4548 * fold-const-call.c (fold_const_fold_left): New function.
4549 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4550 * tree-parloops.c (valid_reduction_p): New function.
4551 (gather_scalar_reductions): Use it.
4552 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4553 (vect_finish_replace_stmt): Declare.
4554 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4555 (needs_fold_left_reduction_p): New function, split out from...
4556 (vect_is_simple_reduction): ...here. Accept reductions that
4557 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4558 (vect_force_simple_reduction): Also store the reduction type in
4559 the assignment's STMT_VINFO_REDUC_TYPE.
4560 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4561 (merge_with_identity): New function.
4562 (vect_expand_fold_left): Likewise.
4563 (vectorize_fold_left_reduction): Likewise.
4564 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4565 scalar phi in place for it. Check for target support and reject
4566 cases that would reassociate the operation. Defer the transform
4567 phase to vectorize_fold_left_reduction.
4568 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4569 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4570 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4574 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4575 call to ifc_temp_var.
4577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4578 Alan Hayward <alan.hayward@arm.com>
4579 David Sherwood <david.sherwood@arm.com>
4581 * target.def (legitimize_address_displacement): Take the original
4582 offset as a poly_int.
4583 * targhooks.h (default_legitimize_address_displacement): Update
4585 * targhooks.c (default_legitimize_address_displacement): Likewise.
4586 * doc/tm.texi: Regenerate.
4587 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4588 as an argument, moving assert of ad->disp == ad->disp_term to...
4589 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4590 Try calling targetm.legitimize_address_displacement before expanding
4591 the address rather than afterwards, and adjust for the new interface.
4592 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4593 Match the new hook interface. Handle SVE addresses.
4594 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4599 * Makefile.in (OBJS): Add early-remat.o.
4600 * target.def (select_early_remat_modes): New hook.
4601 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4602 * doc/tm.texi: Regenerate.
4603 * targhooks.h (default_select_early_remat_modes): Declare.
4604 * targhooks.c (default_select_early_remat_modes): New function.
4605 * timevar.def (TV_EARLY_REMAT): New timevar.
4606 * passes.def (pass_early_remat): New pass.
4607 * tree-pass.h (make_pass_early_remat): Declare.
4608 * early-remat.c: New file.
4609 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4611 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4614 Alan Hayward <alan.hayward@arm.com>
4615 David Sherwood <david.sherwood@arm.com>
4617 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4618 vfm1 with a bound_epilog parameter.
4619 (vect_do_peeling): Update calls accordingly, and move the prologue
4620 call earlier in the function. Treat the base bound_epilog as 0 for
4621 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4622 this base when peeling for gaps.
4623 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4624 with fully-masked loops.
4625 (vect_estimate_min_profitable_iters): Handle the single peeled
4626 iteration in that case.
4628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4629 Alan Hayward <alan.hayward@arm.com>
4630 David Sherwood <david.sherwood@arm.com>
4632 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4633 single-element interleaving even if the size is not a power of 2.
4634 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4635 accesses for single-element interleaving if the group size is
4638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4639 Alan Hayward <alan.hayward@arm.com>
4640 David Sherwood <david.sherwood@arm.com>
4642 * doc/md.texi (fold_extract_last_@var{m}): Document.
4643 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4644 * optabs.def (fold_extract_last_optab): New optab.
4645 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4646 * internal-fn.c (fold_extract_direct): New macro.
4647 (expand_fold_extract_optab_fn): Likewise.
4648 (direct_fold_extract_optab_supported_p): Likewise.
4649 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4650 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4651 EXTRACT_LAST_REDUCTION.
4652 (get_initial_def_for_reduction): Do not create an initial vector
4653 for EXTRACT_LAST_REDUCTION reductions.
4654 (vectorizable_reduction): Leave the scalar phi in place for
4655 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4656 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4657 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4658 transform phase to vectorizable_condition.
4659 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4661 (vect_finish_stmt_generation): ...here.
4662 (vect_finish_replace_stmt): New function.
4663 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4664 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4666 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4668 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4669 Alan Hayward <alan.hayward@arm.com>
4670 David Sherwood <david.sherwood@arm.com>
4672 * doc/md.texi (extract_last_@var{m}): Document.
4673 * optabs.def (extract_last_optab): New optab.
4674 * internal-fn.def (EXTRACT_LAST): New internal function.
4675 * internal-fn.c (cond_unary_direct): New macro.
4676 (expand_cond_unary_optab_fn): Likewise.
4677 (direct_cond_unary_optab_supported_p): Likewise.
4678 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4679 loops using EXTRACT_LAST.
4680 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4681 (extract_last_<mode>): ...this optab.
4682 (vec_extract<mode><Vel>): Update accordingly.
4684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4685 Alan Hayward <alan.hayward@arm.com>
4686 David Sherwood <david.sherwood@arm.com>
4688 * target.def (empty_mask_is_expensive): New hook.
4689 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4690 * doc/tm.texi: Regenerate.
4691 * targhooks.h (default_empty_mask_is_expensive): Declare.
4692 * targhooks.c (default_empty_mask_is_expensive): New function.
4693 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4694 if the target says that empty masks are expensive.
4695 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4697 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4700 Alan Hayward <alan.hayward@arm.com>
4701 David Sherwood <david.sherwood@arm.com>
4703 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4704 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4705 (vect_use_loop_mask_for_alignment_p): New function.
4706 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4707 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4708 niters_skip argument. Make sure that the first niters_skip elements
4709 of the first iteration are inactive.
4710 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4711 Update call to vect_set_loop_masks_directly.
4712 (get_misalign_in_elems): New function, split out from...
4713 (vect_gen_prolog_loop_niters): ...here.
4714 (vect_update_init_of_dr): Take a code argument that specifies whether
4715 the adjustment should be added or subtracted.
4716 (vect_update_init_of_drs): Likewise.
4717 (vect_prepare_for_masked_peels): New function.
4718 (vect_do_peeling): Skip prologue peeling if we're using a mask
4719 instead. Update call to vect_update_inits_of_drs.
4720 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4722 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4723 alignment. Do not include the number of peeled iterations in
4724 the minimum threshold in that case.
4725 (vectorizable_induction): Adjust the start value down by
4726 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4727 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4728 Take the number of skipped iterations into account when calculating
4730 * tree-vect-stmts.c (vect_gen_while_not): New function.
4732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4733 Alan Hayward <alan.hayward@arm.com>
4734 David Sherwood <david.sherwood@arm.com>
4736 * doc/sourcebuild.texi (vect_fully_masked): Document.
4737 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4739 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4741 (vect_analyze_loop_2): ...here. Don't check the vectorization
4742 factor against the number of loop iterations if the loop is
4745 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4746 Alan Hayward <alan.hayward@arm.com>
4747 David Sherwood <david.sherwood@arm.com>
4749 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4750 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4751 (dump_groups): Update accordingly.
4752 (iv_use::mem_type): New member variable.
4753 (address_p): New function.
4754 (record_use): Add a mem_type argument and initialize the new
4756 (record_group_use): Add a mem_type argument. Use address_p.
4757 Remove obsolete null checks of base_object. Update call to record_use.
4758 (find_interesting_uses_op): Update call to record_group_use.
4759 (find_interesting_uses_cond): Likewise.
4760 (find_interesting_uses_address): Likewise.
4761 (get_mem_type_for_internal_fn): New function.
4762 (find_address_like_use): Likewise.
4763 (find_interesting_uses_stmt): Try find_address_like_use before
4764 calling find_interesting_uses_op.
4765 (addr_offset_valid_p): Use the iv mem_type field as the type
4766 of the addressed memory.
4767 (add_autoinc_candidates): Likewise.
4768 (get_address_cost): Likewise.
4769 (split_small_address_groups_p): Use address_p.
4770 (split_address_groups): Likewise.
4771 (add_iv_candidate_for_use): Likewise.
4772 (autoinc_possible_for_pair): Likewise.
4773 (rewrite_groups): Likewise.
4774 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4775 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4776 (get_alias_ptr_type_for_ptr_address): New function.
4777 (rewrite_use_address): Rewrite address uses in calls that were
4778 identified by find_address_like_use.
4780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4781 Alan Hayward <alan.hayward@arm.com>
4782 David Sherwood <david.sherwood@arm.com>
4784 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4786 * gimple-expr.h (is_gimple_addressable: Likewise.
4787 * gimple-expr.c (is_gimple_address): Likewise.
4788 * internal-fn.c (expand_call_mem_ref): New function.
4789 (expand_mask_load_optab_fn): Use it.
4790 (expand_mask_store_optab_fn): Likewise.
4792 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4793 Alan Hayward <alan.hayward@arm.com>
4794 David Sherwood <david.sherwood@arm.com>
4796 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4797 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4798 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4799 (cond_umax@var{mode}): Document.
4800 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4801 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4802 (cond_umin_optab, cond_umax_optab): New optabs.
4803 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4804 (COND_IOR, COND_XOR): New internal functions.
4805 * internal-fn.h (get_conditional_internal_fn): Declare.
4806 * internal-fn.c (cond_binary_direct): New macro.
4807 (expand_cond_binary_optab_fn): Likewise.
4808 (direct_cond_binary_optab_supported_p): Likewise.
4809 (get_conditional_internal_fn): New function.
4810 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4811 Cope with reduction statements that are vectorized as calls rather
4813 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4814 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4815 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4816 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4817 (UNSPEC_COND_EOR): New unspecs.
4818 (optab): Add mappings for them.
4819 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4820 (sve_int_op, sve_fp_op): New int attributes.
4822 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4823 Alan Hayward <alan.hayward@arm.com>
4824 David Sherwood <david.sherwood@arm.com>
4826 * optabs.def (while_ult_optab): New optab.
4827 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4828 * internal-fn.def (WHILE_ULT): New internal function.
4829 * internal-fn.h (direct_internal_fn_supported_p): New override
4830 that takes two types as argument.
4831 * internal-fn.c (while_direct): New macro.
4832 (expand_while_optab_fn): New function.
4833 (convert_optab_supported_p): Likewise.
4834 (direct_while_optab_supported_p): New macro.
4835 * wide-int.h (wi::udiv_ceil): New function.
4836 * tree-vectorizer.h (rgroup_masks): New structure.
4837 (vec_loop_masks): New typedef.
4838 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4840 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4841 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4842 (vect_max_vf): New function.
4843 (slpeel_make_loop_iterate_ntimes): Delete.
4844 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4845 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4846 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4847 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4848 internal-fn.h, stor-layout.h and optabs-query.h.
4849 (vect_set_loop_mask): New function.
4850 (add_preheader_seq): Likewise.
4851 (add_header_seq): Likewise.
4852 (interleave_supported_p): Likewise.
4853 (vect_maybe_permute_loop_masks): Likewise.
4854 (vect_set_loop_masks_directly): Likewise.
4855 (vect_set_loop_condition_masked): Likewise.
4856 (vect_set_loop_condition_unmasked): New function, split out from
4857 slpeel_make_loop_iterate_ntimes.
4858 (slpeel_make_loop_iterate_ntimes): Rename to..
4859 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4860 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4861 (vect_do_peeling): Update call accordingly.
4862 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4864 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4865 mask_compare_type, can_fully_mask_p and fully_masked_p.
4866 (release_vec_loop_masks): New function.
4867 (_loop_vec_info): Use it to free the loop masks.
4868 (can_produce_all_loop_masks_p): New function.
4869 (vect_get_max_nscalars_per_iter): Likewise.
4870 (vect_verify_full_masking): Likewise.
4871 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4872 retries, and free the mask rgroups before retrying. Check loop-wide
4873 reasons for disallowing fully-masked loops. Make the final decision
4874 about whether use a fully-masked loop or not.
4875 (vect_estimate_min_profitable_iters): Do not assume that peeling
4876 for the number of iterations will be needed for fully-masked loops.
4877 (vectorizable_reduction): Disable fully-masked loops.
4878 (vectorizable_live_operation): Likewise.
4879 (vect_halve_mask_nunits): New function.
4880 (vect_double_mask_nunits): Likewise.
4881 (vect_record_loop_mask): Likewise.
4882 (vect_get_loop_mask): Likewise.
4883 (vect_transform_loop): Handle the case in which the final loop
4884 iteration might handle a partial vector. Call vect_set_loop_condition
4885 instead of slpeel_make_loop_iterate_ntimes.
4886 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4887 (check_load_store_masking): New function.
4888 (prepare_load_store_mask): Likewise.
4889 (vectorizable_store): Handle fully-masked loops.
4890 (vectorizable_load): Likewise.
4891 (supportable_widening_operation): Use vect_halve_mask_nunits for
4893 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4894 (vect_gen_while): New function.
4895 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4896 (aarch64_uqdec<mode>): New insn.
4898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4899 Alan Hayward <alan.hayward@arm.com>
4900 David Sherwood <david.sherwood@arm.com>
4902 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4903 (reduc_xor_scal_optab): New optabs.
4904 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4905 (reduc_xor_scal_@var{m}): Document.
4906 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4907 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4909 * fold-const-call.c (fold_const_call): Handle them.
4910 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4911 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4912 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4913 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4914 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4915 (UNSPEC_XORV): New unspecs.
4916 (optab): Add entries for them.
4917 (BITWISEV): New int iterator.
4918 (bit_reduc_op): New int attributes.
4920 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4921 Alan Hayward <alan.hayward@arm.com>
4922 David Sherwood <david.sherwood@arm.com>
4924 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4925 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4926 * optabs.def (vec_shl_insert_optab): New optab.
4927 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4928 (duplicate_and_interleave): Likewise.
4929 * tree-vect-loop.c: Include internal-fn.h.
4930 (neutral_op_for_slp_reduction): New function, split out from
4931 get_initial_defs_for_reduction.
4932 (get_initial_def_for_reduction): Handle option 2 for variable-length
4933 vectors by loading the neutral value into a vector and then shifting
4934 the initial value into element 0.
4935 (get_initial_defs_for_reduction): Replace the code argument with
4936 the neutral value calculated by neutral_op_for_slp_reduction.
4937 Use gimple_build_vector for constant-length vectors.
4938 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4939 but the first group_size elements have a neutral value.
4940 Use duplicate_and_interleave otherwise.
4941 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4942 Update call to get_initial_defs_for_reduction. Handle SLP
4943 reductions for variable-length vectors by creating one vector
4944 result for each scalar result, with the elements associated
4945 with other scalar results stubbed out with the neutral value.
4946 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4947 Require IFN_VEC_SHL_INSERT for double reductions on
4948 variable-length vectors, or SLP reductions that have
4949 a neutral value. Require can_duplicate_and_interleave_p
4950 support for variable-length unchained SLP reductions if there
4951 is no neutral value, such as for MIN/MAX reductions. Also require
4952 the number of vector elements to be a multiple of the number of
4953 SLP statements when doing variable-length unchained SLP reductions.
4954 Update call to vect_create_epilog_for_reduction.
4955 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4956 and remove initial values.
4957 (duplicate_and_interleave): Make public.
4958 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4959 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4961 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4962 Alan Hayward <alan.hayward@arm.com>
4963 David Sherwood <david.sherwood@arm.com>
4965 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4966 (can_duplicate_and_interleave_p): New function.
4967 (vect_get_and_check_slp_defs): Take the vector of statements
4968 rather than just the current one. Remove excess parentheses.
4969 Restriction rejectinon of vect_constant_def and vect_external_def
4970 for variable-length vectors to boolean types, or types for which
4971 can_duplicate_and_interleave_p is false.
4972 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4973 (duplicate_and_interleave): New function.
4974 (vect_get_constant_vectors): Use gimple_build_vector for
4975 constant-length vectors and suitable variable-length constant
4976 vectors. Use duplicate_and_interleave for other variable-length
4977 vectors. Don't defer the update when inserting new statements.
4979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4980 Alan Hayward <alan.hayward@arm.com>
4981 David Sherwood <david.sherwood@arm.com>
4983 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4984 min_profitable_iters doesn't go negative.
4986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4987 Alan Hayward <alan.hayward@arm.com>
4988 David Sherwood <david.sherwood@arm.com>
4990 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4991 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4992 * optabs.def (vec_mask_load_lanes_optab): New optab.
4993 (vec_mask_store_lanes_optab): Likewise.
4994 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4995 (MASK_STORE_LANES): Likewise.
4996 * internal-fn.c (mask_load_lanes_direct): New macro.
4997 (mask_store_lanes_direct): Likewise.
4998 (expand_mask_load_optab_fn): Handle masked operations.
4999 (expand_mask_load_lanes_optab_fn): New macro.
5000 (expand_mask_store_optab_fn): Handle masked operations.
5001 (expand_mask_store_lanes_optab_fn): New macro.
5002 (direct_mask_load_lanes_optab_supported_p): Likewise.
5003 (direct_mask_store_lanes_optab_supported_p): Likewise.
5004 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5006 (vect_load_lanes_supported): Likewise.
5007 * tree-vect-data-refs.c (strip_conversion): New function.
5008 (can_group_stmts_p): Likewise.
5009 (vect_analyze_data_ref_accesses): Use it instead of checking
5010 for a pair of assignments.
5011 (vect_store_lanes_supported): Take a masked_p parameter.
5012 (vect_load_lanes_supported): Likewise.
5013 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5014 vect_store_lanes_supported and vect_load_lanes_supported.
5015 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5016 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5017 parameter. Don't allow gaps for masked accesses.
5018 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5019 and vect_load_lanes_supported.
5020 (get_load_store_type): Take a masked_p parameter and update
5021 call to get_group_load_store_type.
5022 (vectorizable_store): Update call to get_load_store_type.
5023 Handle IFN_MASK_STORE_LANES.
5024 (vectorizable_load): Update call to get_load_store_type.
5025 Handle IFN_MASK_LOAD_LANES.
5027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5028 Alan Hayward <alan.hayward@arm.com>
5029 David Sherwood <david.sherwood@arm.com>
5031 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5033 * config/aarch64/aarch64-protos.h
5034 (aarch64_sve_struct_memory_operand_p): Declare.
5035 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5036 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5037 (VPRED, vpred): Handle SVE structure modes.
5038 * config/aarch64/constraints.md (Utx): New constraint.
5039 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5040 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5041 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5042 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5043 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5044 structure modes. Split into pieces after RA.
5045 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5046 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5048 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5049 SVE structure modes.
5050 (aarch64_classify_address): Likewise.
5051 (sizetochar): Move earlier in file.
5052 (aarch64_print_operand): Handle SVE register lists.
5053 (aarch64_array_mode): New function.
5054 (aarch64_sve_struct_memory_operand_p): Likewise.
5055 (TARGET_ARRAY_MODE): Redefine.
5057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5058 Alan Hayward <alan.hayward@arm.com>
5059 David Sherwood <david.sherwood@arm.com>
5061 * target.def (array_mode): New target hook.
5062 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5063 * doc/tm.texi: Regenerate.
5064 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5065 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5066 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5068 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5072 Alan Hayward <alan.hayward@arm.com>
5073 David Sherwood <david.sherwood@arm.com>
5075 * fold-const.c (fold_binary_loc): Check the argument types
5076 rather than the result type when testing for a vector operation.
5078 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5080 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5081 * doc/tm.texi: Regenerate.
5083 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5084 Alan Hayward <alan.hayward@arm.com>
5085 David Sherwood <david.sherwood@arm.com>
5087 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5088 (sve): Document new AArch64 extension.
5089 * doc/md.texi (w): Extend the description of the AArch64
5090 constraint to include SVE vectors.
5091 (Upl, Upa): Document new AArch64 predicate constraints.
5092 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5094 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5095 (msve-vector-bits=): New option.
5096 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5097 SVE when these are disabled.
5098 (sve): New extension.
5099 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5100 modes. Adjust their number of units based on aarch64_sve_vg.
5101 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5102 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5103 aarch64_addr_query_type.
5104 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5105 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5106 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5107 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5108 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5109 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5110 (aarch64_simd_imm_zero_p): Delete.
5111 (aarch64_check_zero_based_sve_index_immediate): Declare.
5112 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5113 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5114 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5115 (aarch64_sve_float_mul_immediate_p): Likewise.
5116 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5118 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5119 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5120 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5121 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5122 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5123 (aarch64_regmode_natural_size): Likewise.
5124 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5125 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5127 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5128 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5129 for VG and the SVE predicate registers.
5130 (V_ALIASES): Add a "z"-prefixed alias.
5131 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5132 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5133 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5134 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5135 (REG_CLASS_NAMES): Add entries for them.
5136 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5137 and the predicate registers.
5138 (aarch64_sve_vg): Declare.
5139 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5140 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5141 (REGMODE_NATURAL_SIZE): Define.
5142 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5144 * config/aarch64/aarch64.c: Include cfgrtl.h.
5145 (simd_immediate_info): Add a constructor for series vectors,
5146 and an associated step field.
5147 (aarch64_sve_vg): New variable.
5148 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5149 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5150 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5151 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5152 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5153 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5154 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5155 (aarch64_get_mask_mode): New functions.
5156 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5157 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5158 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5159 predicate modes and predicate registers. Explicitly restrict
5160 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5161 to store a vector mode if it is recognized by
5162 aarch64_classify_vector_mode.
5163 (aarch64_regmode_natural_size): New function.
5164 (aarch64_hard_regno_caller_save_mode): Return the original mode
5166 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5167 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5168 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5169 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5171 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5172 does not overlap dest if the function is frame-related. Handle
5174 (aarch64_split_add_offset): New function.
5175 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5176 them aarch64_add_offset.
5177 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5178 and update call to aarch64_sub_sp.
5179 (aarch64_add_cfa_expression): New function.
5180 (aarch64_expand_prologue): Pass extra temporary registers to the
5181 functions above. Handle the case in which we need to emit new
5182 DW_CFA_expressions for registers that were originally saved
5183 relative to the stack pointer, but now have to be expressed
5184 relative to the frame pointer.
5185 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5187 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5188 IP0 and IP1 values for SVE frames.
5189 (aarch64_expand_vec_series): New function.
5190 (aarch64_expand_sve_widened_duplicate): Likewise.
5191 (aarch64_expand_sve_const_vector): Likewise.
5192 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5193 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5194 into the register, rather than emitting a SET directly.
5195 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5196 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5197 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5198 (offset_9bit_signed_scaled_p): New functions.
5199 (aarch64_replicate_bitmask_imm): New function.
5200 (aarch64_bitmask_imm): Use it.
5201 (aarch64_cannot_force_const_mem): Reject expressions involving
5202 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5203 (aarch64_classify_index): Handle SVE indices, by requiring
5204 a plain register index with a scale that matches the element size.
5205 (aarch64_classify_address): Handle SVE addresses. Assert that
5206 the mode of the address is VOIDmode or an integer mode.
5207 Update call to aarch64_classify_symbol.
5208 (aarch64_classify_symbolic_expression): Update call to
5209 aarch64_classify_symbol.
5210 (aarch64_const_vec_all_in_range_p): New function.
5211 (aarch64_print_vector_float_operand): Likewise.
5212 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5213 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5214 and the FP immediates 1.0 and 0.5.
5215 (aarch64_print_address_internal): Handle SVE addresses.
5216 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5217 (aarch64_regno_regclass): Handle predicate registers.
5218 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5220 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5221 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5222 (aarch64_convert_sve_vector_bits): New function.
5223 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5224 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5226 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5227 Handle SVE vector and predicate modes. Accept VL-based constants
5228 that need only one temporary register, and VL offsets that require
5229 no temporary registers.
5230 (aarch64_conditional_register_usage): Mark the predicate registers
5231 as fixed if SVE isn't available.
5232 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5233 Return true for SVE vector and predicate modes.
5234 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5235 rather than an unsigned int. Handle SVE modes.
5236 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5238 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5240 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5241 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5242 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5243 (aarch64_sve_float_mul_immediate_p): New functions.
5244 (aarch64_sve_valid_immediate): New function.
5245 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5246 Explicitly reject structure modes. Check for INDEX constants.
5247 Handle PTRUE and PFALSE constants.
5248 (aarch64_check_zero_based_sve_index_immediate): New function.
5249 (aarch64_simd_imm_zero_p): Delete.
5250 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5251 vector modes. Accept constants in the range of CNT[BHWD].
5252 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5253 ask for an Advanced SIMD mode.
5254 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5255 (aarch64_simd_vector_alignment): Handle SVE predicates.
5256 (aarch64_vectorize_preferred_vector_alignment): New function.
5257 (aarch64_simd_vector_alignment_reachable): Use it instead of
5259 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5260 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5262 (MAX_VECT_LEN): Delete.
5263 (expand_vec_perm_d): Add a vec_flags field.
5264 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5265 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5266 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5268 (aarch64_evpc_rev): Rename to...
5269 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5270 (aarch64_evpc_rev_global): New function.
5271 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5272 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5274 (aarch64_evpc_sve_tbl): New function.
5275 (aarch64_expand_vec_perm_const_1): Update after rename of
5276 aarch64_evpc_rev. Handle SVE permutes too, trying
5277 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5278 than aarch64_evpc_tbl.
5279 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5280 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5281 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5282 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5283 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5284 (aarch64_expand_sve_vcond): New functions.
5285 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5286 of aarch64_vector_mode_p.
5287 (aarch64_dwarf_poly_indeterminate_value): New function.
5288 (aarch64_compute_pressure_classes): Likewise.
5289 (aarch64_can_change_mode_class): Likewise.
5290 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5291 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5292 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5293 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5294 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5295 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5296 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5297 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5299 (Dn, Dl, Dr): Accept const as well as const_vector.
5300 (Dz): Likewise. Compare against CONST0_RTX.
5301 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5302 of "vector" where appropriate.
5303 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5304 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5305 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5306 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5307 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5308 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5309 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5310 (v_int_equiv): Extend to SVE modes.
5311 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5313 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5314 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5315 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5316 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5317 (SVE_COND_FP_CMP): New int iterators.
5318 (perm_hilo): Handle the new unpack unspecs.
5319 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5321 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5322 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5323 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5324 (aarch64_equality_operator, aarch64_constant_vector_operand)
5325 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5326 (aarch64_sve_nonimmediate_operand): Likewise.
5327 (aarch64_sve_general_operand): Likewise.
5328 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5329 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5330 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5331 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5332 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5333 (aarch64_sve_float_arith_immediate): Likewise.
5334 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5335 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5336 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5337 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5338 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5339 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5340 (aarch64_sve_float_arith_operand): Likewise.
5341 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5342 (aarch64_sve_float_mul_operand): Likewise.
5343 (aarch64_sve_vec_perm_operand): Likewise.
5344 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5345 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5346 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5347 as well as const_vector.
5348 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5349 in file. Use CONST0_RTX and CONSTM1_RTX.
5350 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5351 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5352 Use aarch64_simd_imm_zero.
5353 * config/aarch64/aarch64-sve.md: New file.
5354 * config/aarch64/aarch64.md: Include it.
5355 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5356 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5357 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5358 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5359 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5360 (sve): New attribute.
5361 (enabled): Disable instructions with the sve attribute unless
5363 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5364 aarch64_expand_mov_immediate.
5365 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5366 CNT[BHSD] immediates.
5367 (movti): Split CONST_POLY_INT moves into two halves.
5368 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5369 Split additions that need a temporary here if the destination
5370 is the stack pointer.
5371 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5372 (*add<mode>3_poly_1): New instruction.
5373 (set_clobber_cc): New expander.
5375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5377 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5378 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5379 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5380 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5381 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5382 Change innermode from fixed_mode_size to machine_mode.
5383 (simplify_subreg): Update call accordingly. Handle a constant-sized
5384 subreg of a variable-length CONST_VECTOR.
5386 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5387 Alan Hayward <alan.hayward@arm.com>
5388 David Sherwood <david.sherwood@arm.com>
5390 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5391 (add_offset_to_base): New function, split out from...
5392 (create_mem_ref): ...here. When handling a scale other than 1,
5393 check first whether the address is valid without the offset.
5394 Add it into the base if so, leaving the index and scale as-is.
5396 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5399 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5400 fold_for_warn before checking if arg2 is INTEGER_CST.
5402 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5404 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5405 (store_multiple_operation): Delete.
5406 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5407 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5408 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5409 guarded by TARGET_STRING.
5410 (rs6000_output_load_multiple): Delete.
5411 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5412 OPTION_MASK_STRING / TARGET_STRING handling.
5413 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5414 (const rs6000_opt_masks) <"string">: Change mask to 0.
5415 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5416 (MASK_STRING): Delete.
5417 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5419 (load_multiple): Delete.
5426 (store_multiple): Delete.
5433 (movmemsi_8reg): Delete.
5434 (corresponding unnamed define_insn): Delete.
5435 (movmemsi_6reg): Delete.
5436 (corresponding unnamed define_insn): Delete.
5437 (movmemsi_4reg): Delete.
5438 (corresponding unnamed define_insn): Delete.
5439 (movmemsi_2reg): Delete.
5440 (corresponding unnamed define_insn): Delete.
5441 (movmemsi_1reg): Delete.
5442 (corresponding unnamed define_insn): Delete.
5443 * config/rs6000/rs6000.opt (mno-string): New.
5444 (mstring): Replace by deprecation warning stub.
5445 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5447 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5449 * regrename.c (regrename_do_replace): If replacing the same
5450 reg multiple times, try to reuse last created gen_raw_REG.
5453 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5454 main to workaround a bug in GDB.
5456 2018-01-12 Tom de Vries <tom@codesourcery.com>
5459 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5461 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5463 PR rtl-optimization/80481
5464 * ira-color.c (get_cap_member): New function.
5465 (allocnos_conflict_by_live_ranges_p): Use it.
5466 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5467 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5469 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5472 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5473 (*saddl_se_1): Ditto.
5475 (*ssubl_se_1): Ditto.
5477 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5479 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5480 rather than wi::to_widest for DR_INITs.
5481 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5482 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5483 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5485 (vect_analyze_group_access_1): Note that here.
5487 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5489 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5490 polynomial type sizes.
5492 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5494 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5495 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5496 (gimple_add_tmp_var): Likewise.
5498 2018-01-12 Martin Liska <mliska@suse.cz>
5500 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5501 (gimple_alloc_sizes): Likewise.
5502 (dump_gimple_statistics): Use PRIu64 in printf format.
5503 * gimple.h: Change uint64_t to int.
5505 2018-01-12 Martin Liska <mliska@suse.cz>
5507 * tree-core.h: Use uint64_t instead of int.
5508 * tree.c (tree_node_counts): Likewise.
5509 (tree_node_sizes): Likewise.
5510 (dump_tree_statistics): Use PRIu64 in printf format.
5512 2018-01-12 Martin Liska <mliska@suse.cz>
5514 * Makefile.in: As qsort_chk is implemented in vec.c, add
5515 vec.o to linkage of gencfn-macros.
5516 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5517 passing the info to record_node_allocation_statistics.
5518 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5520 * ggc-common.c (struct ggc_usage): Add operator== and use
5521 it in operator< and compare function.
5522 * mem-stats.h (struct mem_usage): Likewise.
5523 * vec.c (struct vec_usage): Remove operator< and compare
5524 function. Can be simply inherited.
5526 2018-01-12 Martin Jambor <mjambor@suse.cz>
5529 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5530 * tree-ssa-math-opts.c: Include domwalk.h.
5531 (convert_mult_to_fma_1): New function.
5532 (fma_transformation_info): New type.
5533 (fma_deferring_state): Likewise.
5534 (cancel_fma_deferring): New function.
5535 (result_of_phi): Likewise.
5536 (last_fma_candidate_feeds_initial_phi): Likewise.
5537 (convert_mult_to_fma): Added deferring logic, split actual
5538 transformation to convert_mult_to_fma_1.
5539 (math_opts_dom_walker): New type.
5540 (math_opts_dom_walker::after_dom_children): New method, body moved
5541 here from pass_optimize_widening_mul::execute, added deferring logic
5543 (pass_optimize_widening_mul::execute): Moved most of code to
5544 math_opts_dom_walker::after_dom_children.
5545 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5546 * config/i386/i386.c (ix86_option_override_internal): Added
5547 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5549 2018-01-12 Richard Biener <rguenther@suse.de>
5552 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5553 inline instance vars.
5555 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5558 * config/rx/rx.c (rx_is_restricted_memory_address):
5561 2018-01-12 Richard Biener <rguenther@suse.de>
5563 PR tree-optimization/80846
5564 * target.def (split_reduction): New target hook.
5565 * targhooks.c (default_split_reduction): New function.
5566 * targhooks.h (default_split_reduction): Declare.
5567 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5568 target requests first reduce vectors by combining low and high
5570 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5571 (get_vectype_for_scalar_type_and_size): Export.
5572 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5573 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5574 * doc/tm.texi: Regenerate.
5575 * config/i386/i386.c (ix86_split_reduction): Implement
5576 TARGET_VECTORIZE_SPLIT_REDUCTION.
5578 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5581 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5582 in PIC mode except for TARGET_VXWORKS_RTP.
5583 * config/sparc/sparc.c: Include cfgrtl.h.
5584 (TARGET_INIT_PIC_REG): Define.
5585 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5586 (sparc_pic_register_p): New predicate.
5587 (sparc_legitimate_address_p): Use it.
5588 (sparc_legitimize_pic_address): Likewise.
5589 (sparc_delegitimize_address): Likewise.
5590 (sparc_mode_dependent_address_p): Likewise.
5591 (gen_load_pcrel_sym): Remove 4th parameter.
5592 (load_got_register): Adjust call to above. Remove obsolete stuff.
5593 (sparc_expand_prologue): Do not call load_got_register here.
5594 (sparc_flat_expand_prologue): Likewise.
5595 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5596 (sparc_use_pseudo_pic_reg): New function.
5597 (sparc_init_pic_reg): Likewise.
5598 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5599 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5601 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5603 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5604 Add item for branch_cost.
5606 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5608 PR rtl-optimization/83565
5609 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5610 not extend the result to a larger mode for rotate operations.
5611 (num_sign_bit_copies1): Likewise.
5613 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5616 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5618 Use values-Xc.o for -pedantic.
5619 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5621 2018-01-12 Martin Liska <mliska@suse.cz>
5624 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5626 (possible_polymorphic_call_targets): Use it.
5627 (ipa_devirt): Likewise.
5629 2018-01-12 Martin Liska <mliska@suse.cz>
5631 * profile-count.h (enum profile_quality): Use 0 as invalid
5632 enum value of profile_quality.
5634 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5636 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5637 -mext-string options.
5639 2018-01-12 Richard Biener <rguenther@suse.de>
5641 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5642 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5643 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5645 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5647 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5649 * configure.ac (--with-long-double-format): Add support for the
5650 configuration option to change the default long double format on
5652 * config.gcc (powerpc*-linux*-*): Likewise.
5653 * configure: Regenerate.
5654 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5655 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5656 used without modification.
5658 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5660 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5661 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5662 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5663 MISC_BUILTIN_SPEC_BARRIER.
5664 (rs6000_init_builtins): Likewise.
5665 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5667 (speculation_barrier): New define_insn.
5668 * doc/extend.texi: Document __builtin_speculation_barrier.
5670 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5673 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5674 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5675 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5677 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5678 integral modes instead of "ss" and "sd".
5679 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5680 vectors with 32-bit and 64-bit elements.
5681 (vecdupssescalarmodesuffix): New mode attribute.
5682 (vec_dup<mode>): Use it.
5684 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5687 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5688 frame if argument is passed on stack.
5690 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5693 * ree.c (combine_reaching_defs): Optimize also
5694 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5695 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5697 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5700 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5702 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5705 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5706 after they are computed.
5708 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5710 PR tree-optimization/83695
5711 * gimple-loop-linterchange.cc
5712 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5713 reset cached scev information after interchange.
5714 (pass_linterchange::execute): Remove call to scev_reset_htab.
5716 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5718 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5719 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5720 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5721 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5722 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5723 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5724 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5725 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5726 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5727 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5728 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5729 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5730 (V_lane_reg): Likewise.
5731 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5733 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5734 (vfmal_lane_low<mode>_intrinsic,
5735 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5736 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5737 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5738 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5739 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5740 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5742 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5744 * config/arm/arm-cpus.in (fp16fml): New feature.
5745 (ALL_SIMD): Add fp16fml.
5746 (armv8.2-a): Add fp16fml as an option.
5747 (armv8.3-a): Likewise.
5748 (armv8.4-a): Add fp16fml as part of fp16.
5749 * config/arm/arm.h (TARGET_FP16FML): Define.
5750 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5752 * config/arm/arm-modes.def (V2HF): Define.
5753 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5754 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5755 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5756 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5757 vfmsl_low, vfmsl_high): New set of builtins.
5758 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5759 (vfml_op): New code attribute.
5760 (VFMLHALVES): New int iterator.
5761 (VFML, VFMLSEL): New mode attributes.
5762 (V_reg): Define mapping for V2HF.
5763 (V_hi, V_lo): New mode attributes.
5764 (VF_constraint): Likewise.
5765 (vfml_half, vfml_half_selector): New int attributes.
5766 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5768 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5769 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5771 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5772 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5773 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5774 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5776 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5777 Document new effective target and option set.
5779 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5781 * config/arm/arm-cpus.in (armv8_4): New feature.
5782 (ARMv8_4a): New fgroup.
5783 (armv8.4-a): New arch.
5784 * config/arm/arm-tables.opt: Regenerate.
5785 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5786 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5787 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5788 Add matching rules for -march=armv8.4-a and extensions.
5789 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5791 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5794 * config/rx/rx.md (BW): New mode attribute.
5795 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5797 2018-01-11 Richard Biener <rguenther@suse.de>
5799 PR tree-optimization/83435
5800 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5801 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5802 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5804 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5805 Alan Hayward <alan.hayward@arm.com>
5806 David Sherwood <david.sherwood@arm.com>
5808 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5810 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5811 (aarch64_print_address_internal): Use it to check for a zero offset.
5813 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5814 Alan Hayward <alan.hayward@arm.com>
5815 David Sherwood <david.sherwood@arm.com>
5817 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5818 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5819 Return a poly_int64 rather than a HOST_WIDE_INT.
5820 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5821 rather than a HOST_WIDE_INT.
5822 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5823 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5824 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5825 final_offset from HOST_WIDE_INT to poly_int64.
5826 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5827 to_constant when getting the number of units in an Advanced SIMD
5829 (aarch64_builtin_vectorized_function): Check for a constant number
5831 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5833 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5834 attribute instead of GET_MODE_NUNITS.
5835 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5836 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5837 GET_MODE_SIZE for fixed-size registers.
5838 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5839 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5840 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5841 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5842 (aarch64_print_operand, aarch64_print_address_internal)
5843 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5844 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5845 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5846 Handle polynomial GET_MODE_SIZE.
5847 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5848 wider than SImode without modification.
5849 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5850 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5851 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5852 passing and returning SVE modes.
5853 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5854 rather than GEN_INT.
5855 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5856 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5857 (aarch64_allocate_and_probe_stack_space): Likewise.
5858 (aarch64_layout_frame): Cope with polynomial offsets.
5859 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5860 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5862 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5863 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5864 poly_int64 rather than a HOST_WIDE_INT.
5865 (aarch64_get_separate_components, aarch64_process_components)
5866 (aarch64_expand_prologue, aarch64_expand_epilogue)
5867 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5868 (aarch64_anchor_offset): New function, split out from...
5869 (aarch64_legitimize_address): ...here.
5870 (aarch64_builtin_vectorization_cost): Handle polynomial
5871 TYPE_VECTOR_SUBPARTS.
5872 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5874 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5875 number of elements from the PARALLEL rather than the mode.
5876 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5877 rather than GET_MODE_BITSIZE.
5878 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5879 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5880 (aarch64_expand_vec_perm_const_1): Handle polynomial
5881 d->perm.length () and d->perm elements.
5882 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5883 Apply to_constant to d->perm elements.
5884 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5885 polynomial CONST_VECTOR_NUNITS.
5886 (aarch64_move_pointer): Take amount as a poly_int64 rather
5888 (aarch64_progress_pointer): Avoid temporary variable.
5889 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5890 the mode attribute instead of GET_MODE.
5892 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5893 Alan Hayward <alan.hayward@arm.com>
5894 David Sherwood <david.sherwood@arm.com>
5896 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5897 x exists before using it.
5898 (aarch64_add_constant_internal): Rename to...
5899 (aarch64_add_offset_1): ...this. Replace regnum with separate
5900 src and dest rtxes. Handle the case in which they're different,
5901 including when the offset is zero. Replace scratchreg with an rtx.
5902 Use 2 additions if there is no spare register into which we can
5903 move a 16-bit constant.
5904 (aarch64_add_constant): Delete.
5905 (aarch64_add_offset): Replace reg with separate src and dest
5906 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5907 Use aarch64_add_offset_1.
5908 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5909 an rtx rather than an int. Take the delta as a poly_int64
5910 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5911 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5912 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5913 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5914 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5916 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5917 aarch64_add_constant.
5919 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5921 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5922 Use scalar_float_mode.
5924 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5926 * config/aarch64/aarch64-simd.md
5927 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5928 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5929 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5930 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5931 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5932 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5933 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5934 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5935 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5936 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5938 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5941 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5942 targ_options->x_arm_arch_string is non NULL.
5944 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5946 * config/aarch64/aarch64.h
5947 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5949 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5952 * expmed.c (emit_store_flag_force): Swap if const op0
5953 and change VOIDmode to mode of op0.
5955 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5957 PR rtl-optimization/83761
5958 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5959 than bytes to mode_for_size.
5961 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5964 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5965 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5968 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5971 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5972 when in layout mode.
5973 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5974 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5977 2018-01-10 Michael Collison <michael.collison@arm.com>
5979 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5980 * config/aarch64/aarch64-option-extension.def: Add
5981 AARCH64_OPT_EXTENSION of 'fp16fml'.
5982 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5983 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5984 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5985 * config/aarch64/constraints.md (Ui7): New constraint.
5986 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5987 (VFMLA_SEL_W): Ditto.
5990 (VFMLA16_LOW): New int iterator.
5991 (VFMLA16_HIGH): Ditto.
5992 (UNSPEC_FMLAL): New unspec.
5993 (UNSPEC_FMLSL): Ditto.
5994 (UNSPEC_FMLAL2): Ditto.
5995 (UNSPEC_FMLSL2): Ditto.
5996 (f16mac): New code attribute.
5997 * config/aarch64/aarch64-simd-builtins.def
5998 (aarch64_fmlal_lowv2sf): Ditto.
5999 (aarch64_fmlsl_lowv2sf): Ditto.
6000 (aarch64_fmlalq_lowv4sf): Ditto.
6001 (aarch64_fmlslq_lowv4sf): Ditto.
6002 (aarch64_fmlal_highv2sf): Ditto.
6003 (aarch64_fmlsl_highv2sf): Ditto.
6004 (aarch64_fmlalq_highv4sf): Ditto.
6005 (aarch64_fmlslq_highv4sf): Ditto.
6006 (aarch64_fmlal_lane_lowv2sf): Ditto.
6007 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6008 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6009 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6010 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6011 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6012 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6013 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6014 (aarch64_fmlal_lane_highv2sf): Ditto.
6015 (aarch64_fmlsl_lane_highv2sf): Ditto.
6016 (aarch64_fmlal_laneq_highv2sf): Ditto.
6017 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6018 (aarch64_fmlalq_lane_highv4sf): Ditto.
6019 (aarch64_fmlsl_lane_highv4sf): Ditto.
6020 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6021 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6022 * config/aarch64/aarch64-simd.md:
6023 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6024 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6025 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6026 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6027 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6028 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6029 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6030 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6031 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6032 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6033 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6034 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6035 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6036 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6037 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6038 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6039 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6040 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6041 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6042 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6043 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6044 (vfmlsl_low_u32): Ditto.
6045 (vfmlalq_low_u32): Ditto.
6046 (vfmlslq_low_u32): Ditto.
6047 (vfmlal_high_u32): Ditto.
6048 (vfmlsl_high_u32): Ditto.
6049 (vfmlalq_high_u32): Ditto.
6050 (vfmlslq_high_u32): Ditto.
6051 (vfmlal_lane_low_u32): Ditto.
6052 (vfmlsl_lane_low_u32): Ditto.
6053 (vfmlal_laneq_low_u32): Ditto.
6054 (vfmlsl_laneq_low_u32): Ditto.
6055 (vfmlalq_lane_low_u32): Ditto.
6056 (vfmlslq_lane_low_u32): Ditto.
6057 (vfmlalq_laneq_low_u32): Ditto.
6058 (vfmlslq_laneq_low_u32): Ditto.
6059 (vfmlal_lane_high_u32): Ditto.
6060 (vfmlsl_lane_high_u32): Ditto.
6061 (vfmlal_laneq_high_u32): Ditto.
6062 (vfmlsl_laneq_high_u32): Ditto.
6063 (vfmlalq_lane_high_u32): Ditto.
6064 (vfmlslq_lane_high_u32): Ditto.
6065 (vfmlalq_laneq_high_u32): Ditto.
6066 (vfmlslq_laneq_high_u32): Ditto.
6067 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6068 (AARCH64_FL_FOR_ARCH8_4): New.
6069 (AARCH64_ISA_F16FML): New ISA flag.
6070 (TARGET_F16FML): New feature flag for fp16fml.
6071 (doc/invoke.texi): Document new fp16fml option.
6073 2018-01-10 Michael Collison <michael.collison@arm.com>
6075 * config/aarch64/aarch64-builtins.c:
6076 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6077 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6078 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6079 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6080 (AARCH64_ISA_SHA3): New ISA flag.
6081 (TARGET_SHA3): New feature flag for sha3.
6082 * config/aarch64/iterators.md (sha512_op): New int attribute.
6083 (CRYPTO_SHA512): New int iterator.
6084 (UNSPEC_SHA512H): New unspec.
6085 (UNSPEC_SHA512H2): Ditto.
6086 (UNSPEC_SHA512SU0): Ditto.
6087 (UNSPEC_SHA512SU1): Ditto.
6088 * config/aarch64/aarch64-simd-builtins.def
6089 (aarch64_crypto_sha512hqv2di): New builtin.
6090 (aarch64_crypto_sha512h2qv2di): Ditto.
6091 (aarch64_crypto_sha512su0qv2di): Ditto.
6092 (aarch64_crypto_sha512su1qv2di): Ditto.
6093 (aarch64_eor3qv8hi): Ditto.
6094 (aarch64_rax1qv2di): Ditto.
6095 (aarch64_xarqv2di): Ditto.
6096 (aarch64_bcaxqv8hi): Ditto.
6097 * config/aarch64/aarch64-simd.md:
6098 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6099 (aarch64_crypto_sha512su0qv2di): Ditto.
6100 (aarch64_crypto_sha512su1qv2di): Ditto.
6101 (aarch64_eor3qv8hi): Ditto.
6102 (aarch64_rax1qv2di): Ditto.
6103 (aarch64_xarqv2di): Ditto.
6104 (aarch64_bcaxqv8hi): Ditto.
6105 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6106 (vsha512h2q_u64): Ditto.
6107 (vsha512su0q_u64): Ditto.
6108 (vsha512su1q_u64): Ditto.
6109 (veor3q_u16): Ditto.
6110 (vrax1q_u64): Ditto.
6112 (vbcaxq_u16): Ditto.
6113 * config/arm/types.md (crypto_sha512): New type attribute.
6114 (crypto_sha3): Ditto.
6115 (doc/invoke.texi): Document new sha3 option.
6117 2018-01-10 Michael Collison <michael.collison@arm.com>
6119 * config/aarch64/aarch64-builtins.c:
6120 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6121 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6122 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6123 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6124 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6125 (AARCH64_ISA_SM4): New ISA flag.
6126 (TARGET_SM4): New feature flag for sm4.
6127 * config/aarch64/aarch64-simd-builtins.def
6128 (aarch64_sm3ss1qv4si): Ditto.
6129 (aarch64_sm3tt1aq4si): Ditto.
6130 (aarch64_sm3tt1bq4si): Ditto.
6131 (aarch64_sm3tt2aq4si): Ditto.
6132 (aarch64_sm3tt2bq4si): Ditto.
6133 (aarch64_sm3partw1qv4si): Ditto.
6134 (aarch64_sm3partw2qv4si): Ditto.
6135 (aarch64_sm4eqv4si): Ditto.
6136 (aarch64_sm4ekeyqv4si): Ditto.
6137 * config/aarch64/aarch64-simd.md:
6138 (aarch64_sm3ss1qv4si): Ditto.
6139 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6140 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6141 (aarch64_sm4eqv4si): Ditto.
6142 (aarch64_sm4ekeyqv4si): Ditto.
6143 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6144 (sm3part_op): Ditto.
6145 (CRYPTO_SM3TT): Ditto.
6146 (CRYPTO_SM3PART): Ditto.
6147 (UNSPEC_SM3SS1): New unspec.
6148 (UNSPEC_SM3TT1A): Ditto.
6149 (UNSPEC_SM3TT1B): Ditto.
6150 (UNSPEC_SM3TT2A): Ditto.
6151 (UNSPEC_SM3TT2B): Ditto.
6152 (UNSPEC_SM3PARTW1): Ditto.
6153 (UNSPEC_SM3PARTW2): Ditto.
6154 (UNSPEC_SM4E): Ditto.
6155 (UNSPEC_SM4EKEY): Ditto.
6156 * config/aarch64/constraints.md (Ui2): New constraint.
6157 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6158 * config/arm/types.md (crypto_sm3): New type attribute.
6159 (crypto_sm4): Ditto.
6160 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6161 (vsm3tt1aq_u32): Ditto.
6162 (vsm3tt1bq_u32): Ditto.
6163 (vsm3tt2aq_u32): Ditto.
6164 (vsm3tt2bq_u32): Ditto.
6165 (vsm3partw1q_u32): Ditto.
6166 (vsm3partw2q_u32): Ditto.
6167 (vsm4eq_u32): Ditto.
6168 (vsm4ekeyq_u32): Ditto.
6169 (doc/invoke.texi): Document new sm4 option.
6171 2018-01-10 Michael Collison <michael.collison@arm.com>
6173 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6174 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6175 (AARCH64_FL_FOR_ARCH8_4): New.
6176 (AARCH64_FL_V8_4): New flag.
6177 (doc/invoke.texi): Document new armv8.4-a option.
6179 2018-01-10 Michael Collison <michael.collison@arm.com>
6181 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6182 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6183 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6184 * config/aarch64/aarch64-option-extension.def: Add
6185 AARCH64_OPT_EXTENSION of 'sha2'.
6186 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6187 (crypto): Disable sha2 and aes if crypto disabled.
6188 (crypto): Enable aes and sha2 if enabled.
6189 (simd): Disable sha2 and aes if simd disabled.
6190 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6192 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6193 (TARGET_SHA2): New feature flag for sha2.
6194 (TARGET_AES): New feature flag for aes.
6195 * config/aarch64/aarch64-simd.md:
6196 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6197 conditional on TARGET_AES.
6198 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6199 (aarch64_crypto_sha1hsi): Make pattern conditional
6201 (aarch64_crypto_sha1hv4si): Ditto.
6202 (aarch64_be_crypto_sha1hv4si): Ditto.
6203 (aarch64_crypto_sha1su1v4si): Ditto.
6204 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6205 (aarch64_crypto_sha1su0v4si): Ditto.
6206 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6207 (aarch64_crypto_sha256su0v4si): Ditto.
6208 (aarch64_crypto_sha256su1v4si): Ditto.
6209 (doc/invoke.texi): Document new aes and sha2 options.
6211 2018-01-10 Martin Sebor <msebor@redhat.com>
6213 PR tree-optimization/83781
6214 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6217 2018-01-11 Martin Sebor <msebor@gmail.com>
6218 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6220 PR tree-optimization/83501
6221 PR tree-optimization/81703
6223 * tree-ssa-strlen.c (get_string_cst): Rename...
6224 (get_string_len): ...to this. Handle global constants.
6225 (handle_char_store): Adjust.
6227 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6228 Jim Wilson <jimw@sifive.com>
6230 * config/riscv/riscv-protos.h (riscv_output_return): New.
6231 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6232 (riscv_attribute_table, riscv_output_return),
6233 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6234 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6235 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6237 (riscv_expand_prologue): Add early return for naked function.
6238 (riscv_expand_epilogue): Likewise.
6239 (riscv_function_ok_for_sibcall): Return false for naked function.
6240 (riscv_set_current_function): New.
6241 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6242 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6243 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6244 * doc/extend.texi (RISC-V Function Attributes): New.
6246 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6248 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6249 check for 128-bit long double before checking TCmode.
6250 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6251 128-bit long doubles before checking TFmode or TCmode.
6252 (FLOAT128_IBM_P): Likewise.
6254 2018-01-10 Martin Sebor <msebor@redhat.com>
6256 PR tree-optimization/83671
6257 * builtins.c (c_strlen): Unconditionally return zero for the empty
6259 Use -Warray-bounds for warnings.
6260 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6261 for non-constant array indices with COMPONENT_REF, arrays of
6262 arrays, and pointers to arrays.
6263 (gimple_fold_builtin_strlen): Determine and set length range for
6264 non-constant character arrays.
6266 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6269 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6272 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6274 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6276 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6279 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6280 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6281 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6282 indexed_or_indirect_operand predicate.
6283 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6284 (*vsx_le_perm_load_v8hi): Likewise.
6285 (*vsx_le_perm_load_v16qi): Likewise.
6286 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6287 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6288 (*vsx_le_perm_store_v8hi): Likewise.
6289 (*vsx_le_perm_store_v16qi): Likewise.
6290 (eight unnamed splitters): Likewise.
6292 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6294 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6295 * config/rs6000/emmintrin.h: Likewise.
6296 * config/rs6000/mmintrin.h: Likewise.
6297 * config/rs6000/xmmintrin.h: Likewise.
6299 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6302 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6304 * tree.c (tree_nop_conversion): Return true for location wrapper
6306 (maybe_wrap_with_location): New function.
6307 (selftest::check_strip_nops): New function.
6308 (selftest::test_location_wrappers): New function.
6309 (selftest::tree_c_tests): Call it.
6310 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6311 (maybe_wrap_with_location): New decl.
6312 (EXPR_LOCATION_WRAPPER_P): New macro.
6313 (location_wrapper_p): New inline function.
6314 (tree_strip_any_location_wrapper): New inline function.
6316 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6319 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6320 stack_realign_offset for the largest alignment of stack slot
6322 (ix86_find_max_used_stack_alignment): New function.
6323 (ix86_finalize_stack_frame_flags): Use it. Set
6324 max_used_stack_alignment if we don't realign stack.
6325 * config/i386/i386.h (machine_function): Add
6326 max_used_stack_alignment.
6328 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6330 * config/arm/arm.opt (-mbranch-cost): New option.
6331 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6334 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6337 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6338 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6340 2018-01-10 Richard Biener <rguenther@suse.de>
6343 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6344 early out so it also covers the case where we have a non-NULL
6347 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6349 PR tree-optimization/83753
6350 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6351 for non-strided grouped accesses if the number of elements is 1.
6353 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6356 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6357 * i386.h (TARGET_USE_GATHER): Define.
6358 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6360 2018-01-10 Martin Liska <mliska@suse.cz>
6363 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6364 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6366 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6367 CLEANUP_NO_PARTITIONING is not set.
6369 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6371 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6372 for vectors, as a partial revert of r254296.
6373 * rtl.h (const_vec_p): Delete.
6374 (const_vec_duplicate_p): Don't test for vector CONSTs.
6375 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6376 * expmed.c (make_tree): Likewise.
6379 * common.md (E, F): Use CONSTANT_P instead of checking for
6381 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6382 checking for CONST_VECTOR.
6384 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6387 * predict.c (force_edge_cold): Handle in more sane way edges
6390 2018-01-09 Carl Love <cel@us.ibm.com>
6392 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6394 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6395 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6396 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6397 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6398 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6399 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6400 * config/rs6000/rs6000-protos.h: Add extern defition for
6401 rs6000_generate_float2_double_code.
6402 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6404 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6405 (float2_v2df): Add define_expand.
6407 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6410 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6411 op_mode in the force_to_mode call.
6413 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6415 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6416 instead of checking each element individually.
6417 (aarch64_evpc_uzp): Likewise.
6418 (aarch64_evpc_zip): Likewise.
6419 (aarch64_evpc_ext): Likewise.
6420 (aarch64_evpc_rev): Likewise.
6421 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6422 instead of checking each element individually. Return true without
6424 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6425 whether all selected elements come from the same input, instead of
6426 checking each element individually. Remove calls to gen_rtx_REG,
6427 start_sequence and end_sequence and instead assert that no rtl is
6430 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6432 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6433 order of HIGH and CONST checks.
6435 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6437 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6438 if the destination isn't an SSA_NAME.
6440 2018-01-09 Richard Biener <rguenther@suse.de>
6442 PR tree-optimization/83668
6443 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6445 (canonicalize_loop_form): ... here, renamed from ...
6446 (canonicalize_loop_closed_ssa_form): ... this and amended to
6447 swap successor edges for loop exit blocks to make us use
6448 the RPO order we need for initial schedule generation.
6450 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6452 PR tree-optimization/64811
6453 * match.pd: When optimizing comparisons with Inf, avoid
6454 introducing or losing exceptions from comparisons with NaN.
6456 2018-01-09 Martin Liska <mliska@suse.cz>
6459 * asan.c (shadow_mem_size): Add gcc_assert.
6461 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6463 Don't save registers in main().
6466 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6467 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6468 * config/avr/avr.c (avr_set_current_function): Don't error if
6469 naked, OS_task or OS_main are specified at the same time.
6470 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6472 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6474 * common/config/avr/avr-common.c (avr_option_optimization_table):
6475 Switch on -mmain-is-OS_task for optimizing compilations.
6477 2018-01-09 Richard Biener <rguenther@suse.de>
6479 PR tree-optimization/83572
6480 * graphite.c: Include cfganal.h.
6481 (graphite_transform_loops): Connect infinite loops to exit
6482 and remove fake edges at the end.
6484 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6486 * ipa-inline.c (edge_badness): Revert accidental checkin.
6488 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6491 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6492 symbols; not inline clones.
6494 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6497 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6498 hard registers. Formatting fixes.
6500 PR preprocessor/83722
6501 * gcc.c (try_generate_repro): Pass
6502 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6503 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6506 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6507 Kito Cheng <kito.cheng@gmail.com>
6509 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6510 (riscv_leaf_function_p): Delete.
6511 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6513 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6515 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6517 (do_ifelse): New function.
6518 (do_isel): New function.
6519 (do_sub3): New function.
6520 (do_add3): New function.
6521 (do_load_mask_compare): New function.
6522 (do_overlap_load_compare): New function.
6523 (expand_compare_loop): New function.
6524 (expand_block_compare): Call expand_compare_loop() when appropriate.
6525 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6527 (-mblock-compare-inline-loop-limit): New option.
6529 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6532 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6533 Reverse order of second and third operands in first alternative.
6534 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6535 of first and second elements in UNSPEC_VPERMR vector.
6536 (altivec_expand_vec_perm_le): Likewise.
6538 2018-01-08 Jeff Law <law@redhat.com>
6540 PR rtl-optimizatin/81308
6541 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6542 (process_switch): If group_case_labels makes a change, then set
6544 (pass_convert_switch::execute): If a switch is converted, then
6545 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6547 PR rtl-optimization/81308
6548 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6551 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6553 PR target/83663 - Revert r255946
6554 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6555 generation for cases where splatting a value is not useful.
6556 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6557 across a vec_duplicate and a paradoxical subreg forming a vector
6558 mode to a vec_concat.
6560 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6562 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6563 -march=armv8.3-a variants.
6564 * config/arm/t-multilib: Likewise.
6565 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6567 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6569 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6571 (cceq_ior_compare_complement): Give it a name so I can use it, and
6572 change boolean_or_operator predicate to boolean_operator so it can
6573 be used to generate a crand.
6574 (eqne): New code iterator.
6575 (bd/bd_neg): New code_attrs.
6576 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6577 a single define_insn.
6578 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6579 decrement (bdnzt/bdnzf/bdzt/bdzf).
6580 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6581 with the new names of the branch decrement patterns, and added the
6582 names of the branch decrement conditional patterns.
6584 2018-01-08 Richard Biener <rguenther@suse.de>
6586 PR tree-optimization/83563
6587 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6590 2018-01-08 Richard Biener <rguenther@suse.de>
6593 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6595 2018-01-08 Richard Biener <rguenther@suse.de>
6597 PR tree-optimization/83685
6598 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6599 references to abnormals.
6601 2018-01-08 Richard Biener <rguenther@suse.de>
6604 * dwarf2out.c (output_indirect_strings): Handle empty
6605 skeleton_debug_str_hash.
6606 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6608 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6610 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6611 (emit_store_direct): Likewise.
6612 (arc_trampoline_adjust_address): Likewise.
6613 (arc_asm_trampoline_template): New function.
6614 (arc_initialize_trampoline): Use asm_trampoline_template.
6615 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6616 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6617 * config/arc/arc.md (flush_icache): Delete pattern.
6619 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6621 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6622 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6625 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6628 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6629 by not USED_FOR_TARGET.
6630 (make_pass_resolve_sw_modes): Likewise.
6632 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6634 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6637 2018-01-08 Richard Biener <rguenther@suse.de>
6640 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6642 2018-01-08 Richard Biener <rguenther@suse.de>
6645 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6647 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6650 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6651 basic blocks with a small number of successors.
6652 (convert_control_dep_chain_into_preds): Improve handling of
6654 (dump_predicates): Split apart into...
6655 (dump_pred_chain): ...here...
6656 (dump_pred_info): ...and here.
6657 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6658 (can_chain_union_be_invalidated_p): Improve check for invalidation
6660 (uninit_uses_cannot_happen): Avoid unnecessary if
6661 convert_control_dep_chain_into_preds yielded nothing.
6663 2018-01-06 Martin Sebor <msebor@redhat.com>
6665 PR tree-optimization/83640
6666 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6667 subtracting negative offset from size.
6668 (builtin_access::overlap): Adjust offset bounds of the access to fall
6669 within the size of the object if possible.
6671 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6673 PR rtl-optimization/83699
6674 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6675 extract_bit_field_as_subreg to cases in which the extracted
6676 value is also a vector.
6678 * lra-constraints.c (process_alt_operands): Test for the equivalence
6679 substitutions when detecting a possible reload cycle.
6681 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6684 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6685 by default if flag_selective_schedling{,2}. Formatting fixes.
6687 PR rtl-optimization/83682
6688 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6689 if it has non-VECTOR_MODE element mode.
6690 (vec_duplicate_p): Likewise.
6693 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6694 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6696 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6699 * config/i386/i386-builtin.def
6700 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6701 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6702 Require also OPTION_MASK_ISA_AVX512F in addition to
6703 OPTION_MASK_ISA_GFNI.
6704 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6705 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6706 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6707 to OPTION_MASK_ISA_GFNI.
6708 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6709 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6710 OPTION_MASK_ISA_AVX512BW.
6711 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6712 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6713 addition to OPTION_MASK_ISA_GFNI.
6714 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6715 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6716 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6717 to OPTION_MASK_ISA_GFNI.
6718 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6719 a requirement for all ISAs rather than any of them with a few
6721 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6723 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6724 bitmasks to be enabled with 3 exceptions, instead of requiring any
6725 enabled ISA with lots of exceptions.
6726 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6727 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6728 Change avx512bw in isa attribute to avx512f.
6729 * config/i386/sgxintrin.h: Add license boilerplate.
6730 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6731 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6732 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6733 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6735 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6736 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6737 temporarily sse2 rather than sse if not enabled already.
6740 * config/i386/sse.md (VI248_VLBW): Rename to ...
6741 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6742 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6743 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6744 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6745 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6746 mode iterator instead of VI248_VLBW.
6748 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6750 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6751 (record_modified): Skip clobbers; add debug output.
6752 (param_change_prob): Use sreal frequencies.
6754 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6756 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6757 punt for user-aligned variables.
6759 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6761 * tree-chrec.c (chrec_contains_symbols): Return true for
6764 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6767 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6768 of (x|y) == x for BICS pattern.
6770 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6772 PR tree-optimization/83605
6773 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6774 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6777 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6779 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6780 * config/epiphany/rtems.h: New file.
6782 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6783 Uros Bizjak <ubizjak@gmail.com>
6786 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6787 QIreg_operand instead of register_operand predicate.
6788 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6789 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6790 comments instead of -fmitigate[-_]rop.
6792 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6795 * cgraphunit.c (symbol_table::compile): Switch to text_section
6796 before calling assembly_start debug hook.
6797 * run-rtl-passes.c (run_rtl_passes): Likewise.
6800 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6802 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6803 range_int_cst_p rather than !symbolic_range_p before calling
6804 extract_range_from_multiplicative_op_1.
6806 2018-01-04 Jeff Law <law@redhat.com>
6808 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6809 redundant test in assertion.
6811 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6813 * doc/rtl.texi: Document machine_mode wrapper classes.
6815 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6817 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6820 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6822 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6823 the VEC_PERM_EXPR fold to fail.
6825 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6828 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6829 to switched_sections.
6831 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6834 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6837 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6840 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6841 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6843 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6846 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6847 is BLKmode and bitpos not zero or mode change is needed.
6849 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6852 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6855 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6858 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6859 instead of MULT rtx. Update all corresponding splitters.
6861 (*ssub<modesuffix>): Ditto.
6863 (*cmp_sadd_di): Update split patterns.
6864 (*cmp_sadd_si): Ditto.
6865 (*cmp_sadd_sidi): Ditto.
6866 (*cmp_ssub_di): Ditto.
6867 (*cmp_ssub_si): Ditto.
6868 (*cmp_ssub_sidi): Ditto.
6869 * config/alpha/predicates.md (const23_operand): New predicate.
6870 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6871 Look for ASHIFT, not MULT inner operand.
6872 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6874 2018-01-04 Martin Liska <mliska@suse.cz>
6876 PR gcov-profile/83669
6877 * gcov.c (output_intermediate_file): Add version to intermediate
6879 * doc/gcov.texi: Document new field 'version' in intermediate
6880 file format. Fix location of '-k' option of gcov command.
6882 2018-01-04 Martin Liska <mliska@suse.cz>
6885 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6887 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6889 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6891 2018-01-03 Martin Sebor <msebor@redhat.com>
6893 PR tree-optimization/83655
6894 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6895 checking calls with invalid arguments.
6897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6899 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6900 (vectorizable_mask_load_store): Delete.
6901 (vectorizable_call): Return false for masked loads and stores.
6902 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6903 instead of gimple_assign_rhs1.
6904 (vectorizable_load): Handle IFN_MASK_LOAD.
6905 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6909 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6911 (vectorizable_mask_load_store): ...here.
6912 (vectorizable_load): ...and here.
6914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6916 * tree-vect-stmts.c (vect_build_all_ones_mask)
6917 (vect_build_zero_merge_argument): New functions, split out from...
6918 (vectorizable_load): ...here.
6920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6922 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6924 (vectorizable_mask_load_store): ...here.
6925 (vectorizable_store): ...and here.
6927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6929 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6931 (vectorizable_mask_load_store): ...here.
6933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6935 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6936 (vect_model_store_cost): Take a vec_load_store_type instead of a
6938 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6939 (vect_model_store_cost): Take a vec_load_store_type instead of a
6941 (vectorizable_mask_load_store): Update accordingly.
6942 (vectorizable_store): Likewise.
6943 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6947 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6948 IFN_MASK_LOAD calls here rather than...
6949 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6952 Alan Hayward <alan.hayward@arm.com>
6953 David Sherwood <david.sherwood@arm.com>
6955 * expmed.c (extract_bit_field_1): For vector extracts,
6956 fall back to extract_bit_field_as_subreg if vec_extract
6959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6960 Alan Hayward <alan.hayward@arm.com>
6961 David Sherwood <david.sherwood@arm.com>
6963 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6964 they are variable or constant sized.
6965 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6966 slots for constant-sized data.
6968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6969 Alan Hayward <alan.hayward@arm.com>
6970 David Sherwood <david.sherwood@arm.com>
6972 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6973 handling COND_EXPRs with boolean comparisons, try to find a better
6974 basis for the mask type than the boolean itself.
6976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6978 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6979 is calculated and how it can be overridden.
6980 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6981 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6983 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6987 Alan Hayward <alan.hayward@arm.com>
6988 David Sherwood <david.sherwood@arm.com>
6990 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6991 Remove the mode argument.
6992 (aarch64_simd_valid_immediate): Remove the mode and inverse
6994 * config/aarch64/iterators.md (bitsize): New iterator.
6995 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6996 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6997 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6998 aarch64_simd_valid_immediate.
6999 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7000 (aarch64_reg_or_bic_imm): Likewise.
7001 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7002 with an insn_type enum and msl with a modifier_type enum.
7003 Replace element_width with a scalar_mode. Change the shift
7004 to unsigned int. Add constructors for scalar_float_mode and
7005 scalar_int_mode elements.
7006 (aarch64_vect_float_const_representable_p): Delete.
7007 (aarch64_can_const_movi_rtx_p)
7008 (aarch64_simd_scalar_immediate_valid_for_move)
7009 (aarch64_simd_make_constant): Update call to
7010 aarch64_simd_valid_immediate.
7011 (aarch64_advsimd_valid_immediate_hs): New function.
7012 (aarch64_advsimd_valid_immediate): Likewise.
7013 (aarch64_simd_valid_immediate): Remove mode and inverse
7014 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7015 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7016 and aarch64_float_const_representable_p on the result.
7017 (aarch64_output_simd_mov_immediate): Remove mode argument.
7018 Update call to aarch64_simd_valid_immediate and use of
7019 simd_immediate_info.
7020 (aarch64_output_scalar_simd_mov_immediate): Update call
7023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7024 Alan Hayward <alan.hayward@arm.com>
7025 David Sherwood <david.sherwood@arm.com>
7027 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7028 (mode_nunits): Likewise CONST_MODE_NUNITS.
7029 * machmode.def (ADJUST_NUNITS): Document.
7030 * genmodes.c (mode_data::need_nunits_adj): New field.
7031 (blank_mode): Update accordingly.
7032 (adj_nunits): New variable.
7033 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7035 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7036 listed in adj_nunits.
7037 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7038 listed in adj_nunits. Don't emit case statements for such modes.
7039 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7040 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7041 nothing if adj_nunits is nonnull.
7042 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7043 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7044 (emit_mode_fbit): Update use of print_maybe_const_decl.
7045 (emit_move_size): Likewise. Treat the array as non-const
7047 (emit_mode_adjustments): Handle adj_nunits.
7049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7051 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7052 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7053 (VECTOR_MODES): Use it.
7054 (make_vector_modes): Take the prefix as an argument.
7056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7057 Alan Hayward <alan.hayward@arm.com>
7058 David Sherwood <david.sherwood@arm.com>
7060 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7061 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7062 for MODE_VECTOR_BOOL.
7063 * machmode.def (VECTOR_BOOL_MODE): Document.
7064 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7065 (make_vector_bool_mode): New function.
7066 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7068 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7069 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7071 * stor-layout.c (int_mode_for_mode): Likewise.
7072 * tree.c (build_vector_type_for_mode): Likewise.
7073 * varasm.c (output_constant_pool_2): Likewise.
7074 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7075 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7076 for MODE_VECTOR_BOOL.
7077 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7078 of mode class checks.
7079 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7080 instead of a list of mode class checks.
7081 (expand_vector_scalar_condition): Likewise.
7082 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7085 Alan Hayward <alan.hayward@arm.com>
7086 David Sherwood <david.sherwood@arm.com>
7088 * machmode.h (mode_size): Change from unsigned short to
7090 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7091 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7092 or if measurement_type is not polynomial.
7093 (fixed_size_mode::includes_p): Check for constant-sized modes.
7094 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7095 return a poly_uint16 rather than an unsigned short.
7096 (emit_mode_size): Change the type of mode_size from unsigned short
7097 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7098 (emit_mode_adjustments): Cope with polynomial vector sizes.
7099 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7101 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7103 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7104 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7105 * caller-save.c (setup_save_areas): Likewise.
7106 (replace_reg_with_saved_mem): Likewise.
7107 * calls.c (emit_library_call_value_1): Likewise.
7108 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7109 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7110 (gen_lowpart_for_combine): Likewise.
7111 * convert.c (convert_to_integer_1): Likewise.
7112 * cse.c (equiv_constant, cse_insn): Likewise.
7113 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7114 (cselib_subst_to_values): Likewise.
7115 * dce.c (word_dce_process_block): Likewise.
7116 * df-problems.c (df_word_lr_mark_ref): Likewise.
7117 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7118 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7119 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7120 (rtl_for_decl_location): Likewise.
7121 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7122 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7123 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7124 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7125 (expand_expr_real_1): Likewise.
7126 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7127 (pad_below): Likewise.
7128 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7129 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7130 * ira.c (get_subreg_tracking_sizes): Likewise.
7131 * ira-build.c (ira_create_allocno_objects): Likewise.
7132 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7133 (ira_sort_regnos_for_alter_reg): Likewise.
7134 * ira-costs.c (record_operand_costs): Likewise.
7135 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7136 (resolve_simple_move): Likewise.
7137 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7138 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7139 (lra_constraints): Likewise.
7140 (CONST_POOL_OK_P): Reject variable-sized modes.
7141 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7142 (add_pseudo_to_slot, lra_spill): Likewise.
7143 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7144 * optabs-query.c (get_best_extraction_insn): Likewise.
7145 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7146 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7147 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7148 * recog.c (offsettable_address_addr_space_p): Likewise.
7149 * regcprop.c (maybe_mode_change): Likewise.
7150 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7151 * regrename.c (build_def_use): Likewise.
7152 * regstat.c (dump_reg_info): Likewise.
7153 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7154 (find_reloads, find_reloads_subreg_address): Likewise.
7155 * reload1.c (eliminate_regs_1): Likewise.
7156 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7157 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7158 (simplify_binary_operation_1, simplify_subreg): Likewise.
7159 * targhooks.c (default_function_arg_padding): Likewise.
7160 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7161 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7162 (verify_gimple_assign_ternary): Likewise.
7163 * tree-inline.c (estimate_move_cost): Likewise.
7164 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7165 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7166 (get_address_cost_ainc): Likewise.
7167 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7168 (vect_supportable_dr_alignment): Likewise.
7169 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7170 (vectorizable_reduction): Likewise.
7171 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7172 (vectorizable_operation, vectorizable_load): Likewise.
7173 * tree.c (build_same_sized_truth_vector_type): Likewise.
7174 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7175 * var-tracking.c (emit_note_insn_var_location): Likewise.
7176 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7177 (ADDR_VEC_ALIGN): Likewise.
7179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7180 Alan Hayward <alan.hayward@arm.com>
7181 David Sherwood <david.sherwood@arm.com>
7183 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7185 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7186 or if measurement_type is polynomial.
7187 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7188 * combine.c (make_extraction): Likewise.
7189 * dse.c (find_shift_sequence): Likewise.
7190 * dwarf2out.c (mem_loc_descriptor): Likewise.
7191 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7192 (extract_bit_field, extract_low_bits): Likewise.
7193 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7194 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7195 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7196 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7197 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7198 * reload.c (find_reloads): Likewise.
7199 * reload1.c (alter_reg): Likewise.
7200 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7201 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7202 * tree-if-conv.c (predicate_mem_writes): Likewise.
7203 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7204 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7205 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7206 * valtrack.c (dead_debug_insert_temp): Likewise.
7207 * varasm.c (mergeable_constant_section): Likewise.
7208 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7211 Alan Hayward <alan.hayward@arm.com>
7212 David Sherwood <david.sherwood@arm.com>
7214 * expr.c (expand_assignment): Cope with polynomial mode sizes
7215 when assigning to a CONCAT.
7217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7218 Alan Hayward <alan.hayward@arm.com>
7219 David Sherwood <david.sherwood@arm.com>
7221 * machmode.h (mode_precision): Change from unsigned short to
7223 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7225 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7226 or if measurement_type is not polynomial.
7227 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7228 in which the mode is already known to be a scalar_int_mode.
7229 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7230 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7232 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7233 for GET_MODE_PRECISION.
7234 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7235 for GET_MODE_PRECISION.
7236 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7238 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7239 (expand_field_assignment, make_extraction): Likewise.
7240 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7241 (get_last_value): Likewise.
7242 * convert.c (convert_to_integer_1): Likewise.
7243 * cse.c (cse_insn): Likewise.
7244 * expr.c (expand_expr_real_1): Likewise.
7245 * lra-constraints.c (simplify_operand_subreg): Likewise.
7246 * optabs-query.c (can_atomic_load_p): Likewise.
7247 * optabs.c (expand_atomic_load): Likewise.
7248 (expand_atomic_store): Likewise.
7249 * ree.c (combine_reaching_defs): Likewise.
7250 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7251 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7252 * tree.h (type_has_mode_precision_p): Likewise.
7253 * ubsan.c (instrument_si_overflow): Likewise.
7255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7256 Alan Hayward <alan.hayward@arm.com>
7257 David Sherwood <david.sherwood@arm.com>
7259 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7260 polynomial numbers of units.
7261 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7262 (valid_vector_subparts_p): New function.
7263 (build_vector_type): Remove temporary shim and take the number
7264 of units as a poly_uint64 rather than an int.
7265 (build_opaque_vector_type): Take the number of units as a
7266 poly_uint64 rather than an int.
7267 * tree.c (build_vector_from_ctor): Handle polynomial
7268 TYPE_VECTOR_SUBPARTS.
7269 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7270 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7271 (build_vector_from_val): If the number of units is variable,
7272 use build_vec_duplicate_cst for constant operands and
7273 VEC_DUPLICATE_EXPR otherwise.
7274 (make_vector_type): Remove temporary is_constant ().
7275 (build_vector_type, build_opaque_vector_type): Take the number of
7276 units as a poly_uint64 rather than an int.
7277 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7279 * cfgexpand.c (expand_debug_expr): Likewise.
7280 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7281 (store_constructor, expand_expr_real_1): Likewise.
7282 (const_scalar_mask_from_tree): Likewise.
7283 * fold-const-call.c (fold_const_reduction): Likewise.
7284 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7285 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7286 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7287 (fold_relational_const): Likewise.
7288 (native_interpret_vector): Likewise. Change the size from an
7289 int to an unsigned int.
7290 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7291 TYPE_VECTOR_SUBPARTS.
7292 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7293 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7294 duplicating a non-constant operand into a variable-length vector.
7295 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7296 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7297 * ipa-icf.c (sem_variable::equals): Likewise.
7298 * match.pd: Likewise.
7299 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7300 * print-tree.c (print_node): Likewise.
7301 * stor-layout.c (layout_type): Likewise.
7302 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7303 * tree-cfg.c (verify_gimple_comparison): Likewise.
7304 (verify_gimple_assign_binary): Likewise.
7305 (verify_gimple_assign_ternary): Likewise.
7306 (verify_gimple_assign_single): Likewise.
7307 * tree-pretty-print.c (dump_generic_node): Likewise.
7308 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7309 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7310 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7311 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7312 (vect_shift_permute_load_chain): Likewise.
7313 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7314 (expand_vector_condition, optimize_vector_constructor): Likewise.
7315 (lower_vec_perm, get_compute_type): Likewise.
7316 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7317 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7318 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7319 (vect_recog_mask_conversion_pattern): Likewise.
7320 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7321 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7322 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7323 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7324 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7325 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7326 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7327 (supportable_widening_operation): Likewise.
7328 (supportable_narrowing_operation): Likewise.
7329 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7331 * varasm.c (output_constant): Likewise.
7333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7334 Alan Hayward <alan.hayward@arm.com>
7335 David Sherwood <david.sherwood@arm.com>
7337 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7338 so that both the length == 3 and length != 3 cases set up their
7339 own permute vectors. Add comments explaining why we know the
7340 number of elements is constant.
7341 (vect_permute_load_chain): Likewise.
7343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7344 Alan Hayward <alan.hayward@arm.com>
7345 David Sherwood <david.sherwood@arm.com>
7347 * machmode.h (mode_nunits): Change from unsigned char to
7349 (ONLY_FIXED_SIZE_MODES): New macro.
7350 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7351 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7352 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7354 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7355 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7356 or if measurement_type is not polynomial.
7357 * genmodes.c (ZERO_COEFFS): New macro.
7358 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7360 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7361 Use ZERO_COEFFS when emitting initializers.
7362 * data-streamer.h (bp_pack_poly_value): New function.
7363 (bp_unpack_poly_value): Likewise.
7364 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7365 for GET_MODE_NUNITS.
7366 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7367 for GET_MODE_NUNITS.
7368 * tree.c (make_vector_type): Remove temporary shim and make
7369 the real function take the number of units as a poly_uint64
7371 (build_vector_type_for_mode): Handle polynomial nunits.
7372 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7373 * emit-rtl.c (const_vec_series_p_1): Likewise.
7374 (gen_rtx_CONST_VECTOR): Likewise.
7375 * fold-const.c (test_vec_duplicate_folding): Likewise.
7376 * genrecog.c (validate_pattern): Likewise.
7377 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7378 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7379 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7380 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7381 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7382 * rtlanal.c (subreg_get_info): Likewise.
7383 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7384 (vect_grouped_load_supported): Likewise.
7385 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7386 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7387 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7388 (simplify_const_unary_operation, simplify_binary_operation_1)
7389 (simplify_const_binary_operation, simplify_ternary_operation)
7390 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7391 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7392 instead of CONST_VECTOR_NUNITS.
7393 * varasm.c (output_constant_pool_2): Likewise.
7394 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7395 explicit-encoded elements in the XVEC for variable-length vectors.
7397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7399 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7402 Alan Hayward <alan.hayward@arm.com>
7403 David Sherwood <david.sherwood@arm.com>
7405 * coretypes.h (fixed_size_mode): Declare.
7406 (fixed_size_mode_pod): New typedef.
7407 * builtins.h (target_builtins::x_apply_args_mode)
7408 (target_builtins::x_apply_result_mode): Change type to
7409 fixed_size_mode_pod.
7410 * builtins.c (apply_args_size, apply_result_size, result_vector)
7411 (expand_builtin_apply_args_1, expand_builtin_apply)
7412 (expand_builtin_return): Update accordingly.
7414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7416 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7417 * cselib.c (cselib_hash_rtx): Likewise.
7418 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7419 CONST_VECTOR encoding.
7421 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7422 Jeff Law <law@redhat.com>
7425 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7426 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7427 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7428 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7431 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7432 explicitly probe *sp in a noreturn function if there were any callee
7433 register saves or frame pointer is needed.
7435 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7438 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7439 BLKmode for ternary, binary or unary expressions.
7442 * var-tracking.c (delete_vta_debug_insn): New inline function.
7443 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7444 insns from get_insns () to NULL instead of each bb separately.
7445 Use delete_vta_debug_insn. No longer static.
7446 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7447 delete_vta_debug_insns callers.
7448 * rtl.h (delete_vta_debug_insns): Declare.
7449 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7450 instead of variable_tracking_main.
7452 2018-01-03 Martin Sebor <msebor@redhat.com>
7454 PR tree-optimization/83603
7455 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7456 arguments past the endof the argument list in functions declared
7457 without a prototype.
7458 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7459 Avoid checking when arguments are null.
7461 2018-01-03 Martin Sebor <msebor@redhat.com>
7464 * doc/extend.texi (attribute const): Fix a typo.
7465 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7466 issuing -Wsuggest-attribute for void functions.
7468 2018-01-03 Martin Sebor <msebor@redhat.com>
7470 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7471 offset_int::from instead of wide_int::to_shwi.
7472 (maybe_diag_overlap): Remove assertion.
7473 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7474 * gimple-ssa-sprintf.c (format_directive): Same.
7475 (parse_directive): Same.
7476 (sprintf_dom_walker::compute_format_length): Same.
7477 (try_substitute_return_value): Same.
7479 2018-01-03 Jeff Law <law@redhat.com>
7482 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7483 non-constant residual for zero at runtime and avoid probing in
7484 that case. Reorganize code for trailing problem to mirror handling
7487 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7489 PR tree-optimization/83501
7490 * tree-ssa-strlen.c (get_string_cst): New.
7491 (handle_char_store): Call get_string_cst.
7493 2018-01-03 Martin Liska <mliska@suse.cz>
7495 PR tree-optimization/83593
7496 * tree-ssa-strlen.c: Include tree-cfg.h.
7497 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7498 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7499 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7501 (strlen_dom_walker::before_dom_children): Call
7502 gimple_purge_dead_eh_edges. Dump tranformation with details
7504 (strlen_dom_walker::before_dom_children): Update call by adding
7505 new argument cleanup_eh.
7506 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7508 2018-01-03 Martin Liska <mliska@suse.cz>
7511 * cif-code.def (VARIADIC_THUNK): New enum value.
7512 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7515 2018-01-03 Jan Beulich <jbeulich@suse.com>
7517 * sse.md (mov<mode>_internal): Tighten condition for when to use
7518 vmovdqu<ssescalarsize> for TI and OI modes.
7520 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7522 Update copyright years.
7524 2018-01-03 Martin Liska <mliska@suse.cz>
7527 * ipa-visibility.c (function_and_variable_visibility): Skip
7528 functions with noipa attribure.
7530 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7532 * gcc.c (process_command): Update copyright notice dates.
7533 * gcov-dump.c (print_version): Ditto.
7534 * gcov.c (print_version): Ditto.
7535 * gcov-tool.c (print_version): Ditto.
7536 * gengtype.c (create_file): Ditto.
7537 * doc/cpp.texi: Bump @copying's copyright year.
7538 * doc/cppinternals.texi: Ditto.
7539 * doc/gcc.texi: Ditto.
7540 * doc/gccint.texi: Ditto.
7541 * doc/gcov.texi: Ditto.
7542 * doc/install.texi: Ditto.
7543 * doc/invoke.texi: Ditto.
7545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7547 * vector-builder.h (vector_builder::m_full_nelts): Change from
7548 unsigned int to poly_uint64.
7549 (vector_builder::full_nelts): Update prototype accordingly.
7550 (vector_builder::new_vector): Likewise.
7551 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7552 (vector_builder::operator ==): Likewise.
7553 (vector_builder::finalize): Likewise.
7554 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7555 Take the number of elements as a poly_uint64 rather than an
7557 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7558 from unsigned int to poly_uint64.
7559 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7560 (vec_perm_indices::new_vector): Likewise.
7561 (vec_perm_indices::length): Likewise.
7562 (vec_perm_indices::nelts_per_input): Likewise.
7563 (vec_perm_indices::input_nelts): Likewise.
7564 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7565 number of elements per input as a poly_uint64 rather than an
7566 unsigned int. Use the original encoding for variable-length
7567 vectors, rather than clamping each individual element.
7568 For the second and subsequent elements in each pattern,
7569 clamp the step and base before clamping their sum.
7570 (vec_perm_indices::series_p): Handle polynomial element counts.
7571 (vec_perm_indices::all_in_range_p): Likewise.
7572 (vec_perm_indices_to_tree): Likewise.
7573 (vec_perm_indices_to_rtx): Likewise.
7574 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7575 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7576 (tree_vector_builder::new_binary_operation): Handle polynomial
7577 element counts. Return false if we need to know the number
7578 of elements at compile time.
7579 * fold-const.c (fold_vec_perm): Punt if the number of elements
7580 isn't known at compile time.
7582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7584 * vec-perm-indices.h (vec_perm_builder): Change element type
7585 from HOST_WIDE_INT to poly_int64.
7586 (vec_perm_indices::element_type): Update accordingly.
7587 (vec_perm_indices::clamp): Handle polynomial element_types.
7588 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7589 (vec_perm_indices::all_in_range_p): Likewise.
7590 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7592 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7593 polynomial vec_perm_indices element types.
7594 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7595 * fold-const.c (fold_vec_perm): Likewise.
7596 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7597 * tree-vect-generic.c (lower_vec_perm): Likewise.
7598 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7599 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7600 element type to HOST_WIDE_INT.
7602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7603 Alan Hayward <alan.hayward@arm.com>
7604 David Sherwood <david.sherwood@arm.com>
7606 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7607 rather than an int. Use plus_constant.
7608 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7609 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7612 Alan Hayward <alan.hayward@arm.com>
7613 David Sherwood <david.sherwood@arm.com>
7615 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7616 a HOST_WIDE_INT to a poly_int64.
7618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7619 Alan Hayward <alan.hayward@arm.com>
7620 David Sherwood <david.sherwood@arm.com>
7622 * calls.c (load_register_parameters): Cope with polynomial
7623 mode sizes. Require a constant size for BLKmode parameters
7624 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7625 forces a parameter to be padded at the lsb end in order to
7626 fill a complete number of words, require the parameter size
7627 to be ordered wrt UNITS_PER_WORD.
7629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7630 Alan Hayward <alan.hayward@arm.com>
7631 David Sherwood <david.sherwood@arm.com>
7633 * reload1.c (spill_stack_slot_width): Change element type
7634 from unsigned int to poly_uint64_pod.
7635 (alter_reg): Treat mode sizes as polynomial.
7637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7638 Alan Hayward <alan.hayward@arm.com>
7639 David Sherwood <david.sherwood@arm.com>
7641 * reload.c (complex_word_subreg_p): New function.
7642 (reload_inner_reg_of_subreg, push_reload): Use it.
7644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7645 Alan Hayward <alan.hayward@arm.com>
7646 David Sherwood <david.sherwood@arm.com>
7648 * lra-constraints.c (process_alt_operands): Reject matched
7649 operands whose sizes aren't ordered.
7650 (match_reload): Refer to this check here.
7652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7653 Alan Hayward <alan.hayward@arm.com>
7654 David Sherwood <david.sherwood@arm.com>
7656 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7657 that the mode size is in the set {1, 2, 4, 8, 16}.
7659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7660 Alan Hayward <alan.hayward@arm.com>
7661 David Sherwood <david.sherwood@arm.com>
7663 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7664 Use plus_constant instead of gen_rtx_PLUS.
7666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7667 Alan Hayward <alan.hayward@arm.com>
7668 David Sherwood <david.sherwood@arm.com>
7670 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7671 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7672 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7673 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7674 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7675 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7676 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7677 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7678 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7679 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7681 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7682 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7683 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7684 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7685 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7686 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7687 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7688 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7689 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7690 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7692 * expr.c (emit_move_resolve_push): Treat the input and result
7693 of PUSH_ROUNDING as a poly_int64.
7694 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7695 (emit_push_insn): Likewise.
7696 * lra-eliminations.c (mark_not_eliminable): Likewise.
7697 * recog.c (push_operand): Likewise.
7698 * reload1.c (elimination_effects): Likewise.
7699 * rtlanal.c (nonzero_bits1): Likewise.
7700 * calls.c (store_one_arg): Likewise. Require the padding to be
7701 known at compile time.
7703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7704 Alan Hayward <alan.hayward@arm.com>
7705 David Sherwood <david.sherwood@arm.com>
7707 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7708 Use plus_constant instead of gen_rtx_PLUS.
7710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7711 Alan Hayward <alan.hayward@arm.com>
7712 David Sherwood <david.sherwood@arm.com>
7714 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7718 Alan Hayward <alan.hayward@arm.com>
7719 David Sherwood <david.sherwood@arm.com>
7721 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7722 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7723 via stack temporaries. Treat the mode size as polynomial too.
7725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7726 Alan Hayward <alan.hayward@arm.com>
7727 David Sherwood <david.sherwood@arm.com>
7729 * expr.c (expand_expr_real_2): When handling conversions involving
7730 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7731 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7732 as a poly_uint64 too.
7734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7735 Alan Hayward <alan.hayward@arm.com>
7736 David Sherwood <david.sherwood@arm.com>
7738 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7741 Alan Hayward <alan.hayward@arm.com>
7742 David Sherwood <david.sherwood@arm.com>
7744 * combine.c (can_change_dest_mode): Handle polynomial
7745 REGMODE_NATURAL_SIZE.
7746 * expmed.c (store_bit_field_1): Likewise.
7747 * expr.c (store_constructor): Likewise.
7748 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7749 and polynomial REGMODE_NATURAL_SIZE.
7750 (gen_lowpart_common): Likewise.
7751 * reginfo.c (record_subregs_of_mode): Likewise.
7752 * rtlanal.c (read_modify_subreg_p): Likewise.
7754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7755 Alan Hayward <alan.hayward@arm.com>
7756 David Sherwood <david.sherwood@arm.com>
7758 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7759 numbers of elements.
7761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7762 Alan Hayward <alan.hayward@arm.com>
7763 David Sherwood <david.sherwood@arm.com>
7765 * match.pd: Cope with polynomial numbers of vector elements.
7767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7768 Alan Hayward <alan.hayward@arm.com>
7769 David Sherwood <david.sherwood@arm.com>
7771 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7772 in a POINTER_PLUS_EXPR.
7774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7775 Alan Hayward <alan.hayward@arm.com>
7776 David Sherwood <david.sherwood@arm.com>
7778 * omp-simd-clone.c (simd_clone_subparts): New function.
7779 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7780 (ipa_simd_modify_function_body): Likewise.
7782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7783 Alan Hayward <alan.hayward@arm.com>
7784 David Sherwood <david.sherwood@arm.com>
7786 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7787 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7788 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7789 (expand_vector_condition, vector_element): Likewise.
7790 (subparts_gt): New function.
7791 (get_compute_type): Use subparts_gt.
7792 (count_type_subparts): Delete.
7793 (expand_vector_operations_1): Use subparts_gt instead of
7794 count_type_subparts.
7796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7797 Alan Hayward <alan.hayward@arm.com>
7798 David Sherwood <david.sherwood@arm.com>
7800 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7801 (vect_compile_time_alias): ...this new function. Do the calculation
7802 on poly_ints rather than trees.
7803 (vect_prune_runtime_alias_test_list): Update call accordingly.
7805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7806 Alan Hayward <alan.hayward@arm.com>
7807 David Sherwood <david.sherwood@arm.com>
7809 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7811 (vect_schedule_slp_instance): Likewise.
7813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7814 Alan Hayward <alan.hayward@arm.com>
7815 David Sherwood <david.sherwood@arm.com>
7817 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7818 constant and extern definitions for variable-length vectors.
7819 (vect_get_constant_vectors): Note that the number of units
7820 is known to be constant.
7822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7823 Alan Hayward <alan.hayward@arm.com>
7824 David Sherwood <david.sherwood@arm.com>
7826 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7827 of units as polynomial. Choose between WIDE and NARROW based
7830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7831 Alan Hayward <alan.hayward@arm.com>
7832 David Sherwood <david.sherwood@arm.com>
7834 * tree-vect-stmts.c (simd_clone_subparts): New function.
7835 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7838 Alan Hayward <alan.hayward@arm.com>
7839 David Sherwood <david.sherwood@arm.com>
7841 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7842 vectors as polynomial. Use build_index_vector for
7845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7846 Alan Hayward <alan.hayward@arm.com>
7847 David Sherwood <david.sherwood@arm.com>
7849 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7850 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7851 for variable-length vectors.
7852 (vectorizable_mask_load_store): Treat the number of units as
7853 polynomial, asserting that it is constant if the condition has
7854 already been enforced.
7855 (vectorizable_store, vectorizable_load): Likewise.
7857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7858 Alan Hayward <alan.hayward@arm.com>
7859 David Sherwood <david.sherwood@arm.com>
7861 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7862 of units as polynomial. Punt if we can't tell at compile time
7863 which vector contains the final result.
7865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7866 Alan Hayward <alan.hayward@arm.com>
7867 David Sherwood <david.sherwood@arm.com>
7869 * tree-vect-loop.c (vectorizable_induction): Treat the number
7870 of units as polynomial. Punt on SLP inductions. Use an integer
7871 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7872 cast of such a series for variable-length floating-point
7875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7876 Alan Hayward <alan.hayward@arm.com>
7877 David Sherwood <david.sherwood@arm.com>
7879 * tree.h (build_index_vector): Declare.
7880 * tree.c (build_index_vector): New function.
7881 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7882 of units as polynomial, forcibly converting it to a constant if
7883 vectorizable_reduction has already enforced the condition.
7884 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7885 to create a {1,2,3,...} vector.
7886 (vectorizable_reduction): Treat the number of units as polynomial.
7887 Choose vectype_in based on the largest scalar element size rather
7888 than the smallest number of units. Enforce the restrictions
7891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7892 Alan Hayward <alan.hayward@arm.com>
7893 David Sherwood <david.sherwood@arm.com>
7895 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7896 number of units as polynomial.
7898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7899 Alan Hayward <alan.hayward@arm.com>
7900 David Sherwood <david.sherwood@arm.com>
7902 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7903 * target.def (autovectorize_vector_sizes): Return the vector sizes
7904 by pointer, using vector_sizes rather than a bitmask.
7905 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7906 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7907 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7909 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7910 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7911 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7912 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7913 * omp-general.c (omp_max_vf): Likewise.
7914 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7915 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7916 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7917 * tree-vect-slp.c (vect_slp_bb): Likewise.
7918 * doc/tm.texi: Regenerate.
7919 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7921 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7922 the vector size as a poly_uint64 rather than an unsigned int.
7923 (current_vector_size): Change from an unsigned int to a poly_uint64.
7924 (get_vectype_for_scalar_type): Update accordingly.
7925 * tree.h (build_truth_vector_type): Take the size and number of
7926 units as a poly_uint64 rather than an unsigned int.
7927 (build_vector_type): Add a temporary overload that takes
7928 the number of units as a poly_uint64 rather than an unsigned int.
7929 * tree.c (make_vector_type): Likewise.
7930 (build_truth_vector_type): Take the number of units as a poly_uint64
7931 rather than an unsigned int.
7933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7934 Alan Hayward <alan.hayward@arm.com>
7935 David Sherwood <david.sherwood@arm.com>
7937 * target.def (get_mask_mode): Take the number of units and length
7938 as poly_uint64s rather than unsigned ints.
7939 * targhooks.h (default_get_mask_mode): Update accordingly.
7940 * targhooks.c (default_get_mask_mode): Likewise.
7941 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7942 * doc/tm.texi: Regenerate.
7944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7945 Alan Hayward <alan.hayward@arm.com>
7946 David Sherwood <david.sherwood@arm.com>
7948 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7949 * omp-general.c (omp_max_vf): Likewise.
7950 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7951 (expand_omp_simd): Handle polynomial safelen.
7952 * omp-low.c (omplow_simd_context): Add a default constructor.
7953 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7954 (lower_rec_simd_input_clauses): Update accordingly.
7955 (lower_rec_input_clauses): Likewise.
7957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7958 Alan Hayward <alan.hayward@arm.com>
7959 David Sherwood <david.sherwood@arm.com>
7961 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7962 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7963 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7964 (vect_analyze_slp_cost): Likewise.
7965 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7966 (vect_model_load_cost): Likewise.
7968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7969 Alan Hayward <alan.hayward@arm.com>
7970 David Sherwood <david.sherwood@arm.com>
7972 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7973 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7974 from an unsigned int * to a poly_uint64_pod *.
7975 (calculate_unrolling_factor): New function.
7976 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7979 Alan Hayward <alan.hayward@arm.com>
7980 David Sherwood <david.sherwood@arm.com>
7982 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7983 from an unsigned int to a poly_uint64.
7984 (_loop_vec_info::slp_unrolling_factor): Likewise.
7985 (_loop_vec_info::vectorization_factor): Change from an int
7987 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7988 (vect_get_num_vectors): New function.
7989 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7990 (vect_get_num_copies): Use vect_get_num_vectors.
7991 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7992 to an unsigned int *.
7993 (vect_analyze_data_refs): Change min_vf from an int * to a
7995 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7996 than an unsigned HOST_WIDE_INT.
7997 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7998 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7999 to an unsigned int *.
8000 (vect_analyze_data_ref_dependences): Likewise.
8001 (vect_compute_data_ref_alignment): Handle polynomial vf.
8002 (vect_enhance_data_refs_alignment): Likewise.
8003 (vect_prune_runtime_alias_test_list): Likewise.
8004 (vect_shift_permute_load_chain): Likewise.
8005 (vect_supportable_dr_alignment): Likewise.
8006 (dependence_distance_ge_vf): Take the vectorization factor as a
8007 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8008 (vect_analyze_data_refs): Change min_vf from an int * to a
8010 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8011 vfm1 as a poly_uint64 rather than an int. Make the same change
8012 for the returned bound_scalar.
8013 (vect_gen_vector_loop_niters): Handle polynomial vf.
8014 (vect_do_peeling): Likewise. Update call to
8015 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8016 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8018 * tree-vect-loop.c (vect_determine_vectorization_factor)
8019 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8020 (vect_get_known_peeling_cost): Likewise.
8021 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8022 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8023 (vect_transform_loop): Likewise. Use the lowest possible VF when
8024 updating the upper bounds of the loop.
8025 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8027 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8028 polynomial unroll factors.
8029 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8030 (vect_make_slp_decision): Likewise.
8031 (vect_supported_load_permutation_p): Likewise, and polynomial
8033 (vect_analyze_slp_cost): Handle polynomial vf.
8034 (vect_slp_analyze_node_operations): Likewise.
8035 (vect_slp_analyze_bb_1): Likewise.
8036 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8037 than an unsigned HOST_WIDE_INT.
8038 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8039 (vectorizable_load): Handle polynomial vf.
8040 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8042 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8045 Alan Hayward <alan.hayward@arm.com>
8046 David Sherwood <david.sherwood@arm.com>
8048 * match.pd: Handle bit operations involving three constants
8049 and try to fold one pair.
8051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8053 * tree-vect-loop-manip.c: Include gimple-fold.h.
8054 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8055 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8056 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8057 Add a path that uses a step of VF instead of 1, but disable it
8059 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8060 and niters_no_overflow parameters. Update calls to
8061 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8062 Create a new SSA name if the latter choses to use a ste other
8063 than zero, and return it via niters_vector_mult_vf_var.
8064 * tree-vect-loop.c (vect_transform_loop): Update calls to
8065 vect_do_peeling, vect_gen_vector_loop_niters and
8066 slpeel_make_loop_iterate_ntimes.
8067 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8068 (vect_gen_vector_loop_niters): Update declarations after above changes.
8070 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8072 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8073 128-bit round to integer instructions.
8074 (ceil<mode>2): Likewise.
8075 (btrunc<mode>2): Likewise.
8076 (round<mode>2): Likewise.
8078 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8080 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8081 unaligned VSX load/store on P8/P9.
8082 (expand_block_clear): Allow the use of unaligned VSX
8083 load/store on P8/P9.
8085 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8087 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8089 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8090 swap associated with both a load and a store.
8092 2018-01-02 Andrew Waterman <andrew@sifive.com>
8094 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8095 * config/riscv/riscv.md (clear_cache): Use it.
8097 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8099 * web.c: Remove out-of-date comment.
8101 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8103 * expr.c (fixup_args_size_notes): Check that any existing
8104 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8105 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8106 (emit_single_push_insn): ...here.
8108 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8110 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8111 (const_vector_encoded_nelts): New function.
8112 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8113 (const_vector_int_elt, const_vector_elt): Declare.
8114 * emit-rtl.c (const_vector_int_elt_1): New function.
8115 (const_vector_elt): Likewise.
8116 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8117 of CONST_VECTOR_ELT.
8119 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8121 * expr.c: Include rtx-vector-builder.h.
8122 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8123 directly on the tree encoding.
8124 (const_vector_from_tree): Likewise.
8125 * optabs.c: Include rtx-vector-builder.h.
8126 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8127 sequence of "u" values.
8128 * vec-perm-indices.c: Include rtx-vector-builder.h.
8129 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8130 directly on the vec_perm_indices encoding.
8132 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8134 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8135 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8136 * rtx-vector-builder.h: New file.
8137 * rtx-vector-builder.c: Likewise.
8138 * rtl.h (rtx_def::u2): Add a const_vector field.
8139 (CONST_VECTOR_NPATTERNS): New macro.
8140 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8141 (CONST_VECTOR_DUPLICATE_P): Likewise.
8142 (CONST_VECTOR_STEPPED_P): Likewise.
8143 (CONST_VECTOR_ENCODED_ELT): Likewise.
8144 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8145 (unwrap_const_vec_duplicate): Likewise.
8146 (const_vec_series_p): Check for a non-duplicated vector encoding.
8147 Say that the function only returns true for integer vectors.
8148 * emit-rtl.c: Include rtx-vector-builder.h.
8149 (gen_const_vec_duplicate_1): Delete.
8150 (gen_const_vector): Call gen_const_vec_duplicate instead of
8151 gen_const_vec_duplicate_1.
8152 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8153 (gen_const_vec_duplicate): Use rtx_vector_builder.
8154 (gen_const_vec_series): Likewise.
8155 (gen_rtx_CONST_VECTOR): Likewise.
8156 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8157 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8158 Build a new vector rather than modifying a CONST_VECTOR in-place.
8159 (handle_special_swappables): Update call accordingly.
8160 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8161 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8162 Build a new vector rather than modifying a CONST_VECTOR in-place.
8163 (handle_special_swappables): Update call accordingly.
8165 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8167 * simplify-rtx.c (simplify_const_binary_operation): Use
8168 CONST_VECTOR_ELT instead of XVECEXP.
8170 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8172 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8173 the selector elements to be different from the data elements
8174 if the selector is a VECTOR_CST.
8175 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8176 ssizetype for the selector.
8178 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8180 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8181 before testing each element individually.
8182 * tree-vect-generic.c (lower_vec_perm): Likewise.
8184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8186 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8187 * selftest-run-tests.c (selftest::run_tests): Call it.
8188 * vector-builder.h (vector_builder::operator ==): New function.
8189 (vector_builder::operator !=): Likewise.
8190 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8191 (vec_perm_indices::all_from_input_p): New function.
8192 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8193 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8194 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8195 instead of reading the VECTOR_CST directly. Detect whether both
8196 vector inputs are the same before constructing the vec_perm_indices,
8197 and update the number of inputs argument accordingly. Use the
8198 utility functions added above. Only construct sel2 if we need to.
8200 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8202 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8203 the broadcast of the low byte.
8204 (expand_mult_highpart): Use an explicit encoding for the permutes.
8205 * optabs-query.c (can_mult_highpart_p): Likewise.
8206 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8207 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8208 (vectorizable_bswap): Likewise.
8209 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8210 explicit encoding for the power-of-2 permutes.
8211 (vect_permute_store_chain): Likewise.
8212 (vect_grouped_load_supported): Likewise.
8213 (vect_permute_load_chain): Likewise.
8215 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8217 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8218 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8219 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8220 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8221 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8222 (vect_gen_perm_mask_any): Likewise.
8224 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8226 * int-vector-builder.h: New file.
8227 * vec-perm-indices.h: Include int-vector-builder.h.
8228 (vec_perm_indices): Redefine as an int_vector_builder.
8229 (auto_vec_perm_indices): Delete.
8230 (vec_perm_builder): Redefine as a stand-alone class.
8231 (vec_perm_indices::vec_perm_indices): New function.
8232 (vec_perm_indices::clamp): Likewise.
8233 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8234 (vec_perm_indices::new_vector): New function.
8235 (vec_perm_indices::new_expanded_vector): Update for new
8236 vec_perm_indices class.
8237 (vec_perm_indices::rotate_inputs): New function.
8238 (vec_perm_indices::all_in_range_p): Operate directly on the
8239 encoded form, without computing elided elements.
8240 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8241 encoding. Update for new vec_perm_indices class.
8242 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8243 the given vec_perm_builder.
8244 (expand_vec_perm_var): Update vec_perm_builder constructor.
8245 (expand_mult_highpart): Use vec_perm_builder instead of
8246 auto_vec_perm_indices.
8247 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8248 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8249 or double series encoding as appropriate.
8250 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8251 vec_perm_indices instead of auto_vec_perm_indices.
8252 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8253 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8254 (vect_permute_store_chain): Likewise.
8255 (vect_grouped_load_supported): Likewise.
8256 (vect_permute_load_chain): Likewise.
8257 (vect_shift_permute_load_chain): Likewise.
8258 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8259 (vect_transform_slp_perm_load): Likewise.
8260 (vect_schedule_slp_instance): Likewise.
8261 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8262 (vectorizable_mask_load_store): Likewise.
8263 (vectorizable_bswap): Likewise.
8264 (vectorizable_store): Likewise.
8265 (vectorizable_load): Likewise.
8266 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8267 vec_perm_indices instead of auto_vec_perm_indices. Use
8268 tree_to_vec_perm_builder to read the vector from a tree.
8269 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8270 vec_perm_builder instead of a vec_perm_indices.
8271 (have_whole_vector_shift): Use vec_perm_builder and
8272 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8273 truncation to calc_vec_perm_mask_for_shift.
8274 (vect_create_epilog_for_reduction): Likewise.
8275 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8276 from auto_vec_perm_indices to vec_perm_indices.
8277 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8278 instead of changing individual elements.
8279 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8280 the vector in d.perm.
8281 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8282 from auto_vec_perm_indices to vec_perm_indices.
8283 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8284 instead of changing individual elements.
8285 (arm_vectorize_vec_perm_const): Use new_vector to install
8286 the vector in d.perm.
8287 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8288 Update vec_perm_builder constructor.
8289 (rs6000_expand_interleave): Likewise.
8290 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8291 (rs6000_expand_interleave): Likewise.
8293 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8295 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8296 to qimode could truncate the indices.
8297 * optabs.c (expand_vec_perm_var): Likewise.
8299 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8301 * Makefile.in (OBJS): Add vec-perm-indices.o.
8302 * vec-perm-indices.h: New file.
8303 * vec-perm-indices.c: Likewise.
8304 * target.h (vec_perm_indices): Replace with a forward class
8306 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8307 * optabs.h: Include vec-perm-indices.h.
8308 (expand_vec_perm): Delete.
8309 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8310 (expand_vec_perm_const): Declare.
8311 * target.def (vec_perm_const_ok): Replace with...
8312 (vec_perm_const): ...this new hook.
8313 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8314 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8315 * doc/tm.texi: Regenerate.
8316 * optabs.def (vec_perm_const): Delete.
8317 * doc/md.texi (vec_perm_const): Likewise.
8318 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8319 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8320 expand_vec_perm for constant permutation vectors. Assert that
8321 the mode of variable permutation vectors is the integer equivalent
8322 of the mode that is being permuted.
8323 * optabs-query.h (selector_fits_mode_p): Declare.
8324 * optabs-query.c: Include vec-perm-indices.h.
8325 (selector_fits_mode_p): New function.
8326 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8327 is defined, instead of checking whether the vec_perm_const_optab
8328 exists. Use targetm.vectorize.vec_perm_const instead of
8329 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8330 fit in the vector mode before using a variable permute.
8331 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8332 vec_perm_indices instead of an rtx.
8333 (expand_vec_perm): Replace with...
8334 (expand_vec_perm_const): ...this new function. Take the selector
8335 as a vec_perm_indices rather than an rtx. Also take the mode of
8336 the selector. Update call to shift_amt_for_vec_perm_mask.
8337 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8338 Use vec_perm_indices::new_expanded_vector to expand the original
8339 selector into bytes. Check whether the indices fit in the vector
8340 mode before using a variable permute.
8341 (expand_vec_perm_var): Make global.
8342 (expand_mult_highpart): Use expand_vec_perm_const.
8343 * fold-const.c: Includes vec-perm-indices.h.
8344 * tree-ssa-forwprop.c: Likewise.
8345 * tree-vect-data-refs.c: Likewise.
8346 * tree-vect-generic.c: Likewise.
8347 * tree-vect-loop.c: Likewise.
8348 * tree-vect-slp.c: Likewise.
8349 * tree-vect-stmts.c: Likewise.
8350 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8352 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8353 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8354 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8355 (aarch64_vectorize_vec_perm_const): ...this new function.
8356 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8357 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8358 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8359 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8360 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8361 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8362 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8364 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8365 check for NEON modes.
8366 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8367 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8368 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8369 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8371 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8372 the old VEC_PERM_CONST conditions.
8373 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8374 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8375 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8376 (ia64_vectorize_vec_perm_const_ok): Merge into...
8377 (ia64_vectorize_vec_perm_const): ...this new function.
8378 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8379 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8380 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8381 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8382 * config/mips/mips.c (mips_expand_vec_perm_const)
8383 (mips_vectorize_vec_perm_const_ok): Merge into...
8384 (mips_vectorize_vec_perm_const): ...this new function.
8385 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8386 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8387 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8388 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8389 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8390 (rs6000_expand_vec_perm_const): Delete.
8391 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8393 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8394 (altivec_expand_vec_perm_const_le): Take each operand individually.
8395 Operate on constant selectors rather than rtxes.
8396 (altivec_expand_vec_perm_const): Likewise. Update call to
8397 altivec_expand_vec_perm_const_le.
8398 (rs6000_expand_vec_perm_const): Delete.
8399 (rs6000_vectorize_vec_perm_const_ok): Delete.
8400 (rs6000_vectorize_vec_perm_const): New function.
8401 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8402 an element count and rtx array.
8403 (rs6000_expand_extract_even): Update call accordingly.
8404 (rs6000_expand_interleave): Likewise.
8405 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8406 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8407 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8408 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8409 (rs6000_expand_vec_perm_const): Delete.
8410 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8411 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8412 (altivec_expand_vec_perm_const_le): Take each operand individually.
8413 Operate on constant selectors rather than rtxes.
8414 (altivec_expand_vec_perm_const): Likewise. Update call to
8415 altivec_expand_vec_perm_const_le.
8416 (rs6000_expand_vec_perm_const): Delete.
8417 (rs6000_vectorize_vec_perm_const_ok): Delete.
8418 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8419 reference to the SPE evmerge intructions.
8420 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8421 an element count and rtx array.
8422 (rs6000_expand_extract_even): Update call accordingly.
8423 (rs6000_expand_interleave): Likewise.
8424 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8425 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8427 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8429 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8431 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8432 vector mode and that that mode matches the mode of the data
8434 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8435 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8436 directly using expand_vec_perm_1 when forcing selectors into
8438 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8440 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8442 * optabs-query.h (can_vec_perm_p): Delete.
8443 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8444 * optabs-query.c (can_vec_perm_p): Split into...
8445 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8446 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8447 particular selector is valid.
8448 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8449 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8450 (vect_grouped_load_supported): Likewise.
8451 (vect_shift_permute_load_chain): Likewise.
8452 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8453 (vect_transform_slp_perm_load): Likewise.
8454 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8455 (vectorizable_bswap): Likewise.
8456 (vect_gen_perm_mask_checked): Likewise.
8457 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8458 implementations of variable permutation vectors into account
8459 when deciding which selector to use.
8460 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8461 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8462 with a false third argument.
8463 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8464 to test whether the constant selector is valid and can_vec_perm_var_p
8465 to test whether a variable selector is valid.
8467 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8469 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8470 * optabs-query.c (can_vec_perm_p): Likewise.
8471 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8472 instead of vec_perm_indices.
8473 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8474 (vect_gen_perm_mask_checked): Likewise,
8475 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8476 (vect_gen_perm_mask_checked): Likewise,
8478 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8480 * optabs-query.h (qimode_for_vec_perm): Declare.
8481 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8482 (qimode_for_vec_perm): ...this new function.
8483 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8485 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8487 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8488 does not have a conditional at the top.
8490 2018-01-02 Richard Biener <rguenther@suse.de>
8492 * ipa-inline.c (big_speedup_p): Fix expression.
8494 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8497 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8500 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8504 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8505 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8506 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8507 cond_taken_branch_cost 3->4.
8509 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8511 PR tree-optimization/83581
8512 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8513 TODO_cleanup_cfg if any changes have been made.
8516 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8517 convert_modes if target mode has the right side, but different mode
8521 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8522 last argument when extracting from CONCAT. If either from_real or
8523 from_imag is NULL, use expansion through memory. If result is not
8524 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8525 the parts directly to inner mode, if even that fails, use expansion
8529 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8530 check for bswap in mode rather than HImode and use that in expand_unop
8533 Copyright (C) 2018 Free Software Foundation, Inc.
8535 Copying and distribution of this file, with or without modification,
8536 are permitted in any medium without royalty provided the copyright
8537 notice and this notice are preserved.