1 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5 (dw_val_node): Add val_symbolic_view.
6 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
7 (symview_upper_bound): New.
8 (new_line_info_table): Initialize symviews_since_reset.
9 (dwarf2out_source_line): Count symviews_since_reset and set
11 (dw_val_equal_p): Handle symview.
12 (add_AT_symview): New.
13 (print_dw_val): Handle symview.
14 (attr_checksum, attr_checksum_ordered): Likewise.
15 (same_dw_val_p, size_of_die): Likewise.
16 (value_format, output_die): Likewise.
17 (add_high_low_attributes): Use add_AT_symview for entry_view.
18 (dwarf2out_finish): Reset symview_upper_bound, clear
21 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
24 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
25 Add strict argument and use it.
26 (rs6000_split_multireg_move): Update for new strict argument.
27 (mem_operand_gpr): Disallow all non-offsettable addresses.
28 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
30 2018-03-09 Jakub Jelinek <jakub@redhat.com>
33 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
34 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
35 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
38 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
39 decl, use remap_type if we want to use the type.
41 2018-03-09 Martin Sebor <msebor@redhat.com>
43 PR tree-optimization/84526
44 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
46 (builtin_access::generic_overlap): Be prepared to handle non-array
49 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
51 PR rtl-optimization/84682
52 * lra-constraints.c (process_address_1): Check is_address flag
53 for address constraints.
54 (process_alt_operands): Likewise.
55 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
56 preprocess_constraints.
57 * recog.h (preprocess_constraints): Add oploc parameter.
59 * recog.c (preprocess_constraints): Test address_operand for
60 CT_ADDRESS constraints.
62 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
65 * lra-assigns.c (assign_by_spills): Return a flag of reload
66 assignment failure. Do not process the reload assignment
67 failures. Do not spill other reload pseudos if they has the same
69 (lra_assign): Add a return arg. Set up from the result of
70 assign_by_spills call.
71 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
72 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
73 usage_insns if it is not NULL.
74 (spill_hard_reg_in_range): New function.
75 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
76 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
78 (lra_assign): Change prototype.
79 * lra.c (lra): Add code to deal with fails by splitting hard reg
82 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
85 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
86 Accept complain bool parameter. Only emit errors if it is true.
87 (arm_parse_cpu_option_name): Likewise.
88 (arm_target_thumb_only): Adjust callers of the above.
89 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
90 prototype to take a default true bool parameter.
91 (arm_parse_arch_option_name): Likewise.
93 2018-03-09 David Malcolm <dmalcolm@redhat.com>
94 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
98 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
99 * configure: Regenerate.
100 * configure.ac ("linker --version-script option"): New.
101 ("linker soname option"): New.
103 2018-03-09 Richard Biener <rguenther@suse.de>
105 PR tree-optimization/84775
106 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
107 immediate uses of predicate stmts and mark them modified.
110 PR tree-optimization/84178
111 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
113 (version_loop_for_if_conversion): Delay update_ssa call.
114 (tree_if_conversion): Delay update_ssa until after predicate
117 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
120 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
121 when the function accesses prior frames.
123 2018-03-08 Jakub Jelinek <jakub@redhat.com>
126 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
127 gen_llsym, otherwise call maybe_gen_llsym.
130 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
131 has ',' character inside of it.
133 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
136 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
137 as clobbering CC_REGNUM.
139 2018-03-08 Richard Biener <rguenther@suse.de>
142 * tree-scalar-evolution.c: Include tree-into-ssa.h.
143 (follow_copies_to_constant): Do not follow SSA names registered
146 2018-03-08 Richard Biener <rguenther@suse.de>
148 PR tree-optimization/84178
149 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
151 (version_loop_for_if_conversion): Delay update_ssa call.
152 (tree_if_conversion): Delay update_ssa until after predicate
155 2018-03-08 David Malcolm <dmalcolm@redhat.com>
157 PR tree-optimization/84178
158 * tree-if-conv.c (release_bb_predicate): Remove the
159 the assertion that the stmts have NULL use_ops.
160 Discard the statements, asserting that they haven't
161 yet been added to a BB.
163 2018-03-08 Richard Biener <rguenther@suse.de>
165 PR tree-optimization/84746
166 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
167 (phi_translate): Pass in destination ANTIC_OUT set.
168 (phi_translate_1): Likewise. For a simplified result lookup
169 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
170 (phi_translate_set): Adjust.
171 (do_pre_regular_insertion): Likewise.
172 (do_pre_partial_partial_insertion): Likewise.
174 2018-03-08 Martin Liska <mliska@suse.cz>
176 PR gcov-profile/84735
177 * doc/gcov.texi: Document usage of profile files.
178 * gcov-io.h: Document changes in the format.
180 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
184 * dwarf2out.c (struct dw_line_info_table): Update comments for
186 (FORCE_RESET_NEXT_VIEW): New.
187 (FORCE_RESETTING_VIEW_P): New.
188 (RESETTING_VIEW_P): Check for -1 too.
189 (ZERO_VIEW_P): Likewise.
190 (new_line_info_table): Force-reset next view.
191 (dwarf2out_begin_function): Likewise.
192 (dwarf2out_source_line): Simplify zero_view_p initialization.
193 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
194 view directly. Omit view when omitting .loc at line 0.
196 2018-03-08 Jakub Jelinek <jakub@redhat.com>
198 PR tree-optimization/84740
199 * tree-switch-conversion.c (process_switch): Call build_constructors
200 only if info.phi_count is non-zero.
202 PR tree-optimization/84739
203 * tree-tailcall.c (find_tail_calls): Check call arguments against
204 DECL_ARGUMENTS (current_function_decl) rather than
205 DECL_ARGUMENTS (func) when checking for tail recursion.
207 2018-03-07 Jakub Jelinek <jakub@redhat.com>
209 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
210 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
211 Volker Reichelt's entry and add entries for people that perform
212 GCC fuzzy testing and report numerous bugs.
214 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
217 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
218 readonly data in sdata, if that is disabled.
219 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
220 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
221 -mreadonly-in-sdata option.
223 2018-03-07 Martin Sebor <msebor@redhat.com>
225 PR tree-optimization/84468
226 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
227 basic block when looking for nul assignment.
229 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
232 * except.h (output_function_exception_table): Adjust prototype.
233 * except.c (output_function_exception_table): Remove FNNAME parameter
234 and add SECTION parameter. Ouput one part of the table at a time.
235 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
236 the first part of the exception table and emit unwind directives.
237 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
238 (i386_pe_seh_cold_init): Likewise.
239 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
240 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
241 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
242 (ix86_output_call_insn): Emit a nop in one more case for SEH.
243 * config/i386/winnt.c: Include except.h.
244 (struct seh_frame_state): Add reg_offset, after_prologue and
245 in_cold_section fields.
246 (i386_pe_seh_end_prologue): Set seh->after_prologue.
247 (i386_pe_seh_cold_init): New function.
248 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
249 to seh->in_cold_section.
250 (seh_emit_push): Record the offset of the push.
251 (seh_emit_save): Record the offet of the save.
252 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
253 Test seh->after_prologue to disregard the epilogue.
254 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
255 (i386_pe_end_cold_function): New function.
257 2018-03-07 Jakub Jelinek <jakub@redhat.com>
260 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
261 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
264 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
266 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
267 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
270 * multiple_target.c: Include tree-inline.h and intl.h.
271 (expand_target_clones): Diagnose and fail if node->definition and
272 !tree_versionable_function_p (node->decl).
274 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
276 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
278 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
279 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
280 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
282 2018-03-06 Jakub Jelinek <jakub@redhat.com>
285 * combine.c (try_combine): Use reg_or_subregno instead of handling
286 just paradoxical SUBREGs and REGs.
288 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
290 * config/arc/arc.c (arc_finalize_pic): Remove function.
291 (arc_must_save_register): We use single base PIC register, remove
292 checks to save/restore the PIC register.
293 (arc_expand_prologue): Likewise.
294 * config/arc/arc-protos.h (arc_set_default_type_attributes):
296 (arc_verify_short): Likewise.
297 (arc_attr_type): Likewise.
298 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
299 (walk_stores): Likewise.
300 (arc_address_cost): Make it static.
301 (arc_verify_short): Likewise.
302 (branch_dest): Likewise.
303 (arc_attr_type): Likewise.
304 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
305 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
306 (arc_final_prescan_insn): Remove inserting the nops due to
307 hardware hazards. It is done in reorg step.
308 (insn_length_variant_t): Remove.
309 (insn_length_parameters_t): Likewise.
310 (arc_insn_length_parameters): Likewise.
311 (arc_get_insn_variants): Likewise.
312 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
314 2018-03-06 Jakub Jelinek <jakub@redhat.com>
317 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
320 PR tree-optimization/84687
321 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
323 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
325 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
327 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
328 Rename to ppc_speculation_barrier.
329 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
330 __builtin_ppc_speculation_barrier.
332 2018-03-05 Jakub Jelinek <jakub@redhat.com>
335 * combine.c (combine_simplify_rtx): Don't try to simplify if
336 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
339 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
341 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
342 to 32 bytes when compiling for POWER9.
344 2018-03-05 Jakub Jelinek <jakub@redhat.com>
347 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
348 regparm >= 3 with no arg reg available also for calls with
349 flag_force_indirect_call. Pass decl to ix86_function_regparm.
352 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
354 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
356 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
359 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
361 2018-03-05 Richard Biener <rguenther@suse.de>
363 PR tree-optimization/84486
364 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
365 When inserting a __builtin_assume_aligned call set the LHS
366 SSA name alignment info accordingly.
368 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
370 PR tree-optimization/84114
371 * config/aarch64/aarch64.c (aarch64_reassociation_width)
372 Avoid reassociation of FLOAT_MODE addition.
374 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
376 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
377 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
378 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
379 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
380 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
381 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
382 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
384 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
385 __builtin_ia32_wbinvd): New builtins.
386 (SPECIAL_ARGS2): New.
387 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
388 (SPECIAL_ARGS2): New.
389 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
390 (ix86_valid_target_attribute_inner_p): Ditto.
391 (ix86_init_mmx_sse_builtins): Add special_args2.
392 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
393 TARGET_WBNOINVD_P): New.
394 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
395 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
396 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
397 * config/i386/immintrin.h (_wbinvd): New intrinsic.
398 * config/i386/pconfigintrin.h: New file.
399 * config/i386/wbnoinvdintrin.h: Ditto.
400 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
401 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
403 2018-03-05 Richard Biener <rguenther@suse.de>
405 PR tree-optimization/84670
406 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
408 (BB_VISITED_WITH_VISITED_SUCCS): New define.
409 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
410 (compute_antic_aux): Only assert the number of values in ANTIC_IN
411 doesn't grow if all successors (recursively) were visited at least
414 2018-03-05 Richard Biener <rguenther@suse.de>
416 PR tree-optimization/84650
417 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
418 if executed in the loop pipeline.
420 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
422 * doc/configfiles.texi (Configuration Files): Move info about
423 conditionalizing $target-protos.h to...
424 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
425 differs from $target-protos.h.
427 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
428 Chung-Ju Wu <jasonwucj@gmail.com>
430 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
431 * config/nds32/nds32-multiple.md (setmemsi): Define.
432 * config/nds32/nds32-memory-manipulation.c
433 (nds32_gen_dup_4_byte_to_word_value): New.
434 (emit_setmem_word_loop): New.
435 (emit_setmem_byte_loop): New.
436 (nds32_expand_setmem_loop): New.
437 (nds32_expand_setmem_loop_v3m): New.
438 (nds32_expand_setmem_unroll): New.
439 (nds32_expand_setmem): New.
441 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
442 Chung-Ju Wu <jasonwucj@gmail.com>
444 * config/nds32/nds32-memory-manipulation.c
445 (nds32_emit_load_store): New.
446 (nds32_emit_post_inc_load_store): New.
447 (nds32_emit_mem_move): New.
448 (nds32_emit_mem_move_block): New.
449 (nds32_expand_movmemsi_loop_unknown_size): New.
450 (nds32_expand_movmemsi_loop_known_size): New.
451 (nds32_expand_movmemsi_loop): New.
452 (nds32_expand_movmemsi_unroll): New.
453 (nds32_expand_movmemqi): Rename ...
454 (nds32_expand_movmemsi): ... to this.
455 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
456 (movmemsi): ... to this.
457 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
458 (nds32_expand_movmemsi): ... to this.
460 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
461 Monk Chiang <sh.chiang04@gmail.com>
462 Chung-Ju Wu <jasonwucj@gmail.com>
464 * config/nds32/nds32-protos.h
465 (nds32_expand_load_multiple): New arguments.
466 (nds32_expand_store_multiple): Ditto.
467 (nds32_valid_multiple_load_store): Rename ...
468 (nds32_valid_multiple_load_store_p): ... to this.
469 * config/nds32/nds32-memory-manipulation.c
470 (nds32_expand_load_multiple): Refine implementation.
471 (nds32_expand_store_multiple): Ditto.
472 * config/nds32/nds32-multiple.md
473 (load_multiple): Update nds32_expand_load_multiple interface.
474 (store_multiple): Update nds32_expand_store_multiple interface.
475 * config/nds32/nds32-predicates.c
476 (nds32_valid_multiple_load_store): Rename ...
477 (nds32_valid_multiple_load_store_p): ... to this and refine
479 * config/nds32/predicates.md
480 (nds32_load_multiple_and_update_address_operation): New predicate.
481 (nds32_store_multiple_and_update_address_operation): New predicate.
483 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
484 Chung-Ju Wu <jasonwucj@gmail.com>
486 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
487 (combo): New attribute.
488 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
490 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
492 * config/nds32/nds32.opt: Change -mcmodel= default value.
494 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
495 Monk Chiang <sh.chiang04@gmail.com>
496 Chung-Ju Wu <jasonwucj@gmail.com>
498 * config/nds32/constants.md (unspec_element): New enum.
499 * config/nds32/constraints.md (Umw): New constraint.
500 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
501 * config/nds32/nds32-intrinsic.md: Likewise.
502 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
503 (nds32_valid_smw_lwm_base_p): New.
504 (nds32_output_smw_single_word): New.
505 (nds32_output_lmw_single_word): New.
506 (nds32_expand_unaligned_load): New.
507 (nds32_expand_unaligned_store): New.
508 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
509 (nds32_output_smw_single_word): Declare.
510 (nds32_output_lmw_single_word): Declare.
511 (nds32_expand_unaligned_load): Declare.
512 (nds32_expand_unaligned_store): Declare.
513 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
514 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
515 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
516 NDS32_BUILTIN_UASTORE_DW.
517 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
520 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
521 Kito Cheng <kito.cheng@gmail.com>
522 Chung-Ju Wu <jasonwucj@gmail.com>
524 * config/nds32/nds32-intrinsic.c
525 (nds32_expand_builtin_null_ftype_reg): Delete.
526 (nds32_expand_builtin_reg_ftype_imm): Ditto.
527 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
528 (nds32_read_argument): New.
529 (nds32_legitimize_target): Ditto.
530 (nds32_legitimize_argument): Ditto.
531 (nds32_check_constant_argument): Ditto.
532 (nds32_expand_unop_builtin): Ditto.
533 (nds32_expand_unopimm_builtin): Ditto.
534 (nds32_expand_binop_builtin): Ditto.
535 (nds32_builtin_decl_impl): Ditto.
536 (builtin_description): Ditto.
537 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
538 (nds32_init_builtins_impl): Ditto.
539 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
540 (nds32_builtin_decl): New.
541 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
542 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
544 2018-03-02 Jeff Law <law@redhat.com>
546 * reorg.c (stop_search_p): Handle DEBUG_INSN.
547 (redundant_insn, fill_simple_delay_slots): Likewise.
548 (fill_slots_from_thread): Likewise.
549 * resource.c (mark_referenced_resources): Likewise.
550 (mark_set_resources, find_dead_or_set_registers): Likewise.
552 2018-03-02 Jakub Jelinek <jakub@redhat.com>
554 * substring-locations.h (format_warning_va): Formatting fix for
556 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
558 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
559 * substring-locations.c: Include intl.h.
560 (format_warning_va): Turned into small wrapper around
561 format_warning_n_va, renamed to ...
562 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
563 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
565 (format_warning_at_substring_n): New function.
566 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
567 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
568 format_warning_at_substring with just a shorter name instead of
569 const function pointer.
570 (fmtwarn_n): New function.
571 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
572 appropriate, get rid of all the fmtstr temporaries, move conditionals
573 with G_() wrapped string literals directly into fmtwarn arguments,
574 cast dir.len to (int), formatting fixes.
576 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
578 * doc/invoke.texi: Remove "Cilk Plus" references.
580 2018-03-02 Jakub Jelinek <jakub@redhat.com>
581 Richard Biener <rguenther@suse.de>
584 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
585 for error or warning attributes if CALL_FROM_THUNK_P is set.
588 2018-03-02 Jakub Jelinek <jakub@redhat.com>
591 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
592 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
595 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
596 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
598 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
599 instead of -1U in last predictors element's probability member.
601 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
604 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
605 arguments if they are comparable.
607 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
609 PR tree-optimization/84634
610 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
611 masks and masked_loop_p with a single loop_masks, making sure it's
612 null for bb vectorization.
614 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
616 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
617 (vect_analyze_data_ref_access): Use loop->safe_len rather than
618 loop->force_vectorize to check whether there is no alias.
620 2018-03-02 Jakub Jelinek <jakub@redhat.com>
623 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
625 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
627 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
628 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
629 instead of a loop around prev_real_insn.
630 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
634 * config/i386/i386.c (ix86_print_operand): Use conditional
635 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
638 2018-03-02 Richard Biener <rguenther@suse.de>
640 PR tree-optimization/84427
641 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
642 (bitmap_set_subtract_values): Rewrite to handle multiple
645 (prune_clobbered_mems): Likewise.
646 (phi_translate): Take edge instead of pred/phiblock.
647 (phi_translate_1): Likewise.
648 (phi_translate_set): Likewise. Insert all translated
649 exprs for a value into the set, keeping possibly multiple
650 expressions per value.
651 (compute_antic_aux): Adjust for phi_translate changes.
652 When intersecting union the expressions and prune those
653 not in the final value set, keeping possibly multiple
654 expressions per value. Do not use value-insertion
655 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
656 all expressions. Add verification that the value-sets
657 only shrink during iteration.
658 (compute_partial_antic_aux): Adjust for the phi_translate changes.
659 (do_pre_regular_insertion): Likewise.
660 (do_pre_partial_partial_insertion): Likewise.
662 2018-03-02 Richard Biener <rguenther@suse.de>
665 * config/darwin.c (saved_debug_info_level): New static global.
666 (darwin_asm_lto_start): Disable debug info generation for LTO out.
667 (darwin_asm_lto_end): Restore debug info generation settings.
669 2018-03-01 Martin Liska <mliska@suse.cz>
672 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
675 2018-03-01 Richard Biener <rguenther@suse.de>
678 * dwarf2out.c (gen_variable_die): Properly handle late VLA
679 type annotation with LTO when debug was disabled at compile-time.
681 2018-03-01 Matthew Fortune <mfortune@gmail.com>
683 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
685 (mips_final_postscan_insn): Likewise.
687 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
689 PR rtl-optimization/84528
690 * alias.c (init_alias_target): Add commentary.
691 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
692 a unique base value if the frame pointer is not eliminated
693 to the stack pointer.
695 2018-03-01 Tom de Vries <tom@codesourcery.com>
697 PR rtl-optimization/83327
698 * lra-int.h (hard_regs_spilled_into): Declare.
699 * lra.c (hard_regs_spilled_into): Define.
700 (init_reg_info): Init hard_regs_spilled_into.
701 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
702 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
703 (process_bb_lives): Handle hard_regs_spilled_into.
704 (lra_create_live_ranges_1): Before doing liveness propagation, clear
705 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
707 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
709 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
710 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
711 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
712 * config/rs6000/aix72.h: New file.
714 2018-02-28 Jakub Jelinek <jakub@redhat.com>
716 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
717 instead of warning_at with conditional singular and plural messages
721 * stor-layout.c (update_alignment_for_field): For
722 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
723 && !DECL_PACKED (field), do the alignment update, just use
724 only desired_align instead of MAX (type_align, desired_align)
726 (place_field): Don't do known_align < desired_align handling
727 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
728 is non-NULL, instead do it after rli->prev_field handling and
729 only if not within a bitfield word. For DECL_PACKED (field)
730 use type_align of BITS_PER_UNIT.
732 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
734 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
735 superfluous parentheses and trailing spaces.
737 2018-02-28 Richard Biener <rguenther@suse.de>
739 PR tree-optimization/84584
740 * graphite-scop-detection.c (scop_detection::add_scop): Discard
741 SCoPs with fake exit edge.
743 2018-02-28 Martin Liska <mliska@suse.cz>
746 * timevar.c (timer::print): Fix format to properly print 100%
749 2018-02-28 Richard Biener <rguenther@suse.de>
752 * genmatch.c (capture_info::walk_match): Do not mark
753 captured expressions without operands as expr_p given
754 they act more like predicates and should be subject to
755 "lost tail" side-effect preserving.
757 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
759 PR rtl-optimization/81611
760 * auto-inc-dec.c (attempt_change): Move dead note from
761 mem_insn if it's the next use of regno
762 (find_address): Take address use of reg holding
763 non-incremented value. Add parm to limit search to the named
765 (merge_in_block): Attempt to use a mem insn that is the next
766 use of the original regno.
768 2018-02-27 Martin Sebor <msebor@redhat.com>
771 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
772 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
774 2018-02-27 Martin Sebor <msebor@redhat.com>
777 * diagnostic-core.h (warning_n, error_n, inform_n): Change
778 n argument to unsigned HOST_WIDE_INT.
779 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
780 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
781 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
782 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
784 2018-02-27 Richard Biener <rguenther@suse.de>
786 PR tree-optimization/84512
787 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
788 Do not use the estimate returned from record_stmt_cost for
789 the scalar iteration cost but sum properly using add_stmt_cost.
791 2018-02-27 Richard Biener <rguenther@suse.de>
793 PR tree-optimization/84466
794 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
795 Adjust last change to less strictly validate use operands.
797 2018-02-27 Martin Liska <mliska@suse.cz>
799 PR gcov-profile/84548
800 * gcov.c (process_file): Allow partial overlap and consider it
801 also as group functions.
802 (output_lines): Properly calculate range of lines for a group.
804 2018-02-27 Martin Liska <mliska@suse.cz>
806 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
807 'ggc' suffixes. Change first column width.
808 (timer::print): Fix formatting of the column.
810 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
812 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
813 preserve inline entry blocks for the sake of debug inline
814 entry point markers alone.
815 (remove_unused_locals): Suggest in comments a better place to
816 force the preservation of inline entry blocks that are
817 otherwise unused, but do not preserve them.
819 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
821 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
823 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
826 * config/i386/constraints.md (Bs): Replace
827 ix86_indirect_branch_register with
828 TARGET_INDIRECT_BRANCH_REGISTER.
830 * config/i386/i386.md (indirect_jump): Likewise.
831 (tablejump): Likewise.
832 (*sibcall_memory): Likewise.
833 (*sibcall_value_memory): Likewise.
834 Peepholes of indirect call and jump via memory: Likewise.
835 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
836 (*sibcall_value_GOT_32): Likewise.
837 * config/i386/predicates.md (indirect_branch_operand): Likewise.
838 (GOT_memory_operand): Likewise.
839 (call_insn_operand): Likewise.
840 (sibcall_insn_operand): Likewise.
841 (GOT32_symbol_operand): Likewise.
842 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
844 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
846 PR rtl-optimization/83496
847 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
848 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
849 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
850 redundant insn, if any.
851 (relax_delay_slots): Likewise.
852 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
854 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
856 PR tree-optimization/83965
857 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
858 that grouped statements are part of a reduction chain. Return
859 true if the statement is not marked as a reduction itself but
861 (vect_recog_dot_prod_pattern): Don't check whether the statement
862 is part of a group here.
863 (vect_recog_sad_pattern): Likewise.
864 (vect_recog_widen_sum_pattern): Likewise.
866 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
869 * final.c (rest_of_clean_state): Also look for calls inside sequences.
871 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
874 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
876 (ix86_output_indirect_function_return): New prototype.
877 (ix86_split_simple_return_pop_internal): Likewise.
878 * config/i386/i386.c (indirect_return_via_cx): New.
879 (indirect_return_via_cx_bnd): Likewise.
880 (indirect_thunk_name): Handle return va CX_REG.
881 (output_indirect_thunk_function): Create alias for
882 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
883 (ix86_output_indirect_jmp): Remove the bool argument.
884 (ix86_output_indirect_function_return): New function.
885 (ix86_split_simple_return_pop_internal): Likewise.
886 * config/i386/i386.md (*indirect_jump): Don't pass false
887 to ix86_output_indirect_jmp.
888 (*tablejump_1): Likewise.
889 (simple_return_pop_internal): Change it to define_insn_and_split.
890 Call ix86_split_simple_return_pop_internal to split it for
892 (simple_return_indirect_internal): Call
893 ix86_output_indirect_function_return instead of
894 ix86_output_indirect_jmp.
896 2018-02-26 Jakub Jelinek <jakub@redhat.com>
899 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
900 memset and value initialization afterwards.
902 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
904 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
906 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
909 * common/config/aarch64/aarch64-common.c
910 (aarch_option_optimization_table[]): Switch
911 off fomit-frame-pointer
913 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
914 Chung-Ju Wu <jasonwucj@gmail.com>
916 * config/nds32/nds32-multiple.md (load_multiple): Disallow
918 (store_multiple): Ditto.
920 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
922 * config.gcc: Add --with-cpu support for nds32 target.
923 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
924 * config/nds32/nds32.opt: Add -mcpu= option.
926 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
928 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
929 isel=yes): Warn for these deprecated options.
931 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
933 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
934 ISA_2_5_MASKS_EMBEDDED.
936 2018-02-23 Jakub Jelinek <jakub@redhat.com>
938 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
939 p->max as pointers rather than using iterative_hash_expr.
941 2018-02-23 Carl Love <cel@us.ibm.com>
943 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
944 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
946 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
947 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
948 P8V_BUILTIN_VEC_VUNSIGNED2.
950 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
953 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
954 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
956 * lra-constraints.c (curr_insn_transform): Set up
957 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
959 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
960 * lra-eliminations.c (spill_pseudos): Ditto.
961 (process_insn_for_elimination): Ditto.
962 * lra-lives.c (reg_early_clobber_p): Use the new macros.
963 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
964 LRA_NON_CLOBBERED_ALT.
966 2018-02-22 Martin Sebor <msebor@redhat.com>
968 PR tree-optimization/84480
969 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
970 to maybe_diag_stxncpy_trunc. Call it.
971 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
972 from gimple_fold_builtin_strcpy. Print inlining stack.
973 (handle_builtin_stxncpy): Print inlining stack.
974 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
976 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
979 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
980 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
981 and -fcheck-pointer-bounds are used together.
982 (indirect_thunk_prefix): New enum.
983 (indirect_thunk_need_prefix): New function.
984 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
985 "_nt" instead of "_bnd" for NOTRACK prefix.
986 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
987 (output_indirect_thunk_function): Likewise.
989 (ix86_code_end): Update output_indirect_thunk_function calls.
990 (ix86_output_indirect_branch_via_reg): Replace
991 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
992 (ix86_output_indirect_branch_via_push): Likewise.
993 (ix86_output_function_return): Likewise.
994 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
995 incompatible with -fcf-protection=branch and
996 -fcheck-pointer-bounds.
998 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1001 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1002 Change gcc_assert call to output_operand_lossage.
1004 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1006 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1008 2018-02-22 DJ Delorie <dj@redhat.com>
1009 Sebastian Perta <sebastian.perta@renesas.com>
1010 Oleg Endo <olegendo@gcc.gnu.org>
1012 * config/rx/rx.c (rx_rtx_costs): New function.
1013 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1015 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1017 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1019 2018-02-22 Martin Liska <mliska@suse.cz>
1022 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1023 Add "native" as a possible value.
1025 2018-02-22 Martin Liska <mliska@suse.cz>
1028 * config/i386/i386.c (ix86_option_override_internal):
1029 Add "native" as a possible value for -march and -mtune.
1031 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1034 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1035 to all type variants.
1037 PR tree-optimization/84503
1038 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1039 width as info->bitpos + info->bitsize - start.
1040 (merged_store_group::merge_overlapping): Simplify width computation.
1041 (check_no_overlap): New function.
1042 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1043 start + width and last_order of the group, fail if check_no_overlap
1045 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1046 to group if check_no_overlap fails.
1048 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1050 * config/rs6000/altivec.md: Delete contraint arguments to
1051 define_expand, define_split, and define_peephole2, and in
1052 define_insn_and_split if always unused.
1053 * config/rs6000/darwin.md: Ditto.
1054 * config/rs6000/dfp.md: Ditto.
1055 * config/rs6000/rs6000.md: Ditto.
1056 * config/rs6000/sync.md: Ditto.
1057 * config/rs6000/vector.md: Ditto.
1058 * config/rs6000/vsx.md: Ditto.
1060 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1062 * config/rs6000/altivec.md: Write output control strings as braced
1063 blocks instead of double-quoted strings.
1064 * config/rs6000/darwin.md: Ditto.
1065 * config/rs6000/rs6000.md: Ditto.
1066 * config/rs6000/vector.md: Ditto.
1067 * config/rs6000/vsx.md: Ditto.
1069 2018-02-21 Jason Merrill <jason@redhat.com>
1071 PR c++/84314 - ICE with templates and fastcall attribute.
1072 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1074 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1076 * ipa-cp.c (determine_versionability): Fix comment typos.
1078 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1081 * ipa-cp.c (determine_versionability): Do not version functions caling
1084 2018-02-21 Martin Liska <mliska@suse.cz>
1087 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1088 Add "native" as a possible value.
1089 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1090 the macro when native cpu detection is available.
1092 2018-02-21 Martin Liska <mliska@suse.cz>
1095 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1096 Add "native" as a possible value.
1097 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1098 when native cpu detection is available.
1100 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1101 Martin Sebor <msebor@redhat.com>
1103 PR tree-optimization/84478
1104 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1106 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1107 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1108 support which is conservatively correct, for 2 only stay conservative
1109 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1110 argument to the 2 argument get_range_strlen, adjust 6 arg
1111 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1113 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1114 (gimple_fold_builtin_strlen): Pass true as last argument to
1117 2018-02-20 Martin Sebor <msebor@redhat.com>
1120 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1121 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1122 (builtin_memref::builtin_memref): Factor out parts into
1123 set_base_and_offset and call it.
1125 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1128 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1129 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1130 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1131 search at the associated MODE_INT.
1133 2018-02-20 Jeff Law <law@redhat.com>
1136 PR tree-optimization/81592
1138 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1139 for range data rather than using global data.
1140 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1141 range data rather than using global data.
1142 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1143 pass it to children as needed.
1144 (struct directive::fmtresult): Similarly.
1145 (struct directive::set_width): Similarly.
1146 (struct directive::set_precision): Similarly.
1147 (format_integer, format_directive, parse_directive): Similarly.
1148 (format_none): Accept unnamed vr_values parameter.
1149 (format_percent, format_floating, format_character): Similarly.
1150 (format_string, format_plain): Similarly.
1151 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1152 the EVRP range analyzer for range data rather than using global data.
1153 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1154 gimple-ssa-evrp-analyze.h
1155 (class sprintf_dom_walker): Add after_dom_children member function.
1156 Add evrp_range_analyzer member.
1157 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1158 range analyzer as needed.
1159 (sprintf_dom_walker::after_dom_children): New member function.
1160 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1162 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1163 (evrp_range_analyzer::pop_to_marker): Likewise.
1165 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1167 PR tree-optimization/84419
1168 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1169 with the required type if its current type is compatible but
1172 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1175 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1176 after vectorization.
1178 2018-02-20 Martin Liska <mliska@suse.cz>
1181 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1182 possible values if we don't have a hint.
1184 2018-02-20 Martin Liska <mliska@suse.cz>
1188 * final.c (shorten_branches): Build align_tab array with one
1190 * opts.c (finish_options): Add alignment option limit check.
1191 (MAX_CODE_ALIGN): Likewise.
1192 (MAX_CODE_ALIGN_VALUE): Likewise.
1193 * doc/invoke.texi: Document maximum allowed option value for
1194 all -falign-* options.
1196 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1199 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1200 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1201 * var-tracking.c (emit_note_insn_var_location): Remove all references
1202 to NOTE_INSN_CALL_ARG_LOCATION.
1203 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1204 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1205 Use copy_rtx_if_shared.
1206 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1207 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1208 (dwarf2out_var_location): Remove handling of
1209 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1211 * final.c (final_scan_insn): Remove all references to
1212 NOTE_INSN_CALL_ARG_LOCATION.
1213 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1214 before dumping final insns.
1215 * except.c (emit_note_eh_region_end): Remove all references to
1216 NOTE_INSN_CALL_ARG_LOCATION.
1217 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1218 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1219 * config/arc/arc.c (hwloop_optimize): Likewise.
1220 * config/arm/arm.c (create_fix_barrier): Likewise.
1221 * config/s390/s390.c (s390_chunkify_start): Likewise.
1222 * config/sh/sh.c (find_barrier): Likewise.
1223 * config/i386/i386.c (rest_of_insert_endbranch,
1224 ix86_seh_fixup_eh_fallthru): Likewise.
1225 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1226 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1227 * config/frv/frv.c (frv_function_prologue): Likewise.
1228 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1230 (note_outside_basic_block_p): Remove all references to
1231 NOTE_INSN_CALL_ARG_LOCATION.
1232 * gengtype.c (adjust_field_rtx_def): Likewise.
1233 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1235 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1236 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1239 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1242 PR tree-optimization/84452
1243 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1244 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1247 2018-02-19 Martin Liska <mliska@suse.cz>
1250 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1252 2018-02-19 Martin Liska <mliska@suse.cz>
1253 Richard Sandiford <richard.sandiford@linaro.org>
1255 PR tree-optimization/82491
1256 * gimple-fold.c (get_base_constructor): Make earlier bail out
1259 2018-02-19 Carl Love <cel@us.ibm.com>
1261 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1262 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1264 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1265 P8V_BUILTIN_VEC_NEG.
1267 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1269 * config/rl78/rl78.md (movdf): New define expand.
1271 2018-02-19 Martin Liska <mliska@suse.cz>
1274 * doc/invoke.texi: Fix typo.
1275 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1277 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1279 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1280 handle rs6000_single_float and rs6000_double_float specially for
1283 2018-02-16 Jeff Law <law@redhat.com>
1285 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1288 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1291 * ipa-inline.c (inline_small_functions): Fix a typo.
1293 2018-02-16 Nathan Sidwell <nathan@acm.org>
1295 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1297 2018-02-16 Carl Love <cel@us.ibm.com>
1299 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1300 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1301 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1302 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1303 expansion to P8V_BUILTIN_VEC_FLOAT2.
1305 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1307 PR rtl-optimization/70023
1308 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1309 src_regno into account.
1311 2018-02-16 Carl Love <cel@us.ibm.com>
1313 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1314 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1315 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1316 * config/rs6000/rs6000.c: Remove case statements for
1317 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1318 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1319 and P9V_BUILTIN_VEC_VINSERT4B.
1320 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1321 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1322 * config/rs6000/vsx.md:
1323 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1326 2018-02-16 Carl Love <cel@us.ibm.com>
1328 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1330 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1332 * config/rs6000/rs6000-c.c: Add the definitions for
1333 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1334 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1335 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1336 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1337 definition for insert4b and define insn *insert3b_internal.
1338 * doc/extend.texi: Add documentation for vec_extract4b.
1340 2018-02-16 Nathan Sidwell <nathan@acm.org>
1342 * doc/extend.texi (Backwards Compatibility): Mention friend
1343 injection. Note for-scope is deprecated.
1344 * doc/invoke.texi (-ffriend-injection): Deprecate.
1346 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1348 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1349 that moved to I2, also allow destinations that are a paradoxical
1350 subreg (instead of a normal reg).
1352 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1355 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1358 2018-02-16 Richard Biener <rguenther@suse.de>
1360 PR tree-optimization/84037
1361 PR tree-optimization/84016
1363 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1364 Adjust vec_construct for the fact we need additional higher latency
1365 128bit inserts for AVX256 and AVX512 vector builds.
1366 (ix86_add_stmt_cost): Scale vector construction cost for
1369 2018-02-16 Richard Biener <rguenther@suse.de>
1371 PR tree-optimization/84417
1372 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1373 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1374 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1376 2018-02-16 Martin Liska <mliska@suse.cz>
1379 * internal-fn.def (ASAN_CHECK): Set proper flags.
1380 (ASAN_MARK): Likewise.
1382 2018-02-16 Julia Koval <julia.koval@intel.com>
1384 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1385 from PTA_CANNONLAKE.
1387 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1390 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1391 Use ++iter rather than iter++ for std::list iterators.
1392 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1393 defer deleting them until all nodes in the forest are processed. Do
1394 free even leaf nodes. Change to_process into auto_vec.
1397 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1398 * vec.h (vec_default_construct): Use memset instead of placement new
1399 if BROKEN_VALUE_INITIALIZATION is defined.
1400 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1401 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1404 PR rtl-optimization/83723
1405 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1406 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1407 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1409 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1411 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1413 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1415 PR rtl-optimization/81443
1416 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1417 from inner REGs to paradoxical SUBREGs.
1419 2018-02-16 Richard Biener <rguenther@suse.de>
1421 PR tree-optimization/84399
1422 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1423 For operands we can analyze at their definition make sure we can
1424 analyze them at each use as well.
1426 2018-02-16 Richard Biener <rguenther@suse.de>
1428 PR tree-optimization/84190
1429 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1430 volatile accesses if the decl isn't volatile.
1432 2018-02-15 Jason Merrill <jason@redhat.com>
1434 PR c++/84314 - ICE with templates and fastcall attribute.
1435 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1436 TYPE_CANONICAL on an existing type.
1438 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1440 PR tree-optimization/84383
1441 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1442 dstoff nor call operand_equal_p if dstbase is NULL.
1444 PR tree-optimization/84334
1445 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1446 also a CONSTANT_CLASS_P, punt.
1448 2018-02-14 Jim Wilson <jimw@sifive.com>
1450 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1451 first SMALL_OPERAND check. New local min_second_step. Move assert
1452 to where locals are set. Add TARGET_RVC support.
1453 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1455 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1457 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1459 2018-02-14 Martin Sebor <msebor@redhat.com>
1461 PR tree-optimization/83698
1462 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1463 arrays constrain the offset range to their bounds.
1464 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1465 (builtin_access::overlap): Avoid setting the size of overlap if it's
1467 (maybe_diag_overlap): Also consider arrays when deciding what values
1468 of offsets to include in diagnostics.
1470 2018-02-14 Martin Sebor <msebor@redhat.com>
1473 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1474 that correspond to the kind of a declaration.
1476 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1479 * config/pa/pa.md: Load address of PIC label using the linkage table
1480 if the label is nonlocal.
1482 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1484 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1485 warning message if user requests -maltivec=be.
1486 * doc/invoke.texi: Document deprecation of -maltivec=be.
1488 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1491 * config/rs6000/rs6000-c.c: Update definitions for
1492 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1493 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1495 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1498 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1499 add _get_ssp intrinsics. Remove argument from
1500 __builtin_ia32_rdssp[d|q].
1501 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1502 * config/i386/i386-builtin.def: Remove argument from
1503 __builtin_ia32_rdssp[d|q].
1504 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1505 ix86_expand_special_args_builtin for _rdssp[d|q].
1506 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1507 Clear register before usage.
1508 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1509 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1511 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1513 PR tree-optimization/84357
1514 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1515 operand 1 of an ARRAY_REF too.
1517 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1520 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1521 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1523 (set_of_reg): New struct.
1524 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1525 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1526 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1528 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1529 Split into bitclr, bitset, bitinvert patterns if appropriate.
1530 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1531 use rx_fuse_in_memory_bitop.
1532 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1533 to named insn, correct maximum insn length.
1535 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1538 * machmode.def: Define a complex mode for PARTIAL_INT.
1539 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1541 * doc/rtl.texi: Document CSPImode.
1542 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1544 (msp430_hard_regno_nregs_with_padding): Likewise.
1546 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1549 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1551 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1553 PR rtl-optimization/84169
1554 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1555 we generated a parallel as new i3 and we split that to new i2 and i3
1556 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1557 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1558 those to i2, not i1. Partially rewrite this scan code.
1560 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1563 * stor-layout.c (place_field): For variable length fields, adjust
1564 offset_align afterwards not just based on the field's alignment,
1565 but also on the size.
1568 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1569 of exps and logs in the use_exp2 case.
1571 2018-02-13 Jeff Law <law@redhat.com>
1573 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1576 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1579 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1583 * common.opt (gas-loc-support, gas-locview-support): New.
1584 (ginline-points, ginternal-reset-location-views): New.
1585 * doc/invoke.texi: Document them. Use @itemx where intended.
1586 (gvariable-location-views): Adjust.
1587 * target.def (reset_location_view): New.
1588 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1589 (TARGET_RESET_LOCATION_VIEW): New.
1590 * doc/tm.texi: Rebuilt.
1591 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1592 (dwarf2out_default_as_locview_support): New.
1593 (output_asm_line_debug_info): Use option variables.
1594 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1595 (output_loc_list): Likewise.
1596 (add_high_low_attributes): Check option variables.
1597 Don't output entry view attribute in strict mode.
1598 (gen_inlined_subroutine_die): Check option variables.
1599 (dwarf2out_inline_entry): Likewise.
1600 (init_sections_and_labels): Likewise.
1601 (dwarf2out_early_finish): Likewise.
1602 (maybe_reset_location_view): New, from...
1603 (dwarf2out_var_location): ... here. Call it.
1604 * debug.h (dwarf2out_default_as_loc_support): Declare.
1605 (dwarf2out_default_as_locview_support): Declare.
1606 * hooks.c (hook_int_rtx_insn_0): New.
1607 * hooks.h (hook_int_rtx_insn_0): Declare.
1608 * toplev.c (process_options): Take -gas-loc-support and
1609 -gas-locview-support from dwarf2out. Enable
1610 -gvariable-location-views by default only with locview
1611 assembler support. Enable -ginternal-reset-location-views by
1612 default only if the target defines the corresponding hook.
1613 Enable -ginline-points by default if location views are
1614 enabled; force it disabled if statement frontiers are
1616 * tree-inline.c (expand_call_inline): Check option variables.
1617 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1619 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1621 PR tree-optimization/84321
1622 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1623 handling. Also check whether the anti-range contains any values
1624 that satisfy the mask; switch to a VR_RANGE if not.
1626 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1629 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1631 2018-02-13 Martin Jambor <mjambor@suse.cz>
1634 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1635 of call statements, also set location of a load to a temporary.
1637 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1639 * config/rl78/rl78.c (add_vector_labels): New function.
1640 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1641 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1642 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1643 which checks that no arguments are passed.
1644 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1645 * doc/extend.texi: Documentation for the new attribute.
1647 2018-02-13 Andreas Schwab <schwab@suse.de>
1649 * config/riscv/linux.h (CPP_SPEC): Define.
1651 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1654 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1655 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1656 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1657 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1658 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1659 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1660 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1661 temporarily for AES and PCLMUL builtins.
1663 PR tree-optimization/84339
1664 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1665 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1669 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1670 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1671 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1673 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1675 * omp-simd-clone.h: New file.
1676 * omp-simd-clone.c: Include omp-simd-clone.h.
1677 (expand_simd_clones): No longer static.
1678 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1679 cgraph.h and omp-simd-clone.h.
1680 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1681 (vect_recog_widen_shift_pattern): Formatting fix.
1682 (vect_pattern_recog_1): Don't check optab for calls.
1685 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1686 operands[2] into a REG before using gen_lowpart on it.
1688 2018-02-12 Jeff Law <law@redhat.com>
1691 * config/sh/sh.c (find_barrier): Consider a sibling call
1694 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1695 successfully back substituting a reg.
1697 2018-02-12 Richard Biener <rguenther@suse.de>
1699 PR tree-optimization/84037
1700 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1701 parameter, move visited init to caller.
1702 (vect_slp_analyze_operations): Separate cost from validity
1703 check, initialize visited once for all instances.
1704 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1706 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1708 (vect_model_promotion_demotion_cost): Likewise.
1709 (vectorizable_bswap): Guard cost modeling with !slp_node
1710 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1712 (vectorizable_call): Likewise.
1713 (vectorizable_conversion): Likewise.
1714 (vectorizable_assignment): Likewise.
1715 (vectorizable_shift): Likewise.
1716 (vectorizable_operation): Likewise.
1717 (vectorizable_store): Likewise.
1718 (vectorizable_load): Likewise.
1719 (vectorizable_condition): Likewise.
1720 (vectorizable_comparison): Likewise.
1722 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1725 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1726 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1727 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1729 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1732 * params.def (inline-min-speedup): Increase from 8 to 15.
1733 (max-inline-insns-auto): Decrease from 40 to 30.
1734 * ipa-split.c (consider_split): Add some buffer for function to
1735 be considered inlining candidate.
1736 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1739 2018-02-12 Richard Biener <rguenther@suse.de>
1741 PR tree-optimization/84037
1742 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1743 matched stmts if we cannot swap the non-matched ones.
1745 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1747 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1748 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1749 _mm_maskz_scalef_round_ss): New intrinsics.
1750 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1751 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1752 __builtin_ia32_scalefss_round): Remove.
1753 (__builtin_ia32_scalefsd_mask_round,
1754 __builtin_ia32_scalefss_mask_round): New intrinsics.
1755 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1756 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1757 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1758 "<round_constraint>")): Changed to ...
1759 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1760 "<round_scalar_constraint>")): ... this.
1761 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1762 %0, %1, %2<round_op3>}"): Changed to ...
1763 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1764 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1765 %2<round_scalar_mask_op3>}"): ... this.
1766 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1768 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1770 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1771 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1772 (_mm_maskz_sqrt_round_ss): New intrinsics.
1773 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1774 (__builtin_ia32_sqrtsd_mask_round)
1775 (__builtin_ia32_sqrtss_mask_round): New builtins.
1776 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1777 (__builtin_ia32_sqrtss_round): Remove.
1778 (__builtin_ia32_sqrtsd_mask_round)
1779 (__builtin_ia32_sqrtss_mask_round): New builtins.
1780 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1781 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1782 ((match_operand:VF_128 1 "vector_operand"
1783 "xBm,<round_constraint>")): Changed to ...
1784 ((match_operand:VF_128 1 "vector_operand"
1785 "xBm,<round_scalar_constraint>")): ... this.
1786 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1787 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1788 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1789 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1790 %<iptr>1<round_scalar_mask_op3>}): ... this.
1791 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1792 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1794 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1797 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1798 Cast vec_cmpeq result to correct type.
1799 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1800 Cast vec_cmpgt result to correct type.
1802 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1804 * final.c (final_scan_insn_1): Renamed from...
1805 (final_scan_insn): ... this. New wrapper, to recover
1806 seen from the outermost call in recursive ones.
1807 * config/sparc/sparc.c (output_return): Drop seen from call.
1808 (output_sibcall): Likewise.
1809 * config/visium/visium.c (output_branch): Likewise.
1811 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1813 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1816 2018-02-10 Alan Modra <amodra@gmail.com>
1819 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1820 Specify LR as an input.
1822 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1825 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1826 remove_member_access_dummy_vars): New functions.
1827 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1828 lower_omp_1, execute_lower_omp): Use them.
1830 PR rtl-optimization/84308
1831 * shrink-wrap.c (spread_components): Release todo vector.
1833 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1835 PR rtl-optimization/57193
1836 * ira-color.c (struct allocno_color_data): Add member
1837 conflict_allocno_hard_prefs.
1838 (update_conflict_allocno_hard_prefs): New.
1839 (bucket_allocno_compare_func): Add a preference based on
1840 conflict_allocno_hard_prefs.
1841 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1842 (color_allocnos): Remove a dead code. Initiate
1843 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1845 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1848 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1849 constraint from =wa to wa. Avoid a subreg on the output operand,
1850 instead use a pseudo and subreg it in a move.
1851 (p9_xxbrd_<mode>): Changed to ...
1852 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1853 (p9_xxbrd_v2df): New expander.
1854 (p9_xxbrw_<mode>): Changed to ...
1855 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1856 (p9_xxbrw_v4sf): New expander.
1858 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1860 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1862 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1865 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1866 multiply in 32-bit mode.
1867 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1868 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1871 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1873 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1874 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1875 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1876 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1878 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1881 * dwarf2out.c (is_trivial_indirect_ref): New function.
1882 (dwarf2out_late_global_decl): Do not generate a location
1883 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1884 and that are not defined in the current unit.
1886 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1888 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1889 instead of a libcall for UNORDERED.
1891 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1894 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1895 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1897 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1900 * config/s390/s390.c (s390_set_current_function): Invoke
1901 s390_indirect_branch_settings also if fndecl didn't change.
1903 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1905 * config/rs6000/rs6000.md (blockage): Set length to zero.
1907 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1909 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1911 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1914 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1915 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1919 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1920 PARALLEL incoming that failed vt_get_decl_and_offset check.
1923 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1924 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1926 (get_variable_section): For decls in named .bss* sections pass true as
1927 second argument to bss_initializer_p.
1929 2018-02-09 Marek Polacek <polacek@redhat.com>
1930 Jakub Jelinek <jakub@redhat.com>
1933 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1934 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1935 Sync some changes from cxx_fold_indirect_ref.
1937 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1939 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1941 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1942 (BLOCK_INLINE_ENTRY_LABEL): New.
1943 (dwarf2out_var_location): Disregard inline entry markers.
1944 (inline_entry_data): New struct.
1945 (inline_entry_data_hasher): New hashtable type.
1946 (inline_entry_data_hasher::hash): New.
1947 (inline_entry_data_hasher::equal): New.
1948 (inline_entry_data_table): New variable.
1949 (add_high_low_attributes): Add DW_AT_entry_pc and
1950 DW_AT_GNU_entry_view attributes if a pending entry is found
1951 in inline_entry_data_table. Add old entry_pc attribute only
1952 if debug nonbinding markers are disabled.
1953 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1954 markers are enabled.
1955 (block_within_block_p, dwarf2out_inline_entry): New.
1956 (dwarf2out_finish): Check that no entries remained in
1957 inline_entry_data_table.
1958 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1959 (final_scan_insn, notice_source_line): Likewise.
1960 (rest_of_clean_state): Skip inline entry markers.
1961 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1963 * gimple.c (gimple_build_debug_inline_entry): New.
1964 * gimple.h (enum gimple_debug_subcode): Add
1965 GIMPLE_DEBUG_INLINE_ENTRY.
1966 (gimple_build_debug_inline_entry): Declare.
1967 (gimple_debug_inline_entry_p): New.
1968 (gimple_debug_nonbind_marker_p): Adjust.
1969 * insn-notes.def (INLINE_ENTRY): New.
1970 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1971 inline entry marker notes.
1972 (print_insn): Likewise.
1973 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1974 (INSN_DEBUG_MARKER_KIND): Likewise.
1975 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1976 * tree-inline.c (expand_call_inline): Build and insert
1977 debug_inline_entry stmt.
1978 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1979 inline entry blocks early, if nonbind markers are enabled.
1980 (dump_scope_block): Dump fragment info.
1981 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1982 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1983 (gimple_build_debug_inline_entry): New.
1984 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1985 Enable/disable inline entry points too.
1986 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1987 (DEBUG_INSN): Describe inline entry markers.
1989 * common.opt (gvariable-location-views): New.
1990 (gvariable-location-views=incompat5): New.
1991 * config.in: Rebuilt.
1992 * configure: Rebuilt.
1993 * configure.ac: Test assembler for view support.
1994 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1995 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1996 * dwarf2out.c (var_loc_view): New typedef.
1997 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1998 (dwarf2out_locviews_in_attribute): New.
1999 (dwarf2out_locviews_in_loclist): New.
2000 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2001 (enum dw_line_info_opcode): Add LI_adv_address.
2002 (struct dw_line_info_table): Add view.
2003 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2004 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2005 (zero_view_p): New variable.
2006 (ZERO_VIEW_P): New macro.
2007 (output_asm_line_debug_info): New.
2008 (struct var_loc_node): Add view.
2009 (add_AT_view_list, AT_loc_list): New.
2010 (add_var_loc_to_decl): Add view param. Test it against last.
2011 (new_loc_list): Add view params. Record them.
2012 (AT_loc_list_ptr): Handle loc and view lists.
2013 (view_list_to_loc_list_val_node): New.
2014 (print_dw_val): Handle dw_val_class_view_list.
2015 (size_of_die): Likewise.
2016 (value_format): Likewise.
2017 (loc_list_has_views): New.
2018 (gen_llsym): Set vl_symbol too.
2019 (maybe_gen_llsym, skip_loc_list_entry): New.
2020 (dwarf2out_maybe_output_loclist_view_pair): New.
2021 (output_loc_list): Output view list or entries too.
2022 (output_view_list_offset): New.
2023 (output_die): Handle dw_val_class_view_list.
2024 (output_dwarf_version): New.
2025 (output_compilation_unit_header): Use it.
2026 (output_skeleton_debug_sections): Likewise.
2027 (output_rnglists, output_line_info): Likewise.
2028 (output_pubnames, output_aranges): Update version comments.
2029 (output_one_line_info_table): Output view numbers in asm comments.
2030 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2031 Call maybe_gen_llsym.
2032 (loc_list_from_tree_1): Adjust.
2033 (add_AT_location_description): Create view list attribute if
2034 needed, check it's absent otherwise.
2035 (convert_cfa_to_fb_loc_list): Adjust.
2036 (maybe_emit_file): Call output_asm_line_debug_info for test.
2037 (dwarf2out_var_location): Reset views as needed. Precompute
2038 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2039 attribute. Set view.
2040 (new_line_info_table): Reset next view.
2041 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2042 (dwarf2out_source_line): Likewise. Output view resets and labels to
2043 the assembler, or select appropriate line info opcodes.
2044 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2045 (optimize_string_length): Catch it. Adjust.
2046 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2047 dw_val_class_view_list, and remove it if no longer needed.
2048 (hash_loc_list): Hash view numbers.
2049 (loc_list_hasher::equal): Compare them.
2050 (optimize_location_lists): Check whether a view list symbol is
2051 needed, and whether the locview attribute is present, and
2052 whether they match. Remove the locview attribute if no longer
2054 (index_location_lists): Call skip_loc_list_entry for test.
2055 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2056 Use output_dwarf_version.
2057 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2058 (struct dw_val_node): Add val_view_list.
2059 * final.c (SEEN_NEXT_VIEW): New.
2060 (set_next_view_needed): New.
2061 (clear_next_view_needed): New.
2062 (maybe_output_next_view): New.
2063 (final_start_function): Rename to...
2064 (final_start_function_1): ... this. Take pointer to FIRST,
2065 add SEEN parameter. Emit param bindings in the initial view.
2066 (final_start_function): Reintroduce SEEN-less interface.
2067 (final): Rename to...
2068 (final_1): ... this. Take SEEN parameter. Output final pending
2069 next view at the end.
2070 (final): Reintroduce seen-less interface.
2071 (final_scan_insn): Output pending next view before switching
2072 sections or ending a block. Mark the next view as needed when
2073 outputting variable locations. Notify debug backend of section
2074 changes, and of location view changes.
2075 (rest_of_handle_final): Adjust.
2076 * toplev.c (process_options): Autodetect value for debug variable
2077 location views option. Warn on incompat5 without -gdwarf-5.
2078 * doc/invoke.texi (gvariable-location-views): New.
2079 (gvariable-location-views=incompat5): New.
2080 (gno-variable-location-views): New.
2082 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2084 PR tree-optimization/84136
2085 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2086 that the result of find_edge is non-NULL.
2088 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2091 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2092 storing integer register in SImode. Fix cost of 256 and 512
2093 byte aligned SSE register store.
2095 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2097 * config/i386/i386.c (ix86_multiplication_cost): Fix
2098 multiplication cost for TARGET_AVX512DQ.
2100 2018-02-08 Marek Polacek <polacek@redhat.com>
2102 PR tree-optimization/84238
2103 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2106 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2108 PR tree-optimization/84265
2109 * tree-vect-stmts.c (vectorizable_store): Don't treat
2110 VMAT_CONTIGUOUS accesses as grouped.
2111 (vectorizable_load): Likewise.
2113 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2115 PR tree-optimization/81635
2116 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2117 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2118 (test_round_for_mask): New functions.
2119 (wide_int_cc_tests): Call test_round_for_mask.
2120 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2121 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2122 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2123 range returned by get_range_info.
2125 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2128 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2129 * symtab.c: Include builtins.h
2130 (symtab_node::output_to_lto_symbol_table_p): Move here
2131 from lto-streamer-out.c:output_symbol_p.
2132 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2133 (output_symbol_p): Move all logic to symtab.c
2134 (produce_symtab): Update.
2136 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2138 * config/s390/s390-opts.h (enum indirect_branch): Define.
2139 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2140 (s390_indirect_branch_via_thunk)
2141 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2142 (enum s390_indirect_branch_type): Define.
2143 * config/s390/s390.c (struct s390_frame_layout, struct
2144 machine_function): Remove.
2145 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2146 (indirect_branch_table_label_no, indirect_branch_table_name):
2148 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2149 (enum s390_indirect_branch_option): Define.
2150 (s390_return_addr_from_memory): New function.
2151 (s390_handle_string_attribute): New function.
2152 (s390_attribute_table): Add new attribute handler.
2153 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2154 (s390_indirect_branch_via_thunk): New function.
2155 (s390_indirect_branch_via_inline_thunk): New function.
2156 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2157 sibling call optimization for non z10 compiles.
2158 (s390_emit_call): Force indirect branch target to be a single
2159 register. Add r1 clobber for non-z10 compiles.
2160 (s390_emit_epilogue): Emit return jump via return_use expander.
2161 (s390_reorg): Handle JUMP_INSNs as execute targets.
2162 (s390_option_override_internal): Perform validity checks for the
2163 new command line options.
2164 (s390_indirect_branch_attrvalue): New function.
2165 (s390_indirect_branch_settings): New function.
2166 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2167 (s390_output_indirect_thunk_function): New function.
2168 (s390_code_end): Implement target hook.
2169 (s390_case_values_threshold): Implement target hook.
2170 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2172 * config/s390/s390.h (struct s390_frame_layout)
2173 (struct machine_function): Move here from s390.c.
2174 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2175 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2176 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2177 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2178 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2179 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2180 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2181 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2182 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2183 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2184 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2185 (mnemonic attribute): Add values which aren't recognized
2187 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2188 pattern for branch conversion. Fix mnemonic attribute.
2189 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2190 indirect branch via thunk if requested.
2191 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2192 ("*indirect_jump"): Disable for branch conversion using out of
2194 ("indirect_jump_via_thunk<mode>_z10")
2195 ("indirect_jump_via_thunk<mode>")
2196 ("indirect_jump_via_inlinethunk<mode>_z10")
2197 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2198 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2199 ("casesi_jump_via_inlinethunk<mode>_z10")
2200 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2201 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2202 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2203 ("*indirect2_jump"): Disable for branch conversion.
2204 ("casesi_jump"): Turn into expander and expand patterns for branch
2206 ("return_use"): New expander.
2207 ("*return"): Emit return via thunk and rename it to ...
2208 ("*return<mode>"): ... this one.
2209 * config/s390/s390.opt: Add new options and and enum for the
2212 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2214 * lra-constraints.c (match_reload): Unconditionally use
2215 gen_lowpart_SUBREG, rather than selecting between that
2216 and equivalent gen_rtx_SUBREG code.
2218 2018-02-08 Richard Biener <rguenther@suse.de>
2220 PR tree-optimization/84233
2221 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2222 changed flag instead of boguously re-using phi_inserted.
2224 2018-02-08 Martin Jambor <mjambor@suse.cz>
2226 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2227 static local variables.
2229 2018-02-08 Richard Biener <rguenther@suse.de>
2231 PR tree-optimization/84278
2232 * tree-vect-stmts.c (vectorizable_store): When looking for
2233 smaller vector types to perform grouped strided loads/stores
2234 make sure the mode is supported by the target.
2235 (vectorizable_load): Likewise.
2237 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2239 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2240 Increase LDP/STP opportunities by adding adjacent callee-saves.
2242 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2244 PR rtl-optimization/84068
2245 PR rtl-optimization/83459
2246 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2248 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2250 PR tree-optimization/84224
2251 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2252 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2255 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2258 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2259 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2260 count, remove one USE.
2262 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2264 * doc/install.texi (Configuration): Document the
2265 --with-long-double-format={ibm,ieee} PowerPC configuration
2269 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2270 Convert from define_expand to be define_insn_and_split. Rework
2271 float/double/_Float128 conversions to QI/HI/SImode to work with
2272 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2273 conversions to QI/HImode types did a store and then a load to
2274 truncate the value. For conversions to VSX registers, don't split
2275 the insn, instead emit the code directly. Use the code iterator
2276 any_fix to combine signed and unsigned conversions.
2277 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2278 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2279 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2280 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2281 (fix_<mode>di2_hw): Likewise.
2282 (fixuns_<mode>di2_hw): Likewise.
2283 (fix_<mode>si2_hw): Likewise.
2284 (fixuns_<mode>si2_hw): Likewise.
2285 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2286 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2287 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2288 fix<uns>_trunc<SFDF:mode>si2_p8.
2289 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2291 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2292 (fix<uns>_<mode>_mem): Likewise.
2293 (fctiw<u>z_<mode>_mem): Likewise.
2294 (fix<uns>_<mode>_mem): Likewise.
2295 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2296 the register allocator from doing a direct move to the GPRs to do
2297 a store, and instead use the ISA 3.0 store byte/half-word from
2298 vector register instruction. For IEEE 128-bit floating point,
2299 also optimize stores of 32-bit ints.
2300 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2302 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2304 * genextract.c (push_pathstr_operand): New function to support
2306 (walk_rtx): Call push_pathstr_operand.
2307 (print_path): Support [a-zA-Z].
2309 2018-02-07 Richard Biener <rguenther@suse.de>
2311 PR tree-optimization/84037
2312 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2313 (cse_and_gimplify_to_preheader): Declare.
2314 (vect_get_place_in_interleaving_chain): Likewise.
2315 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2317 (_loop_vec_info::~_loop_vec_info): Delete it.
2318 (cse_and_gimplify_to_preheader): New function.
2319 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2320 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2321 (vectorizable_load): Likewise. For grouped stores always base
2322 the IV on the first element.
2323 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2324 condition before gimplifying.
2326 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2328 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2329 *DIV_EXPR and *MOD_EXPR.
2331 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2334 * config/i386/i386.c (ix86_option_override_internal): Mask out
2335 the CF_SET bit when checking -fcf-protection.
2337 2018-02-07 Tom de Vries <tom@codesourcery.com>
2340 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2343 2018-02-07 Richard Biener <rguenther@suse.de>
2345 PR tree-optimization/84204
2346 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2349 PR tree-optimization/84205
2350 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2351 special-case isl_ast_op_zdiv_r.
2353 PR tree-optimization/84223
2354 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2355 Only add conditions from within the region.
2356 (gather_bbs::after_dom_children): Adjust.
2358 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2361 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2362 * config/avr/avr.md: Only post-reload split REG-REG moves if
2363 either register is GENERAL_REG_P.
2365 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2367 PR tree-optimization/84235
2368 * tree-ssa-scopedtables.c
2369 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2370 if the subtraction is performed in floating point type where NaNs are
2371 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2372 build 1. Formatting fix.
2374 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2377 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2378 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2379 and skip it regardless of bb boundaries. Use CALL_P macro,
2380 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2381 unnecessarily, formatting fix.
2383 2018-02-06 Michael Collison <michael.collison@arm.com>
2385 * config/arm/thumb2.md:
2386 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2387 (*thumb_mov_notscc): Ditto.
2389 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2392 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2393 unsigned_fix, not "s".
2395 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2397 * configure.ac (gcc_fn_eh_frame_ro): New function.
2398 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2399 correct .eh_frame permissions.
2400 * configure: Regenerate.
2402 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2404 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2407 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2409 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2410 Display warning message for -mno-speculate-indirect-jumps.
2412 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2414 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2416 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2418 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2420 PR tree-optimization/84225
2421 * tree-eh.c (find_trapping_overflow): Only call
2422 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2424 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2427 * config/i386/i386.c: Reimplement the check of possible options
2428 -mibt/-mshstk conbination. Change error messages.
2429 * doc/invoke.texi: Fix a typo: remove extra '='.
2431 2018-02-06 Marek Polacek <polacek@redhat.com>
2433 PR tree-optimization/84228
2434 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2436 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2439 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2440 emitted arch directives.
2441 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2442 __ARM_FEATURE_COPROC before changing architectures.
2444 2018-02-06 Richard Biener <rguenther@suse.de>
2446 * config/i386/i386.c (print_reg): Fix typo.
2447 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2449 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2451 * configure: Regenerate.
2453 2018-02-05 Martin Sebor <msebor@redhat.com>
2455 PR tree-optimization/83369
2456 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2459 2018-02-05 Martin Liska <mliska@suse.cz>
2461 * doc/invoke.texi: Cherry-pick upstream r323995.
2463 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2465 * ira.c (ira_init_register_move_cost): Adjust comment.
2467 2018-02-05 Martin Liska <mliska@suse.cz>
2469 PR gcov-profile/84137
2470 * doc/gcov.texi: Fix typo in documentation.
2472 2018-02-05 Martin Liska <mliska@suse.cz>
2474 PR gcov-profile/83879
2475 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2476 using dlopen functionality.
2478 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2480 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2481 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2482 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2483 _mm_maskz_range_round_ss): New intrinsics.
2484 (__builtin_ia32_rangesd128_round)
2485 (__builtin_ia32_rangess128_round): Remove.
2486 (__builtin_ia32_rangesd128_mask_round,
2487 __builtin_ia32_rangess128_mask_round): New builtins.
2488 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2489 __builtin_ia32_rangess128_round): Remove.
2490 (__builtin_ia32_rangesd128_mask_round,
2491 __builtin_ia32_rangess128_mask_round): New builtins.
2492 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2493 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2494 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2495 "<round_saeonly_constraint>")): Changed to ...
2496 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2497 "<round_saeonly_scalar_constraint>")): ... this.
2498 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2499 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2500 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2501 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2502 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2504 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2506 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2508 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2509 Remove all values except native, 8540 and 8548.
2511 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2513 * config/i386/i386.c (ix86_output_function_return): Pass
2514 INVALID_REGNUM, instead of -1, as invalid register number to
2515 indirect_thunk_name and output_indirect_thunk.
2517 2018-02-02 Julia Koval <julia.koval@intel.com>
2519 * config.gcc: Add -march=icelake.
2520 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2521 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2522 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2523 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2524 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2525 (processor_target_table): Add icelake.
2526 (ix86_option_override_internal): Handle new PTAs.
2527 (get_builtin_code_for_version): Handle icelake.
2528 (M_INTEL_COREI7_ICELAKE): New.
2529 (fold_builtin_cpu): Handle icelake.
2530 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2531 * doc/invoke.texi: Add -march=icelake.
2533 2018-02-02 Julia Koval <julia.koval@intel.com>
2535 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2536 to wide_int_bitmask.
2537 * wide-int-bitmask.h: New.
2539 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2542 * config/i386/i386.md: Replace Pmode with word_mode in
2543 builtin_setjmp_setup and builtin_longjmp to support x32.
2545 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2549 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2551 (rs6000_supported_cpu_names): New static variable.
2552 (linux_cpu_translation_table): Likewise.
2553 (elf_platform) <cpu>: Define new static variable and use it.
2554 Translate kernel AT_PLATFORM name to canonical name if needed.
2555 Error if platform name is unknown.
2557 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2560 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2562 2018-02-01 Jeff Law <law@redhat.com>
2565 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2566 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2567 the scratch if RELEASE_VIA_POP is false.
2568 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2569 If we have to save a temporary register, decrement SIZE appropriately.
2570 Pass new arguments to release_scratch_register_on_entry.
2571 (ix86_adjust_stack_and_probe): Likewise.
2572 (ix86_emit_probe_stack_range): Pass new arguments to
2573 release_scratch_register_on_entry.
2575 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2577 PR rtl-optimization/84157
2578 * combine.c (change_zero_ext): Use REG_P predicate in
2579 front of HARD_REGISTER_P predicate.
2581 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2583 * config/avr/avr.c (avr_option_override): Move disabling of
2584 -fdelete-null-pointer-checks to...
2585 * common/config/avr/avr-common.c (avr_option_optimization_table):
2588 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2590 PR tree-optimization/81635
2591 * tree-data-ref.c (split_constant_offset_1): For types that
2592 wrap on overflow, try to use range info to prove that wrapping
2595 2018-02-01 Renlin Li <renlin.li@arm.com>
2598 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2600 (aarch64_register_move_cost): Likewise.
2601 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2603 (REG_CLASS_NAMES): Likewise.
2604 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2605 TAILCALL_ADDR_REGS. Remove IP registers.
2606 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2608 2018-02-01 Richard Biener <rguenther@suse.de>
2610 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2611 for specifying RPO order and allow NULL for that.
2612 * domwalk.c (dom_walker::dom_walker): Likewise.
2613 (dom_walker::walk): Handle NULL RPO order.
2614 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2616 (rewrite_update_dom_walker): Likewise.
2617 (mark_def_dom_walker): Likewise.
2619 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2621 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2622 (aarch64_maybe_expand_sve_subreg_move): Declare.
2623 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2624 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2626 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2627 that are semantically a reverse operation.
2628 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2629 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2630 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2632 (aarch64_can_change_mode_class): For big-endian, forbid changes
2633 between two SVE modes if they have different element sizes.
2635 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2637 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2638 the TImode handling for big-endian targets.
2640 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2642 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2643 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2645 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2646 Remove BSWAP handing for big-endian targets and use the form of
2647 LD1RQ appropariate for the mode.
2649 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2651 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2652 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2655 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2658 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2659 check for operands that need to go through aarch64_sve_reload_be.
2661 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2663 PR tree-optimization/81661
2664 PR tree-optimization/84117
2665 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2666 * tree-eh.c: Include gimplify.h.
2667 (find_trapping_overflow, replace_trapping_overflow,
2668 rewrite_to_non_trapping_overflow): New functions.
2669 * tree-vect-loop.c: Include tree-eh.h.
2670 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2671 * tree-data-ref.c: Include tree-eh.h.
2672 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2674 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2676 PR rtl-optimization/84123
2677 * combine.c (change_zero_ext): Check if hard register satisfies
2678 can_change_dest_mode before calling gen_lowpart_SUBREG.
2680 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2683 * ira.c (ira_init_register_move_cost): Remove assert.
2685 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2687 PR rtl-optimization/84071
2688 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2689 * doc/tm.texi: Regenerate.
2691 2018-01-31 Richard Biener <rguenther@suse.de>
2693 PR tree-optimization/84132
2694 * tree-data-ref.c (analyze_miv_subscript): Properly
2695 check whether evolution_function_is_affine_multivariate_p
2696 before calling gcd_of_steps_may_divide_p.
2698 2018-01-31 Julia Koval <julia.koval@intel.com>
2701 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2702 * config/i386/i386.md (rdpid_rex64) New.
2703 (rdpid): Make 32bit only.
2705 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2708 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2709 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2711 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2714 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2716 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2718 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2720 PR rtl-optimization/84071
2721 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2722 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2724 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2726 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2727 (arc_attribute_table): Add 'aux' attribute.
2728 (arc_in_small_data_p): Consider aux like variables.
2729 (arc_is_aux_reg_p): New function.
2730 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2731 (arc_get_aux_arg): New function.
2732 (prepare_move_operands): Handle aux-register access.
2733 (arc_handle_aux_attribute): New function.
2734 * doc/extend.texi (ARC Variable attributes): Add subsection.
2736 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2738 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2739 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2740 (arc_attribute_table): Add 'uncached' attribute.
2741 (arc_print_operand): Print '.di' flag for uncached memory
2743 (arc_in_small_data_p): Do not consider for small data the uncached
2745 (arc_is_uncached_mem_p): New function.
2746 * config/arc/predicates.md (compact_store_memory_operand): Check
2747 for uncached memory accesses.
2748 (nonvol_nonimm_operand): Likewise.
2749 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2751 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2754 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2755 falign-loops=): Add Optimization flag.
2757 2018-01-30 Jeff Law <law@redhat.com>
2760 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2761 INT_REGISTERS_SAVED. Check it prior to calling
2762 get_scratch_register_on_entry.
2763 (ix86_adjust_stack_and_probe): Similarly.
2764 (ix86_emit_probe_stack_range): Similarly.
2765 (ix86_expand_prologue): Corresponding changes.
2767 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2770 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2771 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2773 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2776 * lra-constraints.c (curr_insn_transform): Process AND in the
2779 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2781 PR rtl-optimization/83986
2782 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2783 dependence against last_pending_memory_flush in addition to
2786 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2788 PR tree-optimization/81611
2789 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2792 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2795 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2798 2018-01-30 Richard Biener <rguenther@suse.de>
2799 Jakub Jelinek <jakub@redhat.com>
2801 PR tree-optimization/84111
2802 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2803 inner loops added during recursion, as they don't have up-to-date
2806 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2809 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2810 (can_inline_edge_by_limits_p): ... here.
2811 (can_early_inline_edge_p, check_callers,
2812 update_caller_keys, update_callee_keys, recursive_inlining,
2813 add_new_edges_to_heap, speculation_useful_p,
2814 inline_small_functions,
2815 inline_small_functions, flatten_function,
2816 inline_to_all_callers_1): Update.
2818 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2820 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2823 2018-01-30 Richard Biener <rguenther@suse.de>
2825 PR tree-optimization/83008
2826 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2827 invariant and constant vector uses in stmts when they need
2830 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2833 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2834 * configure: Regenerate.
2836 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2838 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2840 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2841 Use gen_rtx_REG rather than gen_lowpart.
2843 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2845 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2846 rather than 0 when creating partial subregs.
2848 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2850 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2853 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2856 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2857 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2858 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2859 flags. This restores the settings used before the 2017-07-24.
2860 Turning off pre increment/decrement/modify allows IVOPTS to
2861 optimize DF/SF loops where the index is an int.
2863 2018-01-29 Richard Biener <rguenther@suse.de>
2864 Kelvin Nilsen <kelvin@gcc.gnu.org>
2867 * tree-vect-stmts.c (vectorizable_call): Don't call
2868 targetm.vectorize_builtin_md_vectorized_function if callee is
2871 2018-01-22 Carl Love <cel@us.ibm.com>
2873 * doc/extend.tex: Fix typo in second arg in
2874 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2876 2018-01-29 Richard Biener <rguenther@suse.de>
2878 PR tree-optimization/84086
2879 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2880 (flush_ssaname_freelist): When SSA names were released reset
2881 the SCEV hash table.
2883 2018-01-29 Richard Biener <rguenther@suse.de>
2885 PR tree-optimization/84057
2886 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2887 removed paths when removing edges.
2889 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2891 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2892 -mfunction-return=@var{choice}.
2894 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2897 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2899 (layout::print_source_line): Likewise.
2900 (test_get_line_width_without_trailing_whitespace): Add test cases.
2902 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2905 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2908 2018-01-26 Jim Wilson <jimw@sifive.com>
2910 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2912 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2915 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2917 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2918 and CMP + SUB-immediate -> SUBS.
2920 2018-01-26 Martin Sebor <msebor@redhat.com>
2922 PR tree-optimization/83896
2923 * tree-ssa-strlen.c (get_string_len): Rename...
2924 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2925 Avoid assuming length is constant.
2926 (handle_char_store): Use HOST_WIDE_INT for string length.
2928 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2931 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2932 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2934 2018-01-26 Richard Biener <rguenther@suse.de>
2936 PR rtl-optimization/84003
2937 * dse.c (record_store): Only record redundant stores when
2938 the earlier store aliases at least all accesses the later one does.
2940 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2942 PR rtl-optimization/83985
2943 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2944 REG_CFA_RESTORE insns.
2945 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2946 REG_CFA_RESTORE insns here.
2949 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2950 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2952 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2954 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2955 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2956 (arc_init): Likewise.
2957 (arc_override_options): Likewise.
2958 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2960 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2962 * config/arc/arc.h (TARGET_DBNZ): Define.
2963 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2964 properly set the tune attribute.
2965 (dbnz): Use TARGET_DBNZ guard.
2966 * config/arc/arc.opt (mtune): Add core3 option.
2968 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2970 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2971 recognize new pic like addresses.
2972 (arc_delegitimize_address): Clean up.
2974 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2976 * config/arc/arc-arches.def: Option mrf16 valid for all
2978 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2979 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2980 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2981 * config/arc/arc-tables.opt: Regenerate.
2982 * config/arc/arc.c (arc_conditional_register_usage): Handle
2983 reduced register file case.
2984 (arc_file_start): Set must have build attributes.
2985 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2987 * config/arc/arc.opt (mrf16): Add new option.
2988 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2989 * config/arc/genmultilib.awk: Handle new mrf16 option.
2990 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2991 * config/arc/t-multilib: Regenerate.
2992 * doc/invoke.texi (ARC Options): Document mrf16 option.
2994 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2996 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2997 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2998 (arc_attribute_table): Add 'secure_call' attribute.
2999 (arc_print_operand): Print secure call operand.
3000 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3002 (arc_is_secure_call_p): New function. * config/arc/arc.md
3003 (call_i): Add support for sjli instruction.
3004 (call_value_i): Likewise.
3005 * config/arc/constraints.md (Csc): New constraint.
3007 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3008 John Eric Martin <John.Martin@emmicro-us.com>
3010 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3011 * config/arc/arc.c (_arc_jli_section): New struct.
3012 (arc_jli_section): New type.
3013 (rc_jli_sections): New static variable.
3014 (arc_handle_jli_attribute): New function.
3015 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3016 (arc_file_end): New function.
3017 (TARGET_ASM_FILE_END): Define.
3018 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3019 (arc_add_jli_section): New function.
3020 (jli_call_scan): Likewise.
3021 (arc_reorg): Call jli_call_scan.
3022 (arc_output_addsi): Remove 'S' from printing asm operand.
3023 (arc_is_jli_call_p): New function.
3024 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3026 (movhi_insn): Likewise.
3027 (movsi_insn): Likewise.
3028 (movsi_set_cc_insn): Likewise.
3029 (loadqi_update): Likewise.
3030 (load_zeroextendqisi_update): Likewise.
3031 (load_signextendqisi_update): Likewise.
3032 (loadhi_update): Likewise.
3033 (load_zeroextendhisi_update): Likewise.
3034 (load_signextendhisi_update): Likewise.
3035 (loadsi_update): Likewise.
3036 (loadsf_update): Likewise.
3037 (movsicc_insn): Likewise.
3038 (bset_insn): Likewise.
3039 (bxor_insn): Likewise.
3040 (bclr_insn): Likewise.
3041 (bmsk_insn): Likewise.
3042 (bicsi3_insn): Likewise.
3043 (cmpsi_cc_c_insn): Likewise.
3044 (movsi_ne): Likewise.
3045 (movsi_cond_exec): Likewise.
3046 (clrsbsi2): Likewise.
3057 (call_i): Remove 'S' asm letter, add jli instruction.
3058 (call_value_i): Likewise.
3059 * config/arc/arc.op (mjli-always): New option.
3060 * config/arc/constraints.md (Cji): New constraint.
3061 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3063 (subsf3_fpx): Likewise.
3064 (mulsf3_fpx): Likewise.
3065 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3067 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3069 * doc/invoke.texi (ARC): Document mjli-always option.
3071 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3073 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3074 avoid addition with 0 and use incw and decw where possible.
3076 2018-01-26 Richard Biener <rguenther@suse.de>
3078 PR tree-optimization/81082
3079 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3080 association if it requires casting to unsigned.
3081 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3082 from fold_plusminus_mult_expr to catch important cases late when
3083 range info is available.
3085 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3087 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3088 * configure.ac (hidden_linkonce): New test.
3089 * configure: Regenerate.
3090 * config.in: Regenerate.
3092 2018-01-26 Julia Koval <julia.koval@intel.com>
3094 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3095 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3096 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3097 _mm_mask_bitshuffle_epi64_mask): Fix type.
3098 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3099 USI_FTYPE_V4DI_V4DI_USI): Remove.
3100 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3101 __builtin_ia32_vpshufbitqmb256_mask,
3102 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3103 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3104 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3106 2018-01-26 Alan Modra <amodra@gmail.com>
3109 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3110 UNSPEC_VBPERMQ. Sort other unspecs.
3112 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3114 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3116 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3119 * predict.c (drop_profile): Do not push/pop cfun; update also
3121 (handle_missing_profiles): Fix logic looking for zero profiles.
3123 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3126 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3127 on functions with #pragma omp declare simd or functions with simd
3129 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3130 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3131 Remove trailing \n from warning_at calls.
3133 2018-01-25 Tom de Vries <tom@codesourcery.com>
3136 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3137 for neutered workers.
3139 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3142 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3143 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3145 2018-01-24 Jeff Law <law@redhat.com>
3148 * i386.c (get_probe_interval): Move to earlier point.
3149 (ix86_compute_frame_layout): If -fstack-clash-protection and
3150 the frame is larger than the probe interval, then use pushes
3151 to save registers rather than reg->mem moves.
3152 (ix86_expand_prologue): Remove conditional for int_registers_saved
3155 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3158 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3159 min/max for never referenced object.
3161 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3164 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3166 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3167 attributes from DECL_ATTRIBUTES (decl) without affecting
3168 DECL_ATTRIBUTES (current_function_decl).
3169 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3170 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3172 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3174 PR tree-optimization/83979
3175 * fold-const.c (fold_comparison): Use constant_boolean_node
3176 instead of boolean_{true,false}_node.
3178 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3180 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3183 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3185 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3186 Simplify the clause that sets the length attribute.
3187 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3188 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3189 clause that sets the length attribute.
3190 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3192 2018-01-24 Tom de Vries <tom@codesourcery.com>
3195 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3196 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3197 Add strict parameter.
3198 (prevent_branch_around_nothing): Insert dummy insn between branch to
3199 label and label with no ptx insn inbetween.
3200 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3202 2018-01-24 Tom de Vries <tom@codesourcery.com>
3205 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3206 for neutered threads in warp.
3207 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3209 2018-01-24 Richard Biener <rguenther@suse.de>
3211 PR tree-optimization/83176
3212 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3215 2018-01-24 Richard Biener <rguenther@suse.de>
3217 PR tree-optimization/82819
3218 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3219 code generating pluses that are no-ops in the target precision.
3221 2018-01-24 Richard Biener <rguenther@suse.de>
3224 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3226 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3228 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3229 to merge probabilities.
3230 * predict.c (probably_never_executed): Also mark as cold functions
3231 with global 0 profile and guessed local profile.
3232 * profile-count.c (profile_probability::combine_with_count): New
3234 * profile-count.h (profile_probability::operator*,
3235 profile_probability::operator*=, profile_probability::operator/,
3236 profile_probability::operator/=): Reduce precision to adjusted
3237 and set value to guessed on contradictory divisions.
3238 (profile_probability::combine_with_freq): Remove.
3239 (profile_probability::combine_wiht_count): Declare.
3240 (profile_count::force_nonzero):: Set to adjusted.
3241 (profile_count::probability_in):: Set quality to adjusted.
3242 * tree-ssa-tail-merge.c (replace_block_by): Use
3245 2018-01-23 Andrew Waterman <andrew@sifive.com>
3246 Jim Wilson <jimw@sifive.com>
3248 * config/riscv/riscv.c (riscv_stack_boundary): New.
3249 (riscv_option_override): Set riscv_stack_boundary. Handle
3250 riscv_preferred_stack_boundary_arg.
3251 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3252 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3253 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3254 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3255 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3256 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3258 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3261 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3262 of struct ix86_frame.
3263 (ix86_expand_epilogue): Likewise. Add a local variable for
3264 the reg_save_offset field in struct ix86_frame.
3266 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3268 PR tree-optimization/82604
3269 * tree-loop-distribution.c (enum partition_kind): New enum item
3270 PKIND_PARTIAL_MEMSET.
3271 (partition_builtin_p): Support above new enum item.
3272 (generate_code_for_partition): Ditto.
3273 (compute_access_range): Differentiate cases that equality can be
3274 proven at all loops, the innermost loops or no loops.
3275 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3276 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3277 (finalize_partitions, distribute_loop): Don't fuse partition of
3278 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3279 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3282 2018-01-23 Martin Liska <mliska@suse.cz>
3284 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3285 order to ignore the predictor.
3286 (PRED_POLYMORPHIC_CALL): Likewise.
3287 (PRED_RECURSIVE_CALL): Likewise.
3289 2018-01-23 Martin Liska <mliska@suse.cz>
3291 * tree-profile.c (tree_profiling): Print function header to
3292 aware reader which function we are working on.
3293 * value-prof.c (gimple_find_values_to_profile): Do not print
3294 not interesting value histograms.
3296 2018-01-23 Martin Liska <mliska@suse.cz>
3298 * profile-count.h (enum profile_quality): Add
3299 profile_uninitialized as the first value. Do not number values
3300 as they are zero based.
3301 (profile_count::verify): Update sanity check.
3302 (profile_probability::verify): Likewise.
3304 2018-01-23 Nathan Sidwell <nathan@acm.org>
3306 * doc/invoke.texi (ffor-scope): Deprecate.
3308 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3310 PR tree-optimization/83510
3311 * domwalk.c (set_all_edges_as_executable): New function.
3312 (dom_walker::dom_walker): Convert bool param
3313 "skip_unreachable_blocks" to enum reachability. Move setup of
3314 edge flags to set_all_edges_as_executable and only do it when
3315 reachability is REACHABLE_BLOCKS.
3316 * domwalk.h (enum dom_walker::reachability): New enum.
3317 (dom_walker::dom_walker): Convert bool param
3318 "skip_unreachable_blocks" to enum reachability.
3319 (set_all_edges_as_executable): New decl.
3320 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3321 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3323 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3324 but converting true to REACHABLE_BLOCKS.
3325 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3327 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3328 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3329 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3331 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3332 if check_all_array_refs will be called.
3334 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3336 * tree.c (selftest::test_location_wrappers): Add more test
3339 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3341 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3342 (selftest::test_bit_in_range): Likewise.
3344 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3347 * doc/sourcebuild.texi (vect_float): Say that the selector
3348 only describes the situation when -funsafe-math-optimizations is on.
3349 (vect_float_strict): Document.
3351 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3353 PR tree-optimization/83965
3354 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3355 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3356 instead of checking only for a reduction.
3357 (vect_recog_widen_sum_pattern): Likewise.
3359 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3361 * predict.c (probably_never_executed): Only use precise profile info.
3362 (compute_function_frequency): Skip after inlining hack since we now
3363 have quality checking.
3365 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3367 * profile-count.h (profile_probability::very_unlikely,
3368 profile_probability::unlikely, profile_probability::even): Set
3369 precision to guessed.
3371 2018-01-23 Richard Biener <rguenther@suse.de>
3373 PR tree-optimization/83963
3374 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3375 Properly terminate dominator walk when crossing the exit edge not
3376 when visiting its source block.
3378 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3381 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3382 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3384 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3386 PR tree-optimization/83957
3387 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3388 semicolon after for body surrounded by braces.
3390 PR tree-optimization/83081
3391 * profile-count.h (profile_probability::split): New method.
3392 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3393 Use profile_probability::split.
3394 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3395 when splitting a single conditional jump into 2.
3397 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3399 PR tree-optimization/69452
3400 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3403 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3405 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3406 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3407 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3409 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3411 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3412 * config/rl78/rl78.md (movdi): New define_expand.
3413 * config/rl78/rl78.c (rl78_split_movdi): New function.
3415 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3418 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3420 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3421 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3422 128-bit to produce an UNSPEC move to get the double word with the
3423 signbit and then a shift directly to do signbit.
3424 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3425 implementation with a new version that just does either a direct
3426 move or a regular move. Move memory interface to separate insns.
3427 Move insns so they are next to the expander.
3428 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3429 with signbit move. Split big and little endian case.
3430 (signbit<mode>2_dm_mem_le): Likewise.
3431 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3432 (signbit<mode>2_dm2): Likewise.
3434 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3436 * config/rl78/rl78.md (anddi3): New define_expand.
3438 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3440 * config/rl78/rl78.md (umindi3): New define_expand.
3442 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3444 * config/rl78/rl78.md (smindi3): New define_expand.
3446 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3448 * config/rl78/rl78.md (smaxdi3): New define_expand.
3450 2018-01-22 Carl Love <cel@us.ibm.com>
3452 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3453 LVX_V1TI): Add macro expansion.
3454 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3455 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3456 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3457 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3458 Change check to determine if the instruction is a byte reversing
3459 entry. Fix typo in comment.
3460 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3461 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3462 Add def_builtin calls for new builtins.
3463 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3464 Add define_insn expansion.
3466 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3468 * config/rl78/rl78.md (umaxdi3): New define_expand.
3470 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3472 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3473 for non-QImode registers.
3475 2018-01-22 Richard Biener <rguenther@suse.de>
3477 PR tree-optimization/83963
3478 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3479 including the loop exit block.
3480 (scop_detection::merge_sese): Likewise.
3481 (scop_detection::add_scop): Do it here instead.
3483 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3485 * doc/sourcebuild.texi (arm_softfloat): Document.
3487 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3490 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3491 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3492 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3494 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3495 David Edelsohn <dje.gcc@gmail.com>
3498 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3499 Change "crset eq" to "crset 2".
3500 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3501 (*call_indirect_aix<mode>_nospec): Likewise.
3502 (*call_value_indirect_aix<mode>_nospec): Likewise.
3503 (*call_indirect_elfv2<mode>_nospec): Likewise.
3504 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3505 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3506 change assembly output from . to $.
3507 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3508 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3509 (*tablejump<mode>_internal1_nospec): Likewise.
3511 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3514 * config/sh/sh_optimize_sett_clrt.cc:
3515 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3517 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3519 PR tree-optimization/83940
3520 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3521 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3522 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3523 use it to pass back the definition type.
3524 (vect_check_store_rhs): Likewise rhs_dt_out.
3525 (vect_build_gather_load_calls): Add a mask_dt argument and use
3526 it instead of a call to vect_is_simple_use.
3527 (vectorizable_store): Update calls to vect_check_load_store_mask
3528 and vect_check_store_rhs. Use the dt returned by the latter instead
3529 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3530 instead of calls to vect_is_simple_use. Pass the scalar rather
3531 than the vector operand to vect_is_simple_use when handling
3532 second and subsequent copies of an rhs value.
3533 (vectorizable_load): Update calls to vect_check_load_store_mask
3534 and vect_build_gather_load_calls. Use the cached mask_dt and
3535 gs_info.offset_dt instead of calls to vect_is_simple_use.
3537 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3540 * tree-emutls.c: Include gimplify.h.
3541 (lower_emutls_2): New function.
3542 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3543 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3544 it before further processing.
3547 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3548 UINTVAL (trueop1) instead of INTVAL (op1).
3550 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3554 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3555 INCOMING_FRAME_SP_OFFSET if not defined.
3556 (scan_trace): Add ENTRY argument. If true and
3557 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3558 emit a note to adjust the CFA offset.
3559 (create_cfi_notes): Adjust scan_trace callers.
3560 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3561 INCOMING_FRAME_SP_OFFSET in the CIE.
3562 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3563 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3565 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3566 * doc/tm.texi: Regenerated.
3568 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3570 PR rtl-optimization/83147
3571 * lra-constraints.c (remove_inheritance_pseudos): Use
3572 lra_substitute_pseudo_within_insn.
3574 2018-01-19 Tom de Vries <tom@codesourcery.com>
3575 Cesar Philippidis <cesar@codesourcery.com>
3578 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3580 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3583 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3584 spaces for function labels.
3586 2018-01-19 Martin Liska <mliska@suse.cz>
3588 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3589 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3590 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3591 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3592 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3593 (PRED_CONST_RETURN): Change from 69 to 65.
3594 (PRED_NULL_RETURN): Change from 91 to 71.
3595 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3596 (PRED_LOOP_GUARD): Change from 66 to 73.
3598 2018-01-19 Martin Liska <mliska@suse.cz>
3600 * predict.c (predict_insn_def): Add new assert.
3601 (struct branch_predictor): Change type to signed integer.
3602 (test_prediction_value_range): Amend test to cover
3604 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3605 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3606 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3607 (PRED_LOOP_IV_COMPARE): Likewise.
3608 * predict.h (PROB_UNINITIALIZED): Define new constant.
3610 2018-01-19 Martin Liska <mliska@suse.cz>
3612 * predict.c (dump_prediction): Add new format for
3613 analyze_brprob.py script which is enabled with -details
3615 * profile-count.h (precise_p): New function.
3617 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3619 PR tree-optimization/83922
3620 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3621 there are no statements that need masking.
3622 (vect_active_double_reduction_p): New function.
3623 (vect_analyze_loop_operations): Use it when handling phis that
3624 are not in the loop header.
3626 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3628 PR tree-optimization/83914
3629 * tree-vect-loop.c (vectorizable_induction): Don't convert
3630 init_expr or apply the peeling adjustment for inductions
3631 that are nested within the vectorized loop.
3633 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3635 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3638 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3642 * function.h (gimplify_parameters): Add gimple_seq * argument.
3643 * function.c: Include gimple.h and options.h.
3644 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3645 for the added local temporaries if needed.
3646 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3647 if there are any parameter cleanups, wrap whole body into a
3648 try/finally with the cleanups.
3650 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3653 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3654 Use GET_MODE_CLASS for scalar floating point.
3656 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3660 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3661 Fix call of call_cgraph_insertion_hooks.
3663 2018-01-18 Martin Sebor <msebor@redhat.com>
3665 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3667 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3670 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3673 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3676 * common.opt: (-ffile-prefix-map): New option.
3677 * opts.c (common_handle_option): Defer it.
3678 * opts-global.c (handle_common_deferred_options): Handle it.
3679 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3680 * file-prefix-map.h: New file.
3681 (remap_debug_filename, add_debug_prefix_map): ...here.
3682 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3683 * final.c (debug_prefix_map, add_debug_prefix_map
3684 remap_debug_filename): Move to...
3685 * file-prefix-map.c: New file.
3686 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3687 generalize, get rid of alloca(), use strrchr() instead of strchr().
3688 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3689 Implement in terms of add_prefix_map().
3690 (remap_macro_filename, remap_debug_filename): Implement in term of
3692 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3693 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3694 * dbxout.c: Include file-prefix-map.h.
3695 * varasm.c: Likewise.
3696 * vmsdbgout.c: Likewise.
3697 * xcoffout.c: Likewise.
3698 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3699 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3700 * doc/invoke.texi (-ffile-prefix-map): Document.
3701 (-fdebug-prefix-map): Update description.
3703 2018-01-18 Martin Liska <mliska@suse.cz>
3705 * config/i386/i386.c (indirect_thunk_name): Document that also
3707 (output_indirect_thunk): Document why both instructions
3708 (pause and lfence) are generated.
3710 2018-01-18 Richard Biener <rguenther@suse.de>
3712 PR tree-optimization/83887
3713 * graphite-scop-detection.c
3714 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3715 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3716 (scop_detection::merge_sese): Re-implement with a flood-fill
3717 algorithm that properly finds a SESE region if it exists.
3719 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3722 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3723 pointer_diff optimizations use view_convert instead of convert.
3725 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3727 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3728 Generate different code for -mno-speculate-indirect-jumps.
3729 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3730 (*call_indirect_aix<mode>): Disable for
3731 -mno-speculate-indirect-jumps.
3732 (*call_indirect_aix<mode>_nospec): New define_insn.
3733 (*call_value_indirect_aix<mode>): Disable for
3734 -mno-speculate-indirect-jumps.
3735 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3736 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3737 -mno-speculate-indirect-jumps.
3738 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3740 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3742 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3743 long double type, set the flags for noting the default long double
3744 type, even if we don't pass or return a long double type.
3746 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3749 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3752 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3754 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3755 support for merge[hl].
3756 (fold_mergehl_helper): New helper function.
3757 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3758 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3759 (altivec_vmrglw_direct): Add xxmrglw insn.
3761 2018-01-17 Andrew Waterman <andrew@sifive.com>
3763 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3764 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3766 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3769 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3770 call the lto_location_cache before reading the
3771 DECL_SOURCE_LOCATION of the types.
3773 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3774 Richard Sandiford <richard.sandiford@linaro.org>
3776 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3777 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3778 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3779 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3780 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3782 * config/aarch64/constraints.md (aarch64_movti_operand):
3784 * config/aarch64/predicates.md (Uti): Add new constraint.
3786 2018-01-17 Carl Love <cel@us.ibm.com>
3788 * config/rs6000/vsx.md (define_expand xl_len_r,
3789 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3790 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3792 (define_expand, define_insn): Move the shift left from the
3793 define_insn to the define_expand for lxvl and stxvl instructions.
3794 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3795 and XL_LEN_R definitions to PURE.
3797 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3799 * config/i386/i386.c (indirect_thunk_name): Declare regno
3800 as unsigned int. Compare regno with INVALID_REGNUM.
3801 (output_indirect_thunk): Ditto.
3802 (output_indirect_thunk_function): Ditto.
3803 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3804 in the call to output_indirect_thunk_function.
3806 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3809 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3810 rather than the size of inner_type to determine the stack slot size
3811 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3813 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3816 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3819 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3821 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3822 endian Linux systems to optionally enable multilibs for selecting
3823 the long double type if the user configured an explicit type.
3824 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3825 have no long double multilibs if not defined.
3826 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3827 warn if the user used -mabi={ieee,ibm}longdouble and we built
3828 multilibs for long double.
3829 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3830 appropriate multilib option.
3831 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3833 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3834 for building long double multilibs.
3835 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3837 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3839 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3842 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3844 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3847 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3850 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3853 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3855 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3856 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3858 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3860 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3861 different rtl trees depending on TARGET_64BIT.
3862 (rs6000_gen_lvx): Likewise.
3864 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3866 * config/visium/visium.md (nop): Tweak comment.
3867 (hazard_nop): Likewise.
3869 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3871 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3872 -mspeculate-indirect-jumps.
3873 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3874 for -mno-speculate-indirect-jumps.
3875 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3876 (*call_value_indirect_elfv2<mode>): Disable for
3877 -mno-speculate-indirect-jumps.
3878 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3879 (indirect_jump): Emit different RTL for
3880 -mno-speculate-indirect-jumps.
3881 (*indirect_jump<mode>): Disable for
3882 -mno-speculate-indirect-jumps.
3883 (*indirect_jump<mode>_nospec): New define_insn.
3884 (tablejump): Emit different RTL for
3885 -mno-speculate-indirect-jumps.
3886 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3887 (tablejumpsi_nospec): New define_expand.
3888 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3889 (tablejumpdi_nospec): New define_expand.
3890 (*tablejump<mode>_internal1): Disable for
3891 -mno-speculate-indirect-jumps.
3892 (*tablejump<mode>_internal1_nospec): New define_insn.
3893 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3896 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3898 * caller-save.c (insert_save): Drop unnecessary parameter. All
3901 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3902 Richard Biener <rguenth@suse.de>
3905 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3906 return early, inline manually is_gimple_sizepos. Make sure if we
3907 call gimplify_expr we don't end up with a gimple constant.
3908 * tree.c (variably_modified_type_p): Don't return true for
3909 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3910 * gimplify.h (is_gimple_sizepos): Remove.
3912 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3914 PR tree-optimization/83857
3915 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3916 vectorizable_live_operation for pure SLP statements.
3917 (vectorizable_live_operation): Handle PHIs.
3919 2018-01-16 Richard Biener <rguenther@suse.de>
3921 PR tree-optimization/83867
3922 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3923 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3925 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3928 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3929 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3930 If off is not INTEGER_CST, issue a may not be aligned warning
3931 rather than isn't aligned. Use isn%'t rather than isn't.
3932 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3934 <case MULT_EXPR>: Improve the case when bottom and one of the
3935 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3936 operand, in that case check if the other operand is multiple of
3937 bottom divided by the INTEGER_CST operand.
3939 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3942 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3943 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3944 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3945 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3946 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3947 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3948 (pa_function_arg_size): New function.
3950 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3952 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3953 in a separate statement.
3955 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3957 PR tree-optimization/83847
3958 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3959 group gathers and scatters.
3961 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3963 PR rtl-optimization/86620
3964 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3966 PR rtl-optimization/83213
3967 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3968 to last if both are JUMP_INSNs.
3970 PR tree-optimization/83843
3971 * gimple-ssa-store-merging.c
3972 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3973 store_immediate_info for bswap/nop orig_stores.
3975 2018-01-15 Andrew Waterman <andrew@sifive.com>
3977 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3979 <UDIV>: Increase cost if !TARGET_DIV.
3981 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3983 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3984 (define_attr "cr_logical_3op"): New.
3985 (cceq_ior_compare): Adjust.
3986 (cceq_ior_compare_complement): Adjust.
3987 (*cceq_rev_compare): Adjust.
3988 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3989 (is_cracked_insn): Adjust.
3990 (insn_must_be_first_in_group): Adjust.
3991 * config/rs6000/40x.md: Adjust.
3992 * config/rs6000/440.md: Adjust.
3993 * config/rs6000/476.md: Adjust.
3994 * config/rs6000/601.md: Adjust.
3995 * config/rs6000/603.md: Adjust.
3996 * config/rs6000/6xx.md: Adjust.
3997 * config/rs6000/7450.md: Adjust.
3998 * config/rs6000/7xx.md: Adjust.
3999 * config/rs6000/8540.md: Adjust.
4000 * config/rs6000/cell.md: Adjust.
4001 * config/rs6000/e300c2c3.md: Adjust.
4002 * config/rs6000/e500mc.md: Adjust.
4003 * config/rs6000/e500mc64.md: Adjust.
4004 * config/rs6000/e5500.md: Adjust.
4005 * config/rs6000/e6500.md: Adjust.
4006 * config/rs6000/mpc.md: Adjust.
4007 * config/rs6000/power4.md: Adjust.
4008 * config/rs6000/power5.md: Adjust.
4009 * config/rs6000/power6.md: Adjust.
4010 * config/rs6000/power7.md: Adjust.
4011 * config/rs6000/power8.md: Adjust.
4012 * config/rs6000/power9.md: Adjust.
4013 * config/rs6000/rs64.md: Adjust.
4014 * config/rs6000/titan.md: Adjust.
4016 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4018 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4019 ix86_indirect_branch_register logic.
4021 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4023 * config/i386/constraints.md (Bs): Update
4024 ix86_indirect_branch_register check. Don't check
4025 ix86_indirect_branch_register with GOT_memory_operand.
4027 * config/i386/predicates.md (GOT_memory_operand): Don't check
4028 ix86_indirect_branch_register here.
4029 (GOT32_symbol_operand): Likewise.
4031 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4033 * config/i386/predicates.md (constant_call_address_operand):
4034 Rewrite ix86_indirect_branch_register logic.
4035 (sibcall_insn_operand): Likewise.
4037 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4039 * config/i386/constraints.md (Bs): Replace
4040 ix86_indirect_branch_thunk_register with
4041 ix86_indirect_branch_register.
4043 * config/i386/i386.md (indirect_jump): Likewise.
4044 (tablejump): Likewise.
4045 (*sibcall_memory): Likewise.
4046 (*sibcall_value_memory): Likewise.
4047 Peepholes of indirect call and jump via memory: Likewise.
4048 * config/i386/i386.opt: Likewise.
4049 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4050 (GOT_memory_operand): Likewise.
4051 (call_insn_operand): Likewise.
4052 (sibcall_insn_operand): Likewise.
4053 (GOT32_symbol_operand): Likewise.
4055 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4058 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4059 type rather than type addr's type points to.
4060 (expand_omp_atomic_mutex): Likewise.
4061 (expand_omp_atomic): Likewise.
4063 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4066 * config/i386/i386.c (output_indirect_thunk_function): Use
4067 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4068 for __x86_return_thunk.
4070 2018-01-15 Richard Biener <rguenther@suse.de>
4073 * expmed.c (extract_bit_field_1): Fix typo.
4075 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4078 * config/arm/iterators.md (VF): New mode iterator.
4079 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4080 Remove integer-related logic from pattern.
4081 (neon_vabd<mode>_3): Likewise.
4083 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4086 * common.opt (fstrict-overflow): No longer an alias.
4087 (fwrapv-pointer): New option.
4088 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4089 also for pointer types based on flag_wrapv_pointer.
4090 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4091 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4092 opts->x_flag_wrapv got set.
4093 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4094 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4095 POINTER_TYPE_OVERFLOW_UNDEFINED.
4096 * match.pd: Likewise in address comparison pattern.
4097 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4099 2018-01-15 Richard Biener <rguenther@suse.de>
4102 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4103 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4104 Reset type names to their identifier if their TYPE_DECL doesn't
4105 have linkage (and thus is used for ODR and devirt).
4106 (save_debug_info_for_decl): Remove.
4107 (save_debug_info_for_type): Likewise.
4108 (add_tree_to_fld_list): Adjust.
4109 * tree-pretty-print.c (dump_generic_node): Make dumping of
4110 type names more robust.
4112 2018-01-15 Richard Biener <rguenther@suse.de>
4114 * BASE-VER: Bump to 8.0.1.
4116 2018-01-14 Martin Sebor <msebor@redhat.com>
4119 * builtins.c (check_access): Avoid warning when the no-warning bit
4122 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4124 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4125 * ira-color (allocno_hard_regs_compare): Likewise.
4127 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4130 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4131 Use .pushsection/.popsection.
4133 2018-01-14 Martin Sebor <msebor@redhat.com>
4136 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4138 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4140 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4141 entry from extra_headers.
4142 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4143 extra_headers, make the list bitwise identical to the i?86-*-* one.
4145 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4147 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4148 -mcmodel=large with -mindirect-branch=thunk,
4149 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4150 -mfunction-return=thunk-extern.
4151 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4152 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4153 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4155 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4157 * config/i386/i386.c (print_reg): Print the name of the full
4158 integer register without '%'.
4159 (ix86_print_operand): Handle 'V'.
4160 * doc/extend.texi: Document 'V' modifier.
4162 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4164 * config/i386/constraints.md (Bs): Disallow memory operand for
4165 -mindirect-branch-register.
4167 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4168 (GOT_memory_operand): Likewise.
4169 (call_insn_operand): Likewise.
4170 (sibcall_insn_operand): Likewise.
4171 (GOT32_symbol_operand): Likewise.
4172 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4173 for -mindirect-branch-register.
4174 (tablejump): Likewise.
4175 (*sibcall_memory): Likewise.
4176 (*sibcall_value_memory): Likewise.
4177 Disallow peepholes of indirect call and jump via memory for
4178 -mindirect-branch-register.
4179 (*call_pop): Replace m with Bw.
4180 (*call_value_pop): Likewise.
4181 (*sibcall_pop_memory): Replace m with Bs.
4182 * config/i386/i386.opt (mindirect-branch-register): New option.
4183 * doc/invoke.texi: Document -mindirect-branch-register option.
4185 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4187 * config/i386/i386-protos.h (ix86_output_function_return): New.
4188 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4189 set function_return_type.
4190 (indirect_thunk_name): Add ret_p to indicate thunk for function
4192 (output_indirect_thunk_function): Pass false to
4193 indirect_thunk_name.
4194 (ix86_output_indirect_branch_via_reg): Likewise.
4195 (ix86_output_indirect_branch_via_push): Likewise.
4196 (output_indirect_thunk_function): Create alias for function
4197 return thunk if regno < 0.
4198 (ix86_output_function_return): New function.
4199 (ix86_handle_fndecl_attribute): Handle function_return.
4200 (ix86_attribute_table): Add function_return.
4201 * config/i386/i386.h (machine_function): Add
4202 function_return_type.
4203 * config/i386/i386.md (simple_return_internal): Use
4204 ix86_output_function_return.
4205 (simple_return_internal_long): Likewise.
4206 * config/i386/i386.opt (mfunction-return=): New option.
4207 (indirect_branch): Mention -mfunction-return=.
4208 * doc/extend.texi: Document function_return function attribute.
4209 * doc/invoke.texi: Document -mfunction-return= option.
4211 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4213 * config/i386/i386-opts.h (indirect_branch): New.
4214 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4215 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4216 with local indirect jump when converting indirect call and jump.
4217 (ix86_set_indirect_branch_type): New.
4218 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4219 (indirectlabelno): New.
4220 (indirect_thunk_needed): Likewise.
4221 (indirect_thunk_bnd_needed): Likewise.
4222 (indirect_thunks_used): Likewise.
4223 (indirect_thunks_bnd_used): Likewise.
4224 (INDIRECT_LABEL): Likewise.
4225 (indirect_thunk_name): Likewise.
4226 (output_indirect_thunk): Likewise.
4227 (output_indirect_thunk_function): Likewise.
4228 (ix86_output_indirect_branch_via_reg): Likewise.
4229 (ix86_output_indirect_branch_via_push): Likewise.
4230 (ix86_output_indirect_branch): Likewise.
4231 (ix86_output_indirect_jmp): Likewise.
4232 (ix86_code_end): Call output_indirect_thunk_function if needed.
4233 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4235 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4236 (ix86_attribute_table): Add indirect_branch.
4237 * config/i386/i386.h (machine_function): Add indirect_branch_type
4238 and has_local_indirect_jump.
4239 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4241 (tablejump): Likewise.
4242 (*indirect_jump): Use ix86_output_indirect_jmp.
4243 (*tablejump_1): Likewise.
4244 (simple_return_indirect_internal): Likewise.
4245 * config/i386/i386.opt (mindirect-branch=): New option.
4246 (indirect_branch): New.
4249 (thunk-inline): Likewise.
4250 (thunk-extern): Likewise.
4251 * doc/extend.texi: Document indirect_branch function attribute.
4252 * doc/invoke.texi: Document -mindirect-branch= option.
4254 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4257 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4259 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4261 * ipa-inline.c (want_inline_small_function_p): Return false if
4262 inlining has already failed with CIF_FINAL_ERROR.
4263 (update_caller_keys): Call want_inline_small_function_p before
4265 (update_callee_keys): Likewise.
4267 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4269 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4271 (rs6000_quadword_masked_address_p): Likewise.
4272 (quad_aligned_load_p): Likewise.
4273 (quad_aligned_store_p): Likewise.
4274 (const_load_sequence_p): Add comment to describe the outer-most loop.
4275 (mimic_memory_attributes_and_flags): New function.
4276 (rs6000_gen_stvx): Likewise.
4277 (replace_swapped_aligned_store): Likewise.
4278 (rs6000_gen_lvx): Likewise.
4279 (replace_swapped_aligned_load): Likewise.
4280 (replace_swapped_load_constant): Capitalize argument name in
4281 comment describing this function.
4282 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4283 and stores that access quad-word aligned addresses and replace
4284 with stvx or lvx instructions when appropriate.
4285 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4286 New function prototype.
4287 (rs6000_quadword_masked_address_p): Likewise.
4288 (rs6000_gen_lvx): Likewise.
4289 (rs6000_gen_stvx): Likewise.
4290 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4291 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4292 when memory address is aligned.
4293 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4294 this split to select lvx instruction when memory address is aligned.
4295 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4296 instruction when memory address is aligned.
4297 (*vsx_le_perm_load_v16qi): Likewise.
4298 (four unnamed splitters): Modify to select the stvx instruction
4299 when memory is aligned.
4301 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4303 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4304 which appears in the queue multiple times.
4306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4307 Alan Hayward <alan.hayward@arm.com>
4308 David Sherwood <david.sherwood@arm.com>
4310 * tree-vectorizer.h (vec_lower_bound): New structure.
4311 (_loop_vec_info): Add check_nonzero and lower_bounds.
4312 (LOOP_VINFO_CHECK_NONZERO): New macro.
4313 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4314 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4315 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4316 fields. Make seg_len the distance travelled, not including the
4318 (dr_direction_indicator): Declare.
4319 (dr_zero_step_indicator): Likewise.
4320 (dr_known_forward_stride_p): Likewise.
4321 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4323 (runtime_alias_check_p): Allow runtime alias checks with
4325 (operator ==): Compare access_size and align.
4326 (prune_runtime_alias_test_list): Rework for new distinction between
4327 the access_size and seg_len.
4328 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4330 (get_segment_min_max): New function.
4331 (create_intersect_range_checks): Use it.
4332 (dr_step_indicator): New function.
4333 (dr_direction_indicator): Likewise.
4334 (dr_zero_step_indicator): Likewise.
4335 (dr_known_forward_stride_p): Likewise.
4336 * tree-loop-distribution.c (data_ref_segment_size): Return
4337 DR_STEP * (niters - 1).
4338 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4340 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4341 (vect_preserves_scalar_order_p): New function, split out from...
4342 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4343 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4344 (vect_vfa_access_size): New function.
4345 (vect_vfa_align): Likewise.
4346 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4347 (dump_lower_bound): New function.
4348 (vect_check_lower_bound): Likewise.
4349 (vect_small_gap_p): Likewise.
4350 (vectorizable_with_step_bound_p): Likewise.
4351 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4352 depencies if the vectorization factor is 1. Convert the checks
4353 for nonzero steps into checks on the bounds of DR_STEP. Try using
4354 a bunds check for variable steps if the minimum required step is
4355 relatively small. Update calls to the dr_with_seg_len
4356 constructor and to vect_compile_time_alias.
4357 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4359 (vect_loop_versioning): Call it.
4360 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4362 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4364 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4365 Alan Hayward <alan.hayward@arm.com>
4366 David Sherwood <david.sherwood@arm.com>
4368 * doc/sourcebuild.texi (vect_scatter_store): Document.
4369 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4371 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4373 * genopinit.c (main): Add supports_vec_scatter_store and
4374 supports_vec_scatter_store_cached to target_optabs.
4375 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4376 IFN_MASK_SCATTER_STORE.
4377 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4379 * internal-fn.h (internal_store_fn_p): Declare.
4380 (internal_fn_stored_value_index): Likewise.
4381 * internal-fn.c (scatter_store_direct): New macro.
4382 (expand_scatter_store_optab_fn): New function.
4383 (direct_scatter_store_optab_supported_p): New macro.
4384 (internal_store_fn_p): New function.
4385 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4386 IFN_MASK_SCATTER_STORE.
4387 (internal_fn_mask_index): Likewise.
4388 (internal_fn_stored_value_index): New function.
4389 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4391 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4392 * optabs-query.c (supports_vec_scatter_store_p): New function.
4393 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4394 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4395 true for scatter stores.
4396 (vect_gather_scatter_fn_p): Handle scatter stores too.
4397 (vect_check_gather_scatter): Consider using scatter stores if
4398 supports_vec_scatter_store_p.
4399 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4401 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4402 internal_fn_stored_value_index.
4403 (check_load_store_masking): Handle scatter stores too.
4404 (vect_get_store_rhs): Make public.
4405 (vectorizable_call): Use internal_store_fn_p.
4406 (vectorizable_store): Handle scatter store internal functions.
4407 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4408 when deciding whether the end of the group has been reached.
4409 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4410 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4411 (mask_scatter_store<mode>): New insns.
4413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4414 Alan Hayward <alan.hayward@arm.com>
4415 David Sherwood <david.sherwood@arm.com>
4417 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4418 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4419 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4421 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4422 Use vect_truncate_gather_scatter_offset if we can't treat the
4423 operation as a normal gather load or scatter store.
4424 (get_group_load_store_type): Take the gather_scatter_info
4425 as argument. Try using a gather load or scatter store for
4426 single-element groups.
4427 (get_load_store_type): Update calls to get_group_load_store_type
4428 and vect_use_strided_gather_scatters_p.
4430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4431 Alan Hayward <alan.hayward@arm.com>
4432 David Sherwood <david.sherwood@arm.com>
4434 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4435 optional tree argument.
4436 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4438 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4439 but continue to use the current value as a fallback.
4440 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4441 to compare the updates.
4442 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4443 (get_load_store_type): Use it when handling a strided access.
4444 (vect_get_strided_load_store_ops): New function.
4445 (vect_get_data_ptr_increment): Likewise.
4446 (vectorizable_load): Handle strided gather loads. Always pass
4447 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4449 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4450 Alan Hayward <alan.hayward@arm.com>
4451 David Sherwood <david.sherwood@arm.com>
4453 * doc/md.texi (gather_load@var{m}): Document.
4454 (mask_gather_load@var{m}): Likewise.
4455 * genopinit.c (main): Add supports_vec_gather_load and
4456 supports_vec_gather_load_cached to target_optabs.
4457 * optabs-tree.c (init_tree_optimization_optabs): Use
4458 ggc_cleared_alloc to allocate target_optabs.
4459 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4460 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4462 * internal-fn.h (internal_load_fn_p): Declare.
4463 (internal_gather_scatter_fn_p): Likewise.
4464 (internal_fn_mask_index): Likewise.
4465 (internal_gather_scatter_fn_supported_p): Likewise.
4466 * internal-fn.c (gather_load_direct): New macro.
4467 (expand_gather_load_optab_fn): New function.
4468 (direct_gather_load_optab_supported_p): New macro.
4469 (direct_internal_fn_optab): New function.
4470 (internal_load_fn_p): Likewise.
4471 (internal_gather_scatter_fn_p): Likewise.
4472 (internal_fn_mask_index): Likewise.
4473 (internal_gather_scatter_fn_supported_p): Likewise.
4474 * optabs-query.c (supports_at_least_one_mode_p): New function.
4475 (supports_vec_gather_load_p): Likewise.
4476 * optabs-query.h (supports_vec_gather_load_p): Declare.
4477 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4478 and memory_type field.
4479 (NUM_PATTERNS): Bump to 15.
4480 * tree-vect-data-refs.c: Include internal-fn.h.
4481 (vect_gather_scatter_fn_p): New function.
4482 (vect_describe_gather_scatter_call): Likewise.
4483 (vect_check_gather_scatter): Try using internal functions for
4484 gather loads. Recognize existing calls to a gather load function.
4485 (vect_analyze_data_refs): Consider using gather loads if
4486 supports_vec_gather_load_p.
4487 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4488 (vect_get_gather_scatter_offset_type): Likewise.
4489 (vect_convert_mask_for_vectype): Likewise.
4490 (vect_add_conversion_to_patterm): Likewise.
4491 (vect_try_gather_scatter_pattern): Likewise.
4492 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4493 (vect_vect_recog_func_ptrs): Add it.
4494 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4495 internal_fn_mask_index and internal_gather_scatter_fn_p.
4496 (check_load_store_masking): Take the gather_scatter_info as an
4497 argument and handle gather loads.
4498 (vect_get_gather_scatter_ops): New function.
4499 (vectorizable_call): Check internal_load_fn_p.
4500 (vectorizable_load): Likewise. Handle gather load internal
4502 (vectorizable_store): Update call to check_load_store_masking.
4503 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4504 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4505 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4506 (aarch64_gather_scale_operand_d): New predicates.
4507 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4508 (mask_gather_load<mode>): New insns.
4510 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4511 Alan Hayward <alan.hayward@arm.com>
4512 David Sherwood <david.sherwood@arm.com>
4514 * optabs.def (fold_left_plus_optab): New optab.
4515 * doc/md.texi (fold_left_plus_@var{m}): Document.
4516 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4517 * internal-fn.c (fold_left_direct): Define.
4518 (expand_fold_left_optab_fn): Likewise.
4519 (direct_fold_left_optab_supported_p): Likewise.
4520 * fold-const-call.c (fold_const_fold_left): New function.
4521 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4522 * tree-parloops.c (valid_reduction_p): New function.
4523 (gather_scalar_reductions): Use it.
4524 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4525 (vect_finish_replace_stmt): Declare.
4526 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4527 (needs_fold_left_reduction_p): New function, split out from...
4528 (vect_is_simple_reduction): ...here. Accept reductions that
4529 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4530 (vect_force_simple_reduction): Also store the reduction type in
4531 the assignment's STMT_VINFO_REDUC_TYPE.
4532 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4533 (merge_with_identity): New function.
4534 (vect_expand_fold_left): Likewise.
4535 (vectorize_fold_left_reduction): Likewise.
4536 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4537 scalar phi in place for it. Check for target support and reject
4538 cases that would reassociate the operation. Defer the transform
4539 phase to vectorize_fold_left_reduction.
4540 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4541 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4542 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4546 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4547 call to ifc_temp_var.
4549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4550 Alan Hayward <alan.hayward@arm.com>
4551 David Sherwood <david.sherwood@arm.com>
4553 * target.def (legitimize_address_displacement): Take the original
4554 offset as a poly_int.
4555 * targhooks.h (default_legitimize_address_displacement): Update
4557 * targhooks.c (default_legitimize_address_displacement): Likewise.
4558 * doc/tm.texi: Regenerate.
4559 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4560 as an argument, moving assert of ad->disp == ad->disp_term to...
4561 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4562 Try calling targetm.legitimize_address_displacement before expanding
4563 the address rather than afterwards, and adjust for the new interface.
4564 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4565 Match the new hook interface. Handle SVE addresses.
4566 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4571 * Makefile.in (OBJS): Add early-remat.o.
4572 * target.def (select_early_remat_modes): New hook.
4573 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4574 * doc/tm.texi: Regenerate.
4575 * targhooks.h (default_select_early_remat_modes): Declare.
4576 * targhooks.c (default_select_early_remat_modes): New function.
4577 * timevar.def (TV_EARLY_REMAT): New timevar.
4578 * passes.def (pass_early_remat): New pass.
4579 * tree-pass.h (make_pass_early_remat): Declare.
4580 * early-remat.c: New file.
4581 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4583 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4585 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4586 Alan Hayward <alan.hayward@arm.com>
4587 David Sherwood <david.sherwood@arm.com>
4589 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4590 vfm1 with a bound_epilog parameter.
4591 (vect_do_peeling): Update calls accordingly, and move the prologue
4592 call earlier in the function. Treat the base bound_epilog as 0 for
4593 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4594 this base when peeling for gaps.
4595 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4596 with fully-masked loops.
4597 (vect_estimate_min_profitable_iters): Handle the single peeled
4598 iteration in that case.
4600 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4601 Alan Hayward <alan.hayward@arm.com>
4602 David Sherwood <david.sherwood@arm.com>
4604 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4605 single-element interleaving even if the size is not a power of 2.
4606 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4607 accesses for single-element interleaving if the group size is
4610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4611 Alan Hayward <alan.hayward@arm.com>
4612 David Sherwood <david.sherwood@arm.com>
4614 * doc/md.texi (fold_extract_last_@var{m}): Document.
4615 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4616 * optabs.def (fold_extract_last_optab): New optab.
4617 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4618 * internal-fn.c (fold_extract_direct): New macro.
4619 (expand_fold_extract_optab_fn): Likewise.
4620 (direct_fold_extract_optab_supported_p): Likewise.
4621 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4622 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4623 EXTRACT_LAST_REDUCTION.
4624 (get_initial_def_for_reduction): Do not create an initial vector
4625 for EXTRACT_LAST_REDUCTION reductions.
4626 (vectorizable_reduction): Leave the scalar phi in place for
4627 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4628 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4629 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4630 transform phase to vectorizable_condition.
4631 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4633 (vect_finish_stmt_generation): ...here.
4634 (vect_finish_replace_stmt): New function.
4635 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4636 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4638 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4640 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4641 Alan Hayward <alan.hayward@arm.com>
4642 David Sherwood <david.sherwood@arm.com>
4644 * doc/md.texi (extract_last_@var{m}): Document.
4645 * optabs.def (extract_last_optab): New optab.
4646 * internal-fn.def (EXTRACT_LAST): New internal function.
4647 * internal-fn.c (cond_unary_direct): New macro.
4648 (expand_cond_unary_optab_fn): Likewise.
4649 (direct_cond_unary_optab_supported_p): Likewise.
4650 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4651 loops using EXTRACT_LAST.
4652 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4653 (extract_last_<mode>): ...this optab.
4654 (vec_extract<mode><Vel>): Update accordingly.
4656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4657 Alan Hayward <alan.hayward@arm.com>
4658 David Sherwood <david.sherwood@arm.com>
4660 * target.def (empty_mask_is_expensive): New hook.
4661 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4662 * doc/tm.texi: Regenerate.
4663 * targhooks.h (default_empty_mask_is_expensive): Declare.
4664 * targhooks.c (default_empty_mask_is_expensive): New function.
4665 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4666 if the target says that empty masks are expensive.
4667 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4669 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4672 Alan Hayward <alan.hayward@arm.com>
4673 David Sherwood <david.sherwood@arm.com>
4675 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4676 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4677 (vect_use_loop_mask_for_alignment_p): New function.
4678 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4679 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4680 niters_skip argument. Make sure that the first niters_skip elements
4681 of the first iteration are inactive.
4682 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4683 Update call to vect_set_loop_masks_directly.
4684 (get_misalign_in_elems): New function, split out from...
4685 (vect_gen_prolog_loop_niters): ...here.
4686 (vect_update_init_of_dr): Take a code argument that specifies whether
4687 the adjustment should be added or subtracted.
4688 (vect_update_init_of_drs): Likewise.
4689 (vect_prepare_for_masked_peels): New function.
4690 (vect_do_peeling): Skip prologue peeling if we're using a mask
4691 instead. Update call to vect_update_inits_of_drs.
4692 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4694 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4695 alignment. Do not include the number of peeled iterations in
4696 the minimum threshold in that case.
4697 (vectorizable_induction): Adjust the start value down by
4698 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4699 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4700 Take the number of skipped iterations into account when calculating
4702 * tree-vect-stmts.c (vect_gen_while_not): New function.
4704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4705 Alan Hayward <alan.hayward@arm.com>
4706 David Sherwood <david.sherwood@arm.com>
4708 * doc/sourcebuild.texi (vect_fully_masked): Document.
4709 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4711 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4713 (vect_analyze_loop_2): ...here. Don't check the vectorization
4714 factor against the number of loop iterations if the loop is
4717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4718 Alan Hayward <alan.hayward@arm.com>
4719 David Sherwood <david.sherwood@arm.com>
4721 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4722 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4723 (dump_groups): Update accordingly.
4724 (iv_use::mem_type): New member variable.
4725 (address_p): New function.
4726 (record_use): Add a mem_type argument and initialize the new
4728 (record_group_use): Add a mem_type argument. Use address_p.
4729 Remove obsolete null checks of base_object. Update call to record_use.
4730 (find_interesting_uses_op): Update call to record_group_use.
4731 (find_interesting_uses_cond): Likewise.
4732 (find_interesting_uses_address): Likewise.
4733 (get_mem_type_for_internal_fn): New function.
4734 (find_address_like_use): Likewise.
4735 (find_interesting_uses_stmt): Try find_address_like_use before
4736 calling find_interesting_uses_op.
4737 (addr_offset_valid_p): Use the iv mem_type field as the type
4738 of the addressed memory.
4739 (add_autoinc_candidates): Likewise.
4740 (get_address_cost): Likewise.
4741 (split_small_address_groups_p): Use address_p.
4742 (split_address_groups): Likewise.
4743 (add_iv_candidate_for_use): Likewise.
4744 (autoinc_possible_for_pair): Likewise.
4745 (rewrite_groups): Likewise.
4746 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4747 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4748 (get_alias_ptr_type_for_ptr_address): New function.
4749 (rewrite_use_address): Rewrite address uses in calls that were
4750 identified by find_address_like_use.
4752 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4753 Alan Hayward <alan.hayward@arm.com>
4754 David Sherwood <david.sherwood@arm.com>
4756 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4758 * gimple-expr.h (is_gimple_addressable: Likewise.
4759 * gimple-expr.c (is_gimple_address): Likewise.
4760 * internal-fn.c (expand_call_mem_ref): New function.
4761 (expand_mask_load_optab_fn): Use it.
4762 (expand_mask_store_optab_fn): Likewise.
4764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4765 Alan Hayward <alan.hayward@arm.com>
4766 David Sherwood <david.sherwood@arm.com>
4768 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4769 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4770 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4771 (cond_umax@var{mode}): Document.
4772 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4773 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4774 (cond_umin_optab, cond_umax_optab): New optabs.
4775 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4776 (COND_IOR, COND_XOR): New internal functions.
4777 * internal-fn.h (get_conditional_internal_fn): Declare.
4778 * internal-fn.c (cond_binary_direct): New macro.
4779 (expand_cond_binary_optab_fn): Likewise.
4780 (direct_cond_binary_optab_supported_p): Likewise.
4781 (get_conditional_internal_fn): New function.
4782 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4783 Cope with reduction statements that are vectorized as calls rather
4785 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4786 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4787 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4788 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4789 (UNSPEC_COND_EOR): New unspecs.
4790 (optab): Add mappings for them.
4791 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4792 (sve_int_op, sve_fp_op): New int attributes.
4794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4795 Alan Hayward <alan.hayward@arm.com>
4796 David Sherwood <david.sherwood@arm.com>
4798 * optabs.def (while_ult_optab): New optab.
4799 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4800 * internal-fn.def (WHILE_ULT): New internal function.
4801 * internal-fn.h (direct_internal_fn_supported_p): New override
4802 that takes two types as argument.
4803 * internal-fn.c (while_direct): New macro.
4804 (expand_while_optab_fn): New function.
4805 (convert_optab_supported_p): Likewise.
4806 (direct_while_optab_supported_p): New macro.
4807 * wide-int.h (wi::udiv_ceil): New function.
4808 * tree-vectorizer.h (rgroup_masks): New structure.
4809 (vec_loop_masks): New typedef.
4810 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4812 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4813 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4814 (vect_max_vf): New function.
4815 (slpeel_make_loop_iterate_ntimes): Delete.
4816 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4817 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4818 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4819 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4820 internal-fn.h, stor-layout.h and optabs-query.h.
4821 (vect_set_loop_mask): New function.
4822 (add_preheader_seq): Likewise.
4823 (add_header_seq): Likewise.
4824 (interleave_supported_p): Likewise.
4825 (vect_maybe_permute_loop_masks): Likewise.
4826 (vect_set_loop_masks_directly): Likewise.
4827 (vect_set_loop_condition_masked): Likewise.
4828 (vect_set_loop_condition_unmasked): New function, split out from
4829 slpeel_make_loop_iterate_ntimes.
4830 (slpeel_make_loop_iterate_ntimes): Rename to..
4831 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4832 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4833 (vect_do_peeling): Update call accordingly.
4834 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4836 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4837 mask_compare_type, can_fully_mask_p and fully_masked_p.
4838 (release_vec_loop_masks): New function.
4839 (_loop_vec_info): Use it to free the loop masks.
4840 (can_produce_all_loop_masks_p): New function.
4841 (vect_get_max_nscalars_per_iter): Likewise.
4842 (vect_verify_full_masking): Likewise.
4843 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4844 retries, and free the mask rgroups before retrying. Check loop-wide
4845 reasons for disallowing fully-masked loops. Make the final decision
4846 about whether use a fully-masked loop or not.
4847 (vect_estimate_min_profitable_iters): Do not assume that peeling
4848 for the number of iterations will be needed for fully-masked loops.
4849 (vectorizable_reduction): Disable fully-masked loops.
4850 (vectorizable_live_operation): Likewise.
4851 (vect_halve_mask_nunits): New function.
4852 (vect_double_mask_nunits): Likewise.
4853 (vect_record_loop_mask): Likewise.
4854 (vect_get_loop_mask): Likewise.
4855 (vect_transform_loop): Handle the case in which the final loop
4856 iteration might handle a partial vector. Call vect_set_loop_condition
4857 instead of slpeel_make_loop_iterate_ntimes.
4858 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4859 (check_load_store_masking): New function.
4860 (prepare_load_store_mask): Likewise.
4861 (vectorizable_store): Handle fully-masked loops.
4862 (vectorizable_load): Likewise.
4863 (supportable_widening_operation): Use vect_halve_mask_nunits for
4865 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4866 (vect_gen_while): New function.
4867 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4868 (aarch64_uqdec<mode>): New insn.
4870 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4871 Alan Hayward <alan.hayward@arm.com>
4872 David Sherwood <david.sherwood@arm.com>
4874 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4875 (reduc_xor_scal_optab): New optabs.
4876 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4877 (reduc_xor_scal_@var{m}): Document.
4878 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4879 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4881 * fold-const-call.c (fold_const_call): Handle them.
4882 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4883 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4884 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4885 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4886 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4887 (UNSPEC_XORV): New unspecs.
4888 (optab): Add entries for them.
4889 (BITWISEV): New int iterator.
4890 (bit_reduc_op): New int attributes.
4892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4893 Alan Hayward <alan.hayward@arm.com>
4894 David Sherwood <david.sherwood@arm.com>
4896 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4897 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4898 * optabs.def (vec_shl_insert_optab): New optab.
4899 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4900 (duplicate_and_interleave): Likewise.
4901 * tree-vect-loop.c: Include internal-fn.h.
4902 (neutral_op_for_slp_reduction): New function, split out from
4903 get_initial_defs_for_reduction.
4904 (get_initial_def_for_reduction): Handle option 2 for variable-length
4905 vectors by loading the neutral value into a vector and then shifting
4906 the initial value into element 0.
4907 (get_initial_defs_for_reduction): Replace the code argument with
4908 the neutral value calculated by neutral_op_for_slp_reduction.
4909 Use gimple_build_vector for constant-length vectors.
4910 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4911 but the first group_size elements have a neutral value.
4912 Use duplicate_and_interleave otherwise.
4913 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4914 Update call to get_initial_defs_for_reduction. Handle SLP
4915 reductions for variable-length vectors by creating one vector
4916 result for each scalar result, with the elements associated
4917 with other scalar results stubbed out with the neutral value.
4918 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4919 Require IFN_VEC_SHL_INSERT for double reductions on
4920 variable-length vectors, or SLP reductions that have
4921 a neutral value. Require can_duplicate_and_interleave_p
4922 support for variable-length unchained SLP reductions if there
4923 is no neutral value, such as for MIN/MAX reductions. Also require
4924 the number of vector elements to be a multiple of the number of
4925 SLP statements when doing variable-length unchained SLP reductions.
4926 Update call to vect_create_epilog_for_reduction.
4927 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4928 and remove initial values.
4929 (duplicate_and_interleave): Make public.
4930 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4931 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4933 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4934 Alan Hayward <alan.hayward@arm.com>
4935 David Sherwood <david.sherwood@arm.com>
4937 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4938 (can_duplicate_and_interleave_p): New function.
4939 (vect_get_and_check_slp_defs): Take the vector of statements
4940 rather than just the current one. Remove excess parentheses.
4941 Restriction rejectinon of vect_constant_def and vect_external_def
4942 for variable-length vectors to boolean types, or types for which
4943 can_duplicate_and_interleave_p is false.
4944 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4945 (duplicate_and_interleave): New function.
4946 (vect_get_constant_vectors): Use gimple_build_vector for
4947 constant-length vectors and suitable variable-length constant
4948 vectors. Use duplicate_and_interleave for other variable-length
4949 vectors. Don't defer the update when inserting new statements.
4951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4952 Alan Hayward <alan.hayward@arm.com>
4953 David Sherwood <david.sherwood@arm.com>
4955 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4956 min_profitable_iters doesn't go negative.
4958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4959 Alan Hayward <alan.hayward@arm.com>
4960 David Sherwood <david.sherwood@arm.com>
4962 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4963 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4964 * optabs.def (vec_mask_load_lanes_optab): New optab.
4965 (vec_mask_store_lanes_optab): Likewise.
4966 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4967 (MASK_STORE_LANES): Likewise.
4968 * internal-fn.c (mask_load_lanes_direct): New macro.
4969 (mask_store_lanes_direct): Likewise.
4970 (expand_mask_load_optab_fn): Handle masked operations.
4971 (expand_mask_load_lanes_optab_fn): New macro.
4972 (expand_mask_store_optab_fn): Handle masked operations.
4973 (expand_mask_store_lanes_optab_fn): New macro.
4974 (direct_mask_load_lanes_optab_supported_p): Likewise.
4975 (direct_mask_store_lanes_optab_supported_p): Likewise.
4976 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4978 (vect_load_lanes_supported): Likewise.
4979 * tree-vect-data-refs.c (strip_conversion): New function.
4980 (can_group_stmts_p): Likewise.
4981 (vect_analyze_data_ref_accesses): Use it instead of checking
4982 for a pair of assignments.
4983 (vect_store_lanes_supported): Take a masked_p parameter.
4984 (vect_load_lanes_supported): Likewise.
4985 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4986 vect_store_lanes_supported and vect_load_lanes_supported.
4987 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4988 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4989 parameter. Don't allow gaps for masked accesses.
4990 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4991 and vect_load_lanes_supported.
4992 (get_load_store_type): Take a masked_p parameter and update
4993 call to get_group_load_store_type.
4994 (vectorizable_store): Update call to get_load_store_type.
4995 Handle IFN_MASK_STORE_LANES.
4996 (vectorizable_load): Update call to get_load_store_type.
4997 Handle IFN_MASK_LOAD_LANES.
4999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5000 Alan Hayward <alan.hayward@arm.com>
5001 David Sherwood <david.sherwood@arm.com>
5003 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5005 * config/aarch64/aarch64-protos.h
5006 (aarch64_sve_struct_memory_operand_p): Declare.
5007 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5008 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5009 (VPRED, vpred): Handle SVE structure modes.
5010 * config/aarch64/constraints.md (Utx): New constraint.
5011 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5012 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5013 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5014 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5015 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5016 structure modes. Split into pieces after RA.
5017 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5018 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5020 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5021 SVE structure modes.
5022 (aarch64_classify_address): Likewise.
5023 (sizetochar): Move earlier in file.
5024 (aarch64_print_operand): Handle SVE register lists.
5025 (aarch64_array_mode): New function.
5026 (aarch64_sve_struct_memory_operand_p): Likewise.
5027 (TARGET_ARRAY_MODE): Redefine.
5029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5030 Alan Hayward <alan.hayward@arm.com>
5031 David Sherwood <david.sherwood@arm.com>
5033 * target.def (array_mode): New target hook.
5034 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5035 * doc/tm.texi: Regenerate.
5036 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5037 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5038 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5040 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5043 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5044 Alan Hayward <alan.hayward@arm.com>
5045 David Sherwood <david.sherwood@arm.com>
5047 * fold-const.c (fold_binary_loc): Check the argument types
5048 rather than the result type when testing for a vector operation.
5050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5052 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5053 * doc/tm.texi: Regenerate.
5055 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5056 Alan Hayward <alan.hayward@arm.com>
5057 David Sherwood <david.sherwood@arm.com>
5059 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5060 (sve): Document new AArch64 extension.
5061 * doc/md.texi (w): Extend the description of the AArch64
5062 constraint to include SVE vectors.
5063 (Upl, Upa): Document new AArch64 predicate constraints.
5064 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5066 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5067 (msve-vector-bits=): New option.
5068 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5069 SVE when these are disabled.
5070 (sve): New extension.
5071 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5072 modes. Adjust their number of units based on aarch64_sve_vg.
5073 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5074 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5075 aarch64_addr_query_type.
5076 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5077 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5078 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5079 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5080 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5081 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5082 (aarch64_simd_imm_zero_p): Delete.
5083 (aarch64_check_zero_based_sve_index_immediate): Declare.
5084 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5085 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5086 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5087 (aarch64_sve_float_mul_immediate_p): Likewise.
5088 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5090 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5091 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5092 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5093 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5094 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5095 (aarch64_regmode_natural_size): Likewise.
5096 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5097 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5099 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5100 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5101 for VG and the SVE predicate registers.
5102 (V_ALIASES): Add a "z"-prefixed alias.
5103 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5104 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5105 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5106 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5107 (REG_CLASS_NAMES): Add entries for them.
5108 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5109 and the predicate registers.
5110 (aarch64_sve_vg): Declare.
5111 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5112 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5113 (REGMODE_NATURAL_SIZE): Define.
5114 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5116 * config/aarch64/aarch64.c: Include cfgrtl.h.
5117 (simd_immediate_info): Add a constructor for series vectors,
5118 and an associated step field.
5119 (aarch64_sve_vg): New variable.
5120 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5121 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5122 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5123 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5124 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5125 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5126 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5127 (aarch64_get_mask_mode): New functions.
5128 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5129 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5130 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5131 predicate modes and predicate registers. Explicitly restrict
5132 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5133 to store a vector mode if it is recognized by
5134 aarch64_classify_vector_mode.
5135 (aarch64_regmode_natural_size): New function.
5136 (aarch64_hard_regno_caller_save_mode): Return the original mode
5138 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5139 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5140 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5141 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5143 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5144 does not overlap dest if the function is frame-related. Handle
5146 (aarch64_split_add_offset): New function.
5147 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5148 them aarch64_add_offset.
5149 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5150 and update call to aarch64_sub_sp.
5151 (aarch64_add_cfa_expression): New function.
5152 (aarch64_expand_prologue): Pass extra temporary registers to the
5153 functions above. Handle the case in which we need to emit new
5154 DW_CFA_expressions for registers that were originally saved
5155 relative to the stack pointer, but now have to be expressed
5156 relative to the frame pointer.
5157 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5159 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5160 IP0 and IP1 values for SVE frames.
5161 (aarch64_expand_vec_series): New function.
5162 (aarch64_expand_sve_widened_duplicate): Likewise.
5163 (aarch64_expand_sve_const_vector): Likewise.
5164 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5165 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5166 into the register, rather than emitting a SET directly.
5167 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5168 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5169 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5170 (offset_9bit_signed_scaled_p): New functions.
5171 (aarch64_replicate_bitmask_imm): New function.
5172 (aarch64_bitmask_imm): Use it.
5173 (aarch64_cannot_force_const_mem): Reject expressions involving
5174 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5175 (aarch64_classify_index): Handle SVE indices, by requiring
5176 a plain register index with a scale that matches the element size.
5177 (aarch64_classify_address): Handle SVE addresses. Assert that
5178 the mode of the address is VOIDmode or an integer mode.
5179 Update call to aarch64_classify_symbol.
5180 (aarch64_classify_symbolic_expression): Update call to
5181 aarch64_classify_symbol.
5182 (aarch64_const_vec_all_in_range_p): New function.
5183 (aarch64_print_vector_float_operand): Likewise.
5184 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5185 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5186 and the FP immediates 1.0 and 0.5.
5187 (aarch64_print_address_internal): Handle SVE addresses.
5188 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5189 (aarch64_regno_regclass): Handle predicate registers.
5190 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5192 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5193 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5194 (aarch64_convert_sve_vector_bits): New function.
5195 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5196 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5198 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5199 Handle SVE vector and predicate modes. Accept VL-based constants
5200 that need only one temporary register, and VL offsets that require
5201 no temporary registers.
5202 (aarch64_conditional_register_usage): Mark the predicate registers
5203 as fixed if SVE isn't available.
5204 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5205 Return true for SVE vector and predicate modes.
5206 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5207 rather than an unsigned int. Handle SVE modes.
5208 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5210 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5212 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5213 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5214 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5215 (aarch64_sve_float_mul_immediate_p): New functions.
5216 (aarch64_sve_valid_immediate): New function.
5217 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5218 Explicitly reject structure modes. Check for INDEX constants.
5219 Handle PTRUE and PFALSE constants.
5220 (aarch64_check_zero_based_sve_index_immediate): New function.
5221 (aarch64_simd_imm_zero_p): Delete.
5222 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5223 vector modes. Accept constants in the range of CNT[BHWD].
5224 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5225 ask for an Advanced SIMD mode.
5226 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5227 (aarch64_simd_vector_alignment): Handle SVE predicates.
5228 (aarch64_vectorize_preferred_vector_alignment): New function.
5229 (aarch64_simd_vector_alignment_reachable): Use it instead of
5231 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5232 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5234 (MAX_VECT_LEN): Delete.
5235 (expand_vec_perm_d): Add a vec_flags field.
5236 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5237 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5238 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5240 (aarch64_evpc_rev): Rename to...
5241 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5242 (aarch64_evpc_rev_global): New function.
5243 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5244 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5246 (aarch64_evpc_sve_tbl): New function.
5247 (aarch64_expand_vec_perm_const_1): Update after rename of
5248 aarch64_evpc_rev. Handle SVE permutes too, trying
5249 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5250 than aarch64_evpc_tbl.
5251 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5252 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5253 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5254 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5255 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5256 (aarch64_expand_sve_vcond): New functions.
5257 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5258 of aarch64_vector_mode_p.
5259 (aarch64_dwarf_poly_indeterminate_value): New function.
5260 (aarch64_compute_pressure_classes): Likewise.
5261 (aarch64_can_change_mode_class): Likewise.
5262 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5263 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5264 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5265 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5266 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5267 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5268 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5269 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5271 (Dn, Dl, Dr): Accept const as well as const_vector.
5272 (Dz): Likewise. Compare against CONST0_RTX.
5273 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5274 of "vector" where appropriate.
5275 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5276 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5277 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5278 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5279 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5280 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5281 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5282 (v_int_equiv): Extend to SVE modes.
5283 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5285 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5286 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5287 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5288 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5289 (SVE_COND_FP_CMP): New int iterators.
5290 (perm_hilo): Handle the new unpack unspecs.
5291 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5293 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5294 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5295 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5296 (aarch64_equality_operator, aarch64_constant_vector_operand)
5297 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5298 (aarch64_sve_nonimmediate_operand): Likewise.
5299 (aarch64_sve_general_operand): Likewise.
5300 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5301 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5302 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5303 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5304 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5305 (aarch64_sve_float_arith_immediate): Likewise.
5306 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5307 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5308 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5309 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5310 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5311 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5312 (aarch64_sve_float_arith_operand): Likewise.
5313 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5314 (aarch64_sve_float_mul_operand): Likewise.
5315 (aarch64_sve_vec_perm_operand): Likewise.
5316 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5317 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5318 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5319 as well as const_vector.
5320 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5321 in file. Use CONST0_RTX and CONSTM1_RTX.
5322 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5323 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5324 Use aarch64_simd_imm_zero.
5325 * config/aarch64/aarch64-sve.md: New file.
5326 * config/aarch64/aarch64.md: Include it.
5327 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5328 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5329 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5330 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5331 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5332 (sve): New attribute.
5333 (enabled): Disable instructions with the sve attribute unless
5335 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5336 aarch64_expand_mov_immediate.
5337 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5338 CNT[BHSD] immediates.
5339 (movti): Split CONST_POLY_INT moves into two halves.
5340 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5341 Split additions that need a temporary here if the destination
5342 is the stack pointer.
5343 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5344 (*add<mode>3_poly_1): New instruction.
5345 (set_clobber_cc): New expander.
5347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5349 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5350 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5351 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5352 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5353 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5354 Change innermode from fixed_mode_size to machine_mode.
5355 (simplify_subreg): Update call accordingly. Handle a constant-sized
5356 subreg of a variable-length CONST_VECTOR.
5358 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5359 Alan Hayward <alan.hayward@arm.com>
5360 David Sherwood <david.sherwood@arm.com>
5362 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5363 (add_offset_to_base): New function, split out from...
5364 (create_mem_ref): ...here. When handling a scale other than 1,
5365 check first whether the address is valid without the offset.
5366 Add it into the base if so, leaving the index and scale as-is.
5368 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5371 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5372 fold_for_warn before checking if arg2 is INTEGER_CST.
5374 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5376 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5377 (store_multiple_operation): Delete.
5378 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5379 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5380 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5381 guarded by TARGET_STRING.
5382 (rs6000_output_load_multiple): Delete.
5383 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5384 OPTION_MASK_STRING / TARGET_STRING handling.
5385 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5386 (const rs6000_opt_masks) <"string">: Change mask to 0.
5387 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5388 (MASK_STRING): Delete.
5389 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5391 (load_multiple): Delete.
5398 (store_multiple): Delete.
5405 (movmemsi_8reg): Delete.
5406 (corresponding unnamed define_insn): Delete.
5407 (movmemsi_6reg): Delete.
5408 (corresponding unnamed define_insn): Delete.
5409 (movmemsi_4reg): Delete.
5410 (corresponding unnamed define_insn): Delete.
5411 (movmemsi_2reg): Delete.
5412 (corresponding unnamed define_insn): Delete.
5413 (movmemsi_1reg): Delete.
5414 (corresponding unnamed define_insn): Delete.
5415 * config/rs6000/rs6000.opt (mno-string): New.
5416 (mstring): Replace by deprecation warning stub.
5417 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5419 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5421 * regrename.c (regrename_do_replace): If replacing the same
5422 reg multiple times, try to reuse last created gen_raw_REG.
5425 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5426 main to workaround a bug in GDB.
5428 2018-01-12 Tom de Vries <tom@codesourcery.com>
5431 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5433 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5435 PR rtl-optimization/80481
5436 * ira-color.c (get_cap_member): New function.
5437 (allocnos_conflict_by_live_ranges_p): Use it.
5438 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5439 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5441 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5444 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5445 (*saddl_se_1): Ditto.
5447 (*ssubl_se_1): Ditto.
5449 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5451 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5452 rather than wi::to_widest for DR_INITs.
5453 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5454 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5455 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5457 (vect_analyze_group_access_1): Note that here.
5459 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5461 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5462 polynomial type sizes.
5464 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5466 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5467 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5468 (gimple_add_tmp_var): Likewise.
5470 2018-01-12 Martin Liska <mliska@suse.cz>
5472 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5473 (gimple_alloc_sizes): Likewise.
5474 (dump_gimple_statistics): Use PRIu64 in printf format.
5475 * gimple.h: Change uint64_t to int.
5477 2018-01-12 Martin Liska <mliska@suse.cz>
5479 * tree-core.h: Use uint64_t instead of int.
5480 * tree.c (tree_node_counts): Likewise.
5481 (tree_node_sizes): Likewise.
5482 (dump_tree_statistics): Use PRIu64 in printf format.
5484 2018-01-12 Martin Liska <mliska@suse.cz>
5486 * Makefile.in: As qsort_chk is implemented in vec.c, add
5487 vec.o to linkage of gencfn-macros.
5488 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5489 passing the info to record_node_allocation_statistics.
5490 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5492 * ggc-common.c (struct ggc_usage): Add operator== and use
5493 it in operator< and compare function.
5494 * mem-stats.h (struct mem_usage): Likewise.
5495 * vec.c (struct vec_usage): Remove operator< and compare
5496 function. Can be simply inherited.
5498 2018-01-12 Martin Jambor <mjambor@suse.cz>
5501 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5502 * tree-ssa-math-opts.c: Include domwalk.h.
5503 (convert_mult_to_fma_1): New function.
5504 (fma_transformation_info): New type.
5505 (fma_deferring_state): Likewise.
5506 (cancel_fma_deferring): New function.
5507 (result_of_phi): Likewise.
5508 (last_fma_candidate_feeds_initial_phi): Likewise.
5509 (convert_mult_to_fma): Added deferring logic, split actual
5510 transformation to convert_mult_to_fma_1.
5511 (math_opts_dom_walker): New type.
5512 (math_opts_dom_walker::after_dom_children): New method, body moved
5513 here from pass_optimize_widening_mul::execute, added deferring logic
5515 (pass_optimize_widening_mul::execute): Moved most of code to
5516 math_opts_dom_walker::after_dom_children.
5517 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5518 * config/i386/i386.c (ix86_option_override_internal): Added
5519 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5521 2018-01-12 Richard Biener <rguenther@suse.de>
5524 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5525 inline instance vars.
5527 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5530 * config/rx/rx.c (rx_is_restricted_memory_address):
5533 2018-01-12 Richard Biener <rguenther@suse.de>
5535 PR tree-optimization/80846
5536 * target.def (split_reduction): New target hook.
5537 * targhooks.c (default_split_reduction): New function.
5538 * targhooks.h (default_split_reduction): Declare.
5539 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5540 target requests first reduce vectors by combining low and high
5542 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5543 (get_vectype_for_scalar_type_and_size): Export.
5544 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5545 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5546 * doc/tm.texi: Regenerate.
5547 * config/i386/i386.c (ix86_split_reduction): Implement
5548 TARGET_VECTORIZE_SPLIT_REDUCTION.
5550 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5553 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5554 in PIC mode except for TARGET_VXWORKS_RTP.
5555 * config/sparc/sparc.c: Include cfgrtl.h.
5556 (TARGET_INIT_PIC_REG): Define.
5557 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5558 (sparc_pic_register_p): New predicate.
5559 (sparc_legitimate_address_p): Use it.
5560 (sparc_legitimize_pic_address): Likewise.
5561 (sparc_delegitimize_address): Likewise.
5562 (sparc_mode_dependent_address_p): Likewise.
5563 (gen_load_pcrel_sym): Remove 4th parameter.
5564 (load_got_register): Adjust call to above. Remove obsolete stuff.
5565 (sparc_expand_prologue): Do not call load_got_register here.
5566 (sparc_flat_expand_prologue): Likewise.
5567 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5568 (sparc_use_pseudo_pic_reg): New function.
5569 (sparc_init_pic_reg): Likewise.
5570 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5571 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5573 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5575 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5576 Add item for branch_cost.
5578 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5580 PR rtl-optimization/83565
5581 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5582 not extend the result to a larger mode for rotate operations.
5583 (num_sign_bit_copies1): Likewise.
5585 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5588 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5590 Use values-Xc.o for -pedantic.
5591 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5593 2018-01-12 Martin Liska <mliska@suse.cz>
5596 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5598 (possible_polymorphic_call_targets): Use it.
5599 (ipa_devirt): Likewise.
5601 2018-01-12 Martin Liska <mliska@suse.cz>
5603 * profile-count.h (enum profile_quality): Use 0 as invalid
5604 enum value of profile_quality.
5606 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5608 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5609 -mext-string options.
5611 2018-01-12 Richard Biener <rguenther@suse.de>
5613 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5614 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5615 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5617 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5619 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5621 * configure.ac (--with-long-double-format): Add support for the
5622 configuration option to change the default long double format on
5624 * config.gcc (powerpc*-linux*-*): Likewise.
5625 * configure: Regenerate.
5626 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5627 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5628 used without modification.
5630 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5632 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5633 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5634 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5635 MISC_BUILTIN_SPEC_BARRIER.
5636 (rs6000_init_builtins): Likewise.
5637 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5639 (speculation_barrier): New define_insn.
5640 * doc/extend.texi: Document __builtin_speculation_barrier.
5642 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5645 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5646 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5647 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5649 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5650 integral modes instead of "ss" and "sd".
5651 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5652 vectors with 32-bit and 64-bit elements.
5653 (vecdupssescalarmodesuffix): New mode attribute.
5654 (vec_dup<mode>): Use it.
5656 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5659 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5660 frame if argument is passed on stack.
5662 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5665 * ree.c (combine_reaching_defs): Optimize also
5666 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5667 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5669 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5672 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5674 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5677 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5678 after they are computed.
5680 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5682 PR tree-optimization/83695
5683 * gimple-loop-linterchange.cc
5684 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5685 reset cached scev information after interchange.
5686 (pass_linterchange::execute): Remove call to scev_reset_htab.
5688 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5690 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5691 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5692 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5693 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5694 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5695 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5696 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5697 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5698 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5699 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5700 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5701 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5702 (V_lane_reg): Likewise.
5703 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5705 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5706 (vfmal_lane_low<mode>_intrinsic,
5707 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5708 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5709 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5710 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5711 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5712 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5714 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5716 * config/arm/arm-cpus.in (fp16fml): New feature.
5717 (ALL_SIMD): Add fp16fml.
5718 (armv8.2-a): Add fp16fml as an option.
5719 (armv8.3-a): Likewise.
5720 (armv8.4-a): Add fp16fml as part of fp16.
5721 * config/arm/arm.h (TARGET_FP16FML): Define.
5722 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5724 * config/arm/arm-modes.def (V2HF): Define.
5725 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5726 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5727 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5728 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5729 vfmsl_low, vfmsl_high): New set of builtins.
5730 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5731 (vfml_op): New code attribute.
5732 (VFMLHALVES): New int iterator.
5733 (VFML, VFMLSEL): New mode attributes.
5734 (V_reg): Define mapping for V2HF.
5735 (V_hi, V_lo): New mode attributes.
5736 (VF_constraint): Likewise.
5737 (vfml_half, vfml_half_selector): New int attributes.
5738 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5740 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5741 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5743 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5744 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5745 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5746 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5748 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5749 Document new effective target and option set.
5751 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5753 * config/arm/arm-cpus.in (armv8_4): New feature.
5754 (ARMv8_4a): New fgroup.
5755 (armv8.4-a): New arch.
5756 * config/arm/arm-tables.opt: Regenerate.
5757 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5758 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5759 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5760 Add matching rules for -march=armv8.4-a and extensions.
5761 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5763 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5766 * config/rx/rx.md (BW): New mode attribute.
5767 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5769 2018-01-11 Richard Biener <rguenther@suse.de>
5771 PR tree-optimization/83435
5772 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5773 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5774 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5776 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5777 Alan Hayward <alan.hayward@arm.com>
5778 David Sherwood <david.sherwood@arm.com>
5780 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5782 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5783 (aarch64_print_address_internal): Use it to check for a zero offset.
5785 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5786 Alan Hayward <alan.hayward@arm.com>
5787 David Sherwood <david.sherwood@arm.com>
5789 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5790 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5791 Return a poly_int64 rather than a HOST_WIDE_INT.
5792 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5793 rather than a HOST_WIDE_INT.
5794 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5795 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5796 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5797 final_offset from HOST_WIDE_INT to poly_int64.
5798 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5799 to_constant when getting the number of units in an Advanced SIMD
5801 (aarch64_builtin_vectorized_function): Check for a constant number
5803 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5805 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5806 attribute instead of GET_MODE_NUNITS.
5807 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5808 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5809 GET_MODE_SIZE for fixed-size registers.
5810 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5811 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5812 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5813 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5814 (aarch64_print_operand, aarch64_print_address_internal)
5815 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5816 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5817 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5818 Handle polynomial GET_MODE_SIZE.
5819 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5820 wider than SImode without modification.
5821 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5822 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5823 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5824 passing and returning SVE modes.
5825 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5826 rather than GEN_INT.
5827 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5828 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5829 (aarch64_allocate_and_probe_stack_space): Likewise.
5830 (aarch64_layout_frame): Cope with polynomial offsets.
5831 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5832 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5834 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5835 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5836 poly_int64 rather than a HOST_WIDE_INT.
5837 (aarch64_get_separate_components, aarch64_process_components)
5838 (aarch64_expand_prologue, aarch64_expand_epilogue)
5839 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5840 (aarch64_anchor_offset): New function, split out from...
5841 (aarch64_legitimize_address): ...here.
5842 (aarch64_builtin_vectorization_cost): Handle polynomial
5843 TYPE_VECTOR_SUBPARTS.
5844 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5846 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5847 number of elements from the PARALLEL rather than the mode.
5848 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5849 rather than GET_MODE_BITSIZE.
5850 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5851 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5852 (aarch64_expand_vec_perm_const_1): Handle polynomial
5853 d->perm.length () and d->perm elements.
5854 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5855 Apply to_constant to d->perm elements.
5856 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5857 polynomial CONST_VECTOR_NUNITS.
5858 (aarch64_move_pointer): Take amount as a poly_int64 rather
5860 (aarch64_progress_pointer): Avoid temporary variable.
5861 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5862 the mode attribute instead of GET_MODE.
5864 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5865 Alan Hayward <alan.hayward@arm.com>
5866 David Sherwood <david.sherwood@arm.com>
5868 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5869 x exists before using it.
5870 (aarch64_add_constant_internal): Rename to...
5871 (aarch64_add_offset_1): ...this. Replace regnum with separate
5872 src and dest rtxes. Handle the case in which they're different,
5873 including when the offset is zero. Replace scratchreg with an rtx.
5874 Use 2 additions if there is no spare register into which we can
5875 move a 16-bit constant.
5876 (aarch64_add_constant): Delete.
5877 (aarch64_add_offset): Replace reg with separate src and dest
5878 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5879 Use aarch64_add_offset_1.
5880 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5881 an rtx rather than an int. Take the delta as a poly_int64
5882 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5883 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5884 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5885 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5886 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5888 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5889 aarch64_add_constant.
5891 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5893 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5894 Use scalar_float_mode.
5896 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5898 * config/aarch64/aarch64-simd.md
5899 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5900 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5901 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5902 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5903 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5904 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5905 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5906 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5907 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5908 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5910 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5913 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5914 targ_options->x_arm_arch_string is non NULL.
5916 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5918 * config/aarch64/aarch64.h
5919 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5921 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5924 * expmed.c (emit_store_flag_force): Swap if const op0
5925 and change VOIDmode to mode of op0.
5927 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5929 PR rtl-optimization/83761
5930 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5931 than bytes to mode_for_size.
5933 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5936 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5937 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5940 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5943 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5944 when in layout mode.
5945 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5946 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5949 2018-01-10 Michael Collison <michael.collison@arm.com>
5951 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5952 * config/aarch64/aarch64-option-extension.def: Add
5953 AARCH64_OPT_EXTENSION of 'fp16fml'.
5954 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5955 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5956 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5957 * config/aarch64/constraints.md (Ui7): New constraint.
5958 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5959 (VFMLA_SEL_W): Ditto.
5962 (VFMLA16_LOW): New int iterator.
5963 (VFMLA16_HIGH): Ditto.
5964 (UNSPEC_FMLAL): New unspec.
5965 (UNSPEC_FMLSL): Ditto.
5966 (UNSPEC_FMLAL2): Ditto.
5967 (UNSPEC_FMLSL2): Ditto.
5968 (f16mac): New code attribute.
5969 * config/aarch64/aarch64-simd-builtins.def
5970 (aarch64_fmlal_lowv2sf): Ditto.
5971 (aarch64_fmlsl_lowv2sf): Ditto.
5972 (aarch64_fmlalq_lowv4sf): Ditto.
5973 (aarch64_fmlslq_lowv4sf): Ditto.
5974 (aarch64_fmlal_highv2sf): Ditto.
5975 (aarch64_fmlsl_highv2sf): Ditto.
5976 (aarch64_fmlalq_highv4sf): Ditto.
5977 (aarch64_fmlslq_highv4sf): Ditto.
5978 (aarch64_fmlal_lane_lowv2sf): Ditto.
5979 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5980 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5981 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5982 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5983 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5984 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5985 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5986 (aarch64_fmlal_lane_highv2sf): Ditto.
5987 (aarch64_fmlsl_lane_highv2sf): Ditto.
5988 (aarch64_fmlal_laneq_highv2sf): Ditto.
5989 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5990 (aarch64_fmlalq_lane_highv4sf): Ditto.
5991 (aarch64_fmlsl_lane_highv4sf): Ditto.
5992 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5993 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5994 * config/aarch64/aarch64-simd.md:
5995 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5996 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5997 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5998 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5999 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6000 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6001 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6002 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6003 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6004 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6005 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6006 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6007 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6008 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6009 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6010 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6011 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6012 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6013 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6014 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6015 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6016 (vfmlsl_low_u32): Ditto.
6017 (vfmlalq_low_u32): Ditto.
6018 (vfmlslq_low_u32): Ditto.
6019 (vfmlal_high_u32): Ditto.
6020 (vfmlsl_high_u32): Ditto.
6021 (vfmlalq_high_u32): Ditto.
6022 (vfmlslq_high_u32): Ditto.
6023 (vfmlal_lane_low_u32): Ditto.
6024 (vfmlsl_lane_low_u32): Ditto.
6025 (vfmlal_laneq_low_u32): Ditto.
6026 (vfmlsl_laneq_low_u32): Ditto.
6027 (vfmlalq_lane_low_u32): Ditto.
6028 (vfmlslq_lane_low_u32): Ditto.
6029 (vfmlalq_laneq_low_u32): Ditto.
6030 (vfmlslq_laneq_low_u32): Ditto.
6031 (vfmlal_lane_high_u32): Ditto.
6032 (vfmlsl_lane_high_u32): Ditto.
6033 (vfmlal_laneq_high_u32): Ditto.
6034 (vfmlsl_laneq_high_u32): Ditto.
6035 (vfmlalq_lane_high_u32): Ditto.
6036 (vfmlslq_lane_high_u32): Ditto.
6037 (vfmlalq_laneq_high_u32): Ditto.
6038 (vfmlslq_laneq_high_u32): Ditto.
6039 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6040 (AARCH64_FL_FOR_ARCH8_4): New.
6041 (AARCH64_ISA_F16FML): New ISA flag.
6042 (TARGET_F16FML): New feature flag for fp16fml.
6043 (doc/invoke.texi): Document new fp16fml option.
6045 2018-01-10 Michael Collison <michael.collison@arm.com>
6047 * config/aarch64/aarch64-builtins.c:
6048 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6049 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6050 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6051 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6052 (AARCH64_ISA_SHA3): New ISA flag.
6053 (TARGET_SHA3): New feature flag for sha3.
6054 * config/aarch64/iterators.md (sha512_op): New int attribute.
6055 (CRYPTO_SHA512): New int iterator.
6056 (UNSPEC_SHA512H): New unspec.
6057 (UNSPEC_SHA512H2): Ditto.
6058 (UNSPEC_SHA512SU0): Ditto.
6059 (UNSPEC_SHA512SU1): Ditto.
6060 * config/aarch64/aarch64-simd-builtins.def
6061 (aarch64_crypto_sha512hqv2di): New builtin.
6062 (aarch64_crypto_sha512h2qv2di): Ditto.
6063 (aarch64_crypto_sha512su0qv2di): Ditto.
6064 (aarch64_crypto_sha512su1qv2di): Ditto.
6065 (aarch64_eor3qv8hi): Ditto.
6066 (aarch64_rax1qv2di): Ditto.
6067 (aarch64_xarqv2di): Ditto.
6068 (aarch64_bcaxqv8hi): Ditto.
6069 * config/aarch64/aarch64-simd.md:
6070 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6071 (aarch64_crypto_sha512su0qv2di): Ditto.
6072 (aarch64_crypto_sha512su1qv2di): Ditto.
6073 (aarch64_eor3qv8hi): Ditto.
6074 (aarch64_rax1qv2di): Ditto.
6075 (aarch64_xarqv2di): Ditto.
6076 (aarch64_bcaxqv8hi): Ditto.
6077 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6078 (vsha512h2q_u64): Ditto.
6079 (vsha512su0q_u64): Ditto.
6080 (vsha512su1q_u64): Ditto.
6081 (veor3q_u16): Ditto.
6082 (vrax1q_u64): Ditto.
6084 (vbcaxq_u16): Ditto.
6085 * config/arm/types.md (crypto_sha512): New type attribute.
6086 (crypto_sha3): Ditto.
6087 (doc/invoke.texi): Document new sha3 option.
6089 2018-01-10 Michael Collison <michael.collison@arm.com>
6091 * config/aarch64/aarch64-builtins.c:
6092 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6093 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6094 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6095 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6096 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6097 (AARCH64_ISA_SM4): New ISA flag.
6098 (TARGET_SM4): New feature flag for sm4.
6099 * config/aarch64/aarch64-simd-builtins.def
6100 (aarch64_sm3ss1qv4si): Ditto.
6101 (aarch64_sm3tt1aq4si): Ditto.
6102 (aarch64_sm3tt1bq4si): Ditto.
6103 (aarch64_sm3tt2aq4si): Ditto.
6104 (aarch64_sm3tt2bq4si): Ditto.
6105 (aarch64_sm3partw1qv4si): Ditto.
6106 (aarch64_sm3partw2qv4si): Ditto.
6107 (aarch64_sm4eqv4si): Ditto.
6108 (aarch64_sm4ekeyqv4si): Ditto.
6109 * config/aarch64/aarch64-simd.md:
6110 (aarch64_sm3ss1qv4si): Ditto.
6111 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6112 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6113 (aarch64_sm4eqv4si): Ditto.
6114 (aarch64_sm4ekeyqv4si): Ditto.
6115 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6116 (sm3part_op): Ditto.
6117 (CRYPTO_SM3TT): Ditto.
6118 (CRYPTO_SM3PART): Ditto.
6119 (UNSPEC_SM3SS1): New unspec.
6120 (UNSPEC_SM3TT1A): Ditto.
6121 (UNSPEC_SM3TT1B): Ditto.
6122 (UNSPEC_SM3TT2A): Ditto.
6123 (UNSPEC_SM3TT2B): Ditto.
6124 (UNSPEC_SM3PARTW1): Ditto.
6125 (UNSPEC_SM3PARTW2): Ditto.
6126 (UNSPEC_SM4E): Ditto.
6127 (UNSPEC_SM4EKEY): Ditto.
6128 * config/aarch64/constraints.md (Ui2): New constraint.
6129 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6130 * config/arm/types.md (crypto_sm3): New type attribute.
6131 (crypto_sm4): Ditto.
6132 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6133 (vsm3tt1aq_u32): Ditto.
6134 (vsm3tt1bq_u32): Ditto.
6135 (vsm3tt2aq_u32): Ditto.
6136 (vsm3tt2bq_u32): Ditto.
6137 (vsm3partw1q_u32): Ditto.
6138 (vsm3partw2q_u32): Ditto.
6139 (vsm4eq_u32): Ditto.
6140 (vsm4ekeyq_u32): Ditto.
6141 (doc/invoke.texi): Document new sm4 option.
6143 2018-01-10 Michael Collison <michael.collison@arm.com>
6145 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6146 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6147 (AARCH64_FL_FOR_ARCH8_4): New.
6148 (AARCH64_FL_V8_4): New flag.
6149 (doc/invoke.texi): Document new armv8.4-a option.
6151 2018-01-10 Michael Collison <michael.collison@arm.com>
6153 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6154 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6155 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6156 * config/aarch64/aarch64-option-extension.def: Add
6157 AARCH64_OPT_EXTENSION of 'sha2'.
6158 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6159 (crypto): Disable sha2 and aes if crypto disabled.
6160 (crypto): Enable aes and sha2 if enabled.
6161 (simd): Disable sha2 and aes if simd disabled.
6162 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6164 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6165 (TARGET_SHA2): New feature flag for sha2.
6166 (TARGET_AES): New feature flag for aes.
6167 * config/aarch64/aarch64-simd.md:
6168 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6169 conditional on TARGET_AES.
6170 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6171 (aarch64_crypto_sha1hsi): Make pattern conditional
6173 (aarch64_crypto_sha1hv4si): Ditto.
6174 (aarch64_be_crypto_sha1hv4si): Ditto.
6175 (aarch64_crypto_sha1su1v4si): Ditto.
6176 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6177 (aarch64_crypto_sha1su0v4si): Ditto.
6178 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6179 (aarch64_crypto_sha256su0v4si): Ditto.
6180 (aarch64_crypto_sha256su1v4si): Ditto.
6181 (doc/invoke.texi): Document new aes and sha2 options.
6183 2018-01-10 Martin Sebor <msebor@redhat.com>
6185 PR tree-optimization/83781
6186 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6189 2018-01-11 Martin Sebor <msebor@gmail.com>
6190 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6192 PR tree-optimization/83501
6193 PR tree-optimization/81703
6195 * tree-ssa-strlen.c (get_string_cst): Rename...
6196 (get_string_len): ...to this. Handle global constants.
6197 (handle_char_store): Adjust.
6199 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6200 Jim Wilson <jimw@sifive.com>
6202 * config/riscv/riscv-protos.h (riscv_output_return): New.
6203 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6204 (riscv_attribute_table, riscv_output_return),
6205 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6206 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6207 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6209 (riscv_expand_prologue): Add early return for naked function.
6210 (riscv_expand_epilogue): Likewise.
6211 (riscv_function_ok_for_sibcall): Return false for naked function.
6212 (riscv_set_current_function): New.
6213 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6214 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6215 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6216 * doc/extend.texi (RISC-V Function Attributes): New.
6218 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6220 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6221 check for 128-bit long double before checking TCmode.
6222 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6223 128-bit long doubles before checking TFmode or TCmode.
6224 (FLOAT128_IBM_P): Likewise.
6226 2018-01-10 Martin Sebor <msebor@redhat.com>
6228 PR tree-optimization/83671
6229 * builtins.c (c_strlen): Unconditionally return zero for the empty
6231 Use -Warray-bounds for warnings.
6232 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6233 for non-constant array indices with COMPONENT_REF, arrays of
6234 arrays, and pointers to arrays.
6235 (gimple_fold_builtin_strlen): Determine and set length range for
6236 non-constant character arrays.
6238 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6241 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6244 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6246 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6248 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6251 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6252 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6253 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6254 indexed_or_indirect_operand predicate.
6255 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6256 (*vsx_le_perm_load_v8hi): Likewise.
6257 (*vsx_le_perm_load_v16qi): Likewise.
6258 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6259 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6260 (*vsx_le_perm_store_v8hi): Likewise.
6261 (*vsx_le_perm_store_v16qi): Likewise.
6262 (eight unnamed splitters): Likewise.
6264 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6266 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6267 * config/rs6000/emmintrin.h: Likewise.
6268 * config/rs6000/mmintrin.h: Likewise.
6269 * config/rs6000/xmmintrin.h: Likewise.
6271 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6274 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6276 * tree.c (tree_nop_conversion): Return true for location wrapper
6278 (maybe_wrap_with_location): New function.
6279 (selftest::check_strip_nops): New function.
6280 (selftest::test_location_wrappers): New function.
6281 (selftest::tree_c_tests): Call it.
6282 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6283 (maybe_wrap_with_location): New decl.
6284 (EXPR_LOCATION_WRAPPER_P): New macro.
6285 (location_wrapper_p): New inline function.
6286 (tree_strip_any_location_wrapper): New inline function.
6288 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6291 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6292 stack_realign_offset for the largest alignment of stack slot
6294 (ix86_find_max_used_stack_alignment): New function.
6295 (ix86_finalize_stack_frame_flags): Use it. Set
6296 max_used_stack_alignment if we don't realign stack.
6297 * config/i386/i386.h (machine_function): Add
6298 max_used_stack_alignment.
6300 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6302 * config/arm/arm.opt (-mbranch-cost): New option.
6303 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6306 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6309 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6310 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6312 2018-01-10 Richard Biener <rguenther@suse.de>
6315 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6316 early out so it also covers the case where we have a non-NULL
6319 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6321 PR tree-optimization/83753
6322 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6323 for non-strided grouped accesses if the number of elements is 1.
6325 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6328 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6329 * i386.h (TARGET_USE_GATHER): Define.
6330 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6332 2018-01-10 Martin Liska <mliska@suse.cz>
6335 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6336 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6338 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6339 CLEANUP_NO_PARTITIONING is not set.
6341 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6343 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6344 for vectors, as a partial revert of r254296.
6345 * rtl.h (const_vec_p): Delete.
6346 (const_vec_duplicate_p): Don't test for vector CONSTs.
6347 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6348 * expmed.c (make_tree): Likewise.
6351 * common.md (E, F): Use CONSTANT_P instead of checking for
6353 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6354 checking for CONST_VECTOR.
6356 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6359 * predict.c (force_edge_cold): Handle in more sane way edges
6362 2018-01-09 Carl Love <cel@us.ibm.com>
6364 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6366 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6367 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6368 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6369 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6370 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6371 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6372 * config/rs6000/rs6000-protos.h: Add extern defition for
6373 rs6000_generate_float2_double_code.
6374 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6376 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6377 (float2_v2df): Add define_expand.
6379 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6382 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6383 op_mode in the force_to_mode call.
6385 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6387 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6388 instead of checking each element individually.
6389 (aarch64_evpc_uzp): Likewise.
6390 (aarch64_evpc_zip): Likewise.
6391 (aarch64_evpc_ext): Likewise.
6392 (aarch64_evpc_rev): Likewise.
6393 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6394 instead of checking each element individually. Return true without
6396 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6397 whether all selected elements come from the same input, instead of
6398 checking each element individually. Remove calls to gen_rtx_REG,
6399 start_sequence and end_sequence and instead assert that no rtl is
6402 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6404 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6405 order of HIGH and CONST checks.
6407 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6409 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6410 if the destination isn't an SSA_NAME.
6412 2018-01-09 Richard Biener <rguenther@suse.de>
6414 PR tree-optimization/83668
6415 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6417 (canonicalize_loop_form): ... here, renamed from ...
6418 (canonicalize_loop_closed_ssa_form): ... this and amended to
6419 swap successor edges for loop exit blocks to make us use
6420 the RPO order we need for initial schedule generation.
6422 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6424 PR tree-optimization/64811
6425 * match.pd: When optimizing comparisons with Inf, avoid
6426 introducing or losing exceptions from comparisons with NaN.
6428 2018-01-09 Martin Liska <mliska@suse.cz>
6431 * asan.c (shadow_mem_size): Add gcc_assert.
6433 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6435 Don't save registers in main().
6438 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6439 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6440 * config/avr/avr.c (avr_set_current_function): Don't error if
6441 naked, OS_task or OS_main are specified at the same time.
6442 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6444 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6446 * common/config/avr/avr-common.c (avr_option_optimization_table):
6447 Switch on -mmain-is-OS_task for optimizing compilations.
6449 2018-01-09 Richard Biener <rguenther@suse.de>
6451 PR tree-optimization/83572
6452 * graphite.c: Include cfganal.h.
6453 (graphite_transform_loops): Connect infinite loops to exit
6454 and remove fake edges at the end.
6456 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6458 * ipa-inline.c (edge_badness): Revert accidental checkin.
6460 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6463 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6464 symbols; not inline clones.
6466 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6469 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6470 hard registers. Formatting fixes.
6472 PR preprocessor/83722
6473 * gcc.c (try_generate_repro): Pass
6474 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6475 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6478 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6479 Kito Cheng <kito.cheng@gmail.com>
6481 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6482 (riscv_leaf_function_p): Delete.
6483 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6485 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6487 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6489 (do_ifelse): New function.
6490 (do_isel): New function.
6491 (do_sub3): New function.
6492 (do_add3): New function.
6493 (do_load_mask_compare): New function.
6494 (do_overlap_load_compare): New function.
6495 (expand_compare_loop): New function.
6496 (expand_block_compare): Call expand_compare_loop() when appropriate.
6497 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6499 (-mblock-compare-inline-loop-limit): New option.
6501 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6504 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6505 Reverse order of second and third operands in first alternative.
6506 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6507 of first and second elements in UNSPEC_VPERMR vector.
6508 (altivec_expand_vec_perm_le): Likewise.
6510 2018-01-08 Jeff Law <law@redhat.com>
6512 PR rtl-optimizatin/81308
6513 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6514 (process_switch): If group_case_labels makes a change, then set
6516 (pass_convert_switch::execute): If a switch is converted, then
6517 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6519 PR rtl-optimization/81308
6520 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6523 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6525 PR target/83663 - Revert r255946
6526 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6527 generation for cases where splatting a value is not useful.
6528 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6529 across a vec_duplicate and a paradoxical subreg forming a vector
6530 mode to a vec_concat.
6532 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6534 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6535 -march=armv8.3-a variants.
6536 * config/arm/t-multilib: Likewise.
6537 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6539 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6541 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6543 (cceq_ior_compare_complement): Give it a name so I can use it, and
6544 change boolean_or_operator predicate to boolean_operator so it can
6545 be used to generate a crand.
6546 (eqne): New code iterator.
6547 (bd/bd_neg): New code_attrs.
6548 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6549 a single define_insn.
6550 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6551 decrement (bdnzt/bdnzf/bdzt/bdzf).
6552 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6553 with the new names of the branch decrement patterns, and added the
6554 names of the branch decrement conditional patterns.
6556 2018-01-08 Richard Biener <rguenther@suse.de>
6558 PR tree-optimization/83563
6559 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6562 2018-01-08 Richard Biener <rguenther@suse.de>
6565 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6567 2018-01-08 Richard Biener <rguenther@suse.de>
6569 PR tree-optimization/83685
6570 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6571 references to abnormals.
6573 2018-01-08 Richard Biener <rguenther@suse.de>
6576 * dwarf2out.c (output_indirect_strings): Handle empty
6577 skeleton_debug_str_hash.
6578 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6580 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6582 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6583 (emit_store_direct): Likewise.
6584 (arc_trampoline_adjust_address): Likewise.
6585 (arc_asm_trampoline_template): New function.
6586 (arc_initialize_trampoline): Use asm_trampoline_template.
6587 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6588 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6589 * config/arc/arc.md (flush_icache): Delete pattern.
6591 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6593 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6594 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6597 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6600 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6601 by not USED_FOR_TARGET.
6602 (make_pass_resolve_sw_modes): Likewise.
6604 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6606 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6609 2018-01-08 Richard Biener <rguenther@suse.de>
6612 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6614 2018-01-08 Richard Biener <rguenther@suse.de>
6617 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6619 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6622 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6623 basic blocks with a small number of successors.
6624 (convert_control_dep_chain_into_preds): Improve handling of
6626 (dump_predicates): Split apart into...
6627 (dump_pred_chain): ...here...
6628 (dump_pred_info): ...and here.
6629 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6630 (can_chain_union_be_invalidated_p): Improve check for invalidation
6632 (uninit_uses_cannot_happen): Avoid unnecessary if
6633 convert_control_dep_chain_into_preds yielded nothing.
6635 2018-01-06 Martin Sebor <msebor@redhat.com>
6637 PR tree-optimization/83640
6638 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6639 subtracting negative offset from size.
6640 (builtin_access::overlap): Adjust offset bounds of the access to fall
6641 within the size of the object if possible.
6643 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6645 PR rtl-optimization/83699
6646 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6647 extract_bit_field_as_subreg to cases in which the extracted
6648 value is also a vector.
6650 * lra-constraints.c (process_alt_operands): Test for the equivalence
6651 substitutions when detecting a possible reload cycle.
6653 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6656 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6657 by default if flag_selective_schedling{,2}. Formatting fixes.
6659 PR rtl-optimization/83682
6660 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6661 if it has non-VECTOR_MODE element mode.
6662 (vec_duplicate_p): Likewise.
6665 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6666 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6668 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6671 * config/i386/i386-builtin.def
6672 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6673 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6674 Require also OPTION_MASK_ISA_AVX512F in addition to
6675 OPTION_MASK_ISA_GFNI.
6676 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6677 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6678 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6679 to OPTION_MASK_ISA_GFNI.
6680 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6681 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6682 OPTION_MASK_ISA_AVX512BW.
6683 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6684 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6685 addition to OPTION_MASK_ISA_GFNI.
6686 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6687 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6688 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6689 to OPTION_MASK_ISA_GFNI.
6690 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6691 a requirement for all ISAs rather than any of them with a few
6693 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6695 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6696 bitmasks to be enabled with 3 exceptions, instead of requiring any
6697 enabled ISA with lots of exceptions.
6698 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6699 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6700 Change avx512bw in isa attribute to avx512f.
6701 * config/i386/sgxintrin.h: Add license boilerplate.
6702 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6703 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6704 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6705 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6707 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6708 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6709 temporarily sse2 rather than sse if not enabled already.
6712 * config/i386/sse.md (VI248_VLBW): Rename to ...
6713 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6714 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6715 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6716 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6717 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6718 mode iterator instead of VI248_VLBW.
6720 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6722 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6723 (record_modified): Skip clobbers; add debug output.
6724 (param_change_prob): Use sreal frequencies.
6726 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6728 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6729 punt for user-aligned variables.
6731 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6733 * tree-chrec.c (chrec_contains_symbols): Return true for
6736 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6739 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6740 of (x|y) == x for BICS pattern.
6742 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6744 PR tree-optimization/83605
6745 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6746 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6749 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6751 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6752 * config/epiphany/rtems.h: New file.
6754 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6755 Uros Bizjak <ubizjak@gmail.com>
6758 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6759 QIreg_operand instead of register_operand predicate.
6760 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6761 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6762 comments instead of -fmitigate[-_]rop.
6764 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6767 * cgraphunit.c (symbol_table::compile): Switch to text_section
6768 before calling assembly_start debug hook.
6769 * run-rtl-passes.c (run_rtl_passes): Likewise.
6772 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6774 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6775 range_int_cst_p rather than !symbolic_range_p before calling
6776 extract_range_from_multiplicative_op_1.
6778 2018-01-04 Jeff Law <law@redhat.com>
6780 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6781 redundant test in assertion.
6783 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6785 * doc/rtl.texi: Document machine_mode wrapper classes.
6787 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6789 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6792 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6794 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6795 the VEC_PERM_EXPR fold to fail.
6797 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6800 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6801 to switched_sections.
6803 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6806 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6809 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6812 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6813 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6815 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6818 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6819 is BLKmode and bitpos not zero or mode change is needed.
6821 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6824 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6827 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6830 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6831 instead of MULT rtx. Update all corresponding splitters.
6833 (*ssub<modesuffix>): Ditto.
6835 (*cmp_sadd_di): Update split patterns.
6836 (*cmp_sadd_si): Ditto.
6837 (*cmp_sadd_sidi): Ditto.
6838 (*cmp_ssub_di): Ditto.
6839 (*cmp_ssub_si): Ditto.
6840 (*cmp_ssub_sidi): Ditto.
6841 * config/alpha/predicates.md (const23_operand): New predicate.
6842 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6843 Look for ASHIFT, not MULT inner operand.
6844 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6846 2018-01-04 Martin Liska <mliska@suse.cz>
6848 PR gcov-profile/83669
6849 * gcov.c (output_intermediate_file): Add version to intermediate
6851 * doc/gcov.texi: Document new field 'version' in intermediate
6852 file format. Fix location of '-k' option of gcov command.
6854 2018-01-04 Martin Liska <mliska@suse.cz>
6857 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6859 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6861 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6863 2018-01-03 Martin Sebor <msebor@redhat.com>
6865 PR tree-optimization/83655
6866 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6867 checking calls with invalid arguments.
6869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6871 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6872 (vectorizable_mask_load_store): Delete.
6873 (vectorizable_call): Return false for masked loads and stores.
6874 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6875 instead of gimple_assign_rhs1.
6876 (vectorizable_load): Handle IFN_MASK_LOAD.
6877 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6881 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6883 (vectorizable_mask_load_store): ...here.
6884 (vectorizable_load): ...and here.
6886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6888 * tree-vect-stmts.c (vect_build_all_ones_mask)
6889 (vect_build_zero_merge_argument): New functions, split out from...
6890 (vectorizable_load): ...here.
6892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6894 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6896 (vectorizable_mask_load_store): ...here.
6897 (vectorizable_store): ...and here.
6899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6901 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6903 (vectorizable_mask_load_store): ...here.
6905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6907 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6908 (vect_model_store_cost): Take a vec_load_store_type instead of a
6910 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6911 (vect_model_store_cost): Take a vec_load_store_type instead of a
6913 (vectorizable_mask_load_store): Update accordingly.
6914 (vectorizable_store): Likewise.
6915 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6919 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6920 IFN_MASK_LOAD calls here rather than...
6921 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6924 Alan Hayward <alan.hayward@arm.com>
6925 David Sherwood <david.sherwood@arm.com>
6927 * expmed.c (extract_bit_field_1): For vector extracts,
6928 fall back to extract_bit_field_as_subreg if vec_extract
6931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6932 Alan Hayward <alan.hayward@arm.com>
6933 David Sherwood <david.sherwood@arm.com>
6935 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6936 they are variable or constant sized.
6937 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6938 slots for constant-sized data.
6940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6941 Alan Hayward <alan.hayward@arm.com>
6942 David Sherwood <david.sherwood@arm.com>
6944 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6945 handling COND_EXPRs with boolean comparisons, try to find a better
6946 basis for the mask type than the boolean itself.
6948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6950 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6951 is calculated and how it can be overridden.
6952 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6953 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6955 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6959 Alan Hayward <alan.hayward@arm.com>
6960 David Sherwood <david.sherwood@arm.com>
6962 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6963 Remove the mode argument.
6964 (aarch64_simd_valid_immediate): Remove the mode and inverse
6966 * config/aarch64/iterators.md (bitsize): New iterator.
6967 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6968 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6969 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6970 aarch64_simd_valid_immediate.
6971 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6972 (aarch64_reg_or_bic_imm): Likewise.
6973 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6974 with an insn_type enum and msl with a modifier_type enum.
6975 Replace element_width with a scalar_mode. Change the shift
6976 to unsigned int. Add constructors for scalar_float_mode and
6977 scalar_int_mode elements.
6978 (aarch64_vect_float_const_representable_p): Delete.
6979 (aarch64_can_const_movi_rtx_p)
6980 (aarch64_simd_scalar_immediate_valid_for_move)
6981 (aarch64_simd_make_constant): Update call to
6982 aarch64_simd_valid_immediate.
6983 (aarch64_advsimd_valid_immediate_hs): New function.
6984 (aarch64_advsimd_valid_immediate): Likewise.
6985 (aarch64_simd_valid_immediate): Remove mode and inverse
6986 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6987 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6988 and aarch64_float_const_representable_p on the result.
6989 (aarch64_output_simd_mov_immediate): Remove mode argument.
6990 Update call to aarch64_simd_valid_immediate and use of
6991 simd_immediate_info.
6992 (aarch64_output_scalar_simd_mov_immediate): Update call
6995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6996 Alan Hayward <alan.hayward@arm.com>
6997 David Sherwood <david.sherwood@arm.com>
6999 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7000 (mode_nunits): Likewise CONST_MODE_NUNITS.
7001 * machmode.def (ADJUST_NUNITS): Document.
7002 * genmodes.c (mode_data::need_nunits_adj): New field.
7003 (blank_mode): Update accordingly.
7004 (adj_nunits): New variable.
7005 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7007 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7008 listed in adj_nunits.
7009 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7010 listed in adj_nunits. Don't emit case statements for such modes.
7011 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7012 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7013 nothing if adj_nunits is nonnull.
7014 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7015 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7016 (emit_mode_fbit): Update use of print_maybe_const_decl.
7017 (emit_move_size): Likewise. Treat the array as non-const
7019 (emit_mode_adjustments): Handle adj_nunits.
7021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7023 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7024 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7025 (VECTOR_MODES): Use it.
7026 (make_vector_modes): Take the prefix as an argument.
7028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7029 Alan Hayward <alan.hayward@arm.com>
7030 David Sherwood <david.sherwood@arm.com>
7032 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7033 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7034 for MODE_VECTOR_BOOL.
7035 * machmode.def (VECTOR_BOOL_MODE): Document.
7036 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7037 (make_vector_bool_mode): New function.
7038 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7040 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7041 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7043 * stor-layout.c (int_mode_for_mode): Likewise.
7044 * tree.c (build_vector_type_for_mode): Likewise.
7045 * varasm.c (output_constant_pool_2): Likewise.
7046 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7047 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7048 for MODE_VECTOR_BOOL.
7049 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7050 of mode class checks.
7051 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7052 instead of a list of mode class checks.
7053 (expand_vector_scalar_condition): Likewise.
7054 (type_for_widest_vector_mode): Handle BImode as an inner mode.
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 * machmode.h (mode_size): Change from unsigned short to
7062 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7063 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7064 or if measurement_type is not polynomial.
7065 (fixed_size_mode::includes_p): Check for constant-sized modes.
7066 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7067 return a poly_uint16 rather than an unsigned short.
7068 (emit_mode_size): Change the type of mode_size from unsigned short
7069 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7070 (emit_mode_adjustments): Cope with polynomial vector sizes.
7071 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7073 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7075 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7076 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7077 * caller-save.c (setup_save_areas): Likewise.
7078 (replace_reg_with_saved_mem): Likewise.
7079 * calls.c (emit_library_call_value_1): Likewise.
7080 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7081 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7082 (gen_lowpart_for_combine): Likewise.
7083 * convert.c (convert_to_integer_1): Likewise.
7084 * cse.c (equiv_constant, cse_insn): Likewise.
7085 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7086 (cselib_subst_to_values): Likewise.
7087 * dce.c (word_dce_process_block): Likewise.
7088 * df-problems.c (df_word_lr_mark_ref): Likewise.
7089 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7090 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7091 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7092 (rtl_for_decl_location): Likewise.
7093 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7094 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7095 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7096 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7097 (expand_expr_real_1): Likewise.
7098 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7099 (pad_below): Likewise.
7100 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7101 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7102 * ira.c (get_subreg_tracking_sizes): Likewise.
7103 * ira-build.c (ira_create_allocno_objects): Likewise.
7104 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7105 (ira_sort_regnos_for_alter_reg): Likewise.
7106 * ira-costs.c (record_operand_costs): Likewise.
7107 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7108 (resolve_simple_move): Likewise.
7109 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7110 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7111 (lra_constraints): Likewise.
7112 (CONST_POOL_OK_P): Reject variable-sized modes.
7113 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7114 (add_pseudo_to_slot, lra_spill): Likewise.
7115 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7116 * optabs-query.c (get_best_extraction_insn): Likewise.
7117 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7118 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7119 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7120 * recog.c (offsettable_address_addr_space_p): Likewise.
7121 * regcprop.c (maybe_mode_change): Likewise.
7122 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7123 * regrename.c (build_def_use): Likewise.
7124 * regstat.c (dump_reg_info): Likewise.
7125 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7126 (find_reloads, find_reloads_subreg_address): Likewise.
7127 * reload1.c (eliminate_regs_1): Likewise.
7128 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7129 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7130 (simplify_binary_operation_1, simplify_subreg): Likewise.
7131 * targhooks.c (default_function_arg_padding): Likewise.
7132 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7133 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7134 (verify_gimple_assign_ternary): Likewise.
7135 * tree-inline.c (estimate_move_cost): Likewise.
7136 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7137 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7138 (get_address_cost_ainc): Likewise.
7139 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7140 (vect_supportable_dr_alignment): Likewise.
7141 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7142 (vectorizable_reduction): Likewise.
7143 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7144 (vectorizable_operation, vectorizable_load): Likewise.
7145 * tree.c (build_same_sized_truth_vector_type): Likewise.
7146 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7147 * var-tracking.c (emit_note_insn_var_location): Likewise.
7148 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7149 (ADDR_VEC_ALIGN): Likewise.
7151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7152 Alan Hayward <alan.hayward@arm.com>
7153 David Sherwood <david.sherwood@arm.com>
7155 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7157 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7158 or if measurement_type is polynomial.
7159 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7160 * combine.c (make_extraction): Likewise.
7161 * dse.c (find_shift_sequence): Likewise.
7162 * dwarf2out.c (mem_loc_descriptor): Likewise.
7163 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7164 (extract_bit_field, extract_low_bits): Likewise.
7165 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7166 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7167 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7168 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7169 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7170 * reload.c (find_reloads): Likewise.
7171 * reload1.c (alter_reg): Likewise.
7172 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7173 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7174 * tree-if-conv.c (predicate_mem_writes): Likewise.
7175 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7176 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7177 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7178 * valtrack.c (dead_debug_insert_temp): Likewise.
7179 * varasm.c (mergeable_constant_section): Likewise.
7180 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7183 Alan Hayward <alan.hayward@arm.com>
7184 David Sherwood <david.sherwood@arm.com>
7186 * expr.c (expand_assignment): Cope with polynomial mode sizes
7187 when assigning to a CONCAT.
7189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7190 Alan Hayward <alan.hayward@arm.com>
7191 David Sherwood <david.sherwood@arm.com>
7193 * machmode.h (mode_precision): Change from unsigned short to
7195 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7197 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7198 or if measurement_type is not polynomial.
7199 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7200 in which the mode is already known to be a scalar_int_mode.
7201 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7202 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7204 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7205 for GET_MODE_PRECISION.
7206 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7207 for GET_MODE_PRECISION.
7208 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7210 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7211 (expand_field_assignment, make_extraction): Likewise.
7212 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7213 (get_last_value): Likewise.
7214 * convert.c (convert_to_integer_1): Likewise.
7215 * cse.c (cse_insn): Likewise.
7216 * expr.c (expand_expr_real_1): Likewise.
7217 * lra-constraints.c (simplify_operand_subreg): Likewise.
7218 * optabs-query.c (can_atomic_load_p): Likewise.
7219 * optabs.c (expand_atomic_load): Likewise.
7220 (expand_atomic_store): Likewise.
7221 * ree.c (combine_reaching_defs): Likewise.
7222 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7223 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7224 * tree.h (type_has_mode_precision_p): Likewise.
7225 * ubsan.c (instrument_si_overflow): Likewise.
7227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7228 Alan Hayward <alan.hayward@arm.com>
7229 David Sherwood <david.sherwood@arm.com>
7231 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7232 polynomial numbers of units.
7233 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7234 (valid_vector_subparts_p): New function.
7235 (build_vector_type): Remove temporary shim and take the number
7236 of units as a poly_uint64 rather than an int.
7237 (build_opaque_vector_type): Take the number of units as a
7238 poly_uint64 rather than an int.
7239 * tree.c (build_vector_from_ctor): Handle polynomial
7240 TYPE_VECTOR_SUBPARTS.
7241 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7242 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7243 (build_vector_from_val): If the number of units is variable,
7244 use build_vec_duplicate_cst for constant operands and
7245 VEC_DUPLICATE_EXPR otherwise.
7246 (make_vector_type): Remove temporary is_constant ().
7247 (build_vector_type, build_opaque_vector_type): Take the number of
7248 units as a poly_uint64 rather than an int.
7249 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7251 * cfgexpand.c (expand_debug_expr): Likewise.
7252 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7253 (store_constructor, expand_expr_real_1): Likewise.
7254 (const_scalar_mask_from_tree): Likewise.
7255 * fold-const-call.c (fold_const_reduction): Likewise.
7256 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7257 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7258 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7259 (fold_relational_const): Likewise.
7260 (native_interpret_vector): Likewise. Change the size from an
7261 int to an unsigned int.
7262 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7263 TYPE_VECTOR_SUBPARTS.
7264 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7265 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7266 duplicating a non-constant operand into a variable-length vector.
7267 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7268 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7269 * ipa-icf.c (sem_variable::equals): Likewise.
7270 * match.pd: Likewise.
7271 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7272 * print-tree.c (print_node): Likewise.
7273 * stor-layout.c (layout_type): Likewise.
7274 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7275 * tree-cfg.c (verify_gimple_comparison): Likewise.
7276 (verify_gimple_assign_binary): Likewise.
7277 (verify_gimple_assign_ternary): Likewise.
7278 (verify_gimple_assign_single): Likewise.
7279 * tree-pretty-print.c (dump_generic_node): Likewise.
7280 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7281 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7282 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7283 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7284 (vect_shift_permute_load_chain): Likewise.
7285 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7286 (expand_vector_condition, optimize_vector_constructor): Likewise.
7287 (lower_vec_perm, get_compute_type): Likewise.
7288 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7289 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7290 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7291 (vect_recog_mask_conversion_pattern): Likewise.
7292 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7293 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7294 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7295 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7296 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7297 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7298 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7299 (supportable_widening_operation): Likewise.
7300 (supportable_narrowing_operation): Likewise.
7301 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7303 * varasm.c (output_constant): Likewise.
7305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7306 Alan Hayward <alan.hayward@arm.com>
7307 David Sherwood <david.sherwood@arm.com>
7309 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7310 so that both the length == 3 and length != 3 cases set up their
7311 own permute vectors. Add comments explaining why we know the
7312 number of elements is constant.
7313 (vect_permute_load_chain): Likewise.
7315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7316 Alan Hayward <alan.hayward@arm.com>
7317 David Sherwood <david.sherwood@arm.com>
7319 * machmode.h (mode_nunits): Change from unsigned char to
7321 (ONLY_FIXED_SIZE_MODES): New macro.
7322 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7323 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7324 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7326 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7327 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7328 or if measurement_type is not polynomial.
7329 * genmodes.c (ZERO_COEFFS): New macro.
7330 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7332 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7333 Use ZERO_COEFFS when emitting initializers.
7334 * data-streamer.h (bp_pack_poly_value): New function.
7335 (bp_unpack_poly_value): Likewise.
7336 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7337 for GET_MODE_NUNITS.
7338 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7339 for GET_MODE_NUNITS.
7340 * tree.c (make_vector_type): Remove temporary shim and make
7341 the real function take the number of units as a poly_uint64
7343 (build_vector_type_for_mode): Handle polynomial nunits.
7344 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7345 * emit-rtl.c (const_vec_series_p_1): Likewise.
7346 (gen_rtx_CONST_VECTOR): Likewise.
7347 * fold-const.c (test_vec_duplicate_folding): Likewise.
7348 * genrecog.c (validate_pattern): Likewise.
7349 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7350 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7351 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7352 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7353 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7354 * rtlanal.c (subreg_get_info): Likewise.
7355 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7356 (vect_grouped_load_supported): Likewise.
7357 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7358 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7359 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7360 (simplify_const_unary_operation, simplify_binary_operation_1)
7361 (simplify_const_binary_operation, simplify_ternary_operation)
7362 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7363 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7364 instead of CONST_VECTOR_NUNITS.
7365 * varasm.c (output_constant_pool_2): Likewise.
7366 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7367 explicit-encoded elements in the XVEC for variable-length vectors.
7369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7371 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7374 Alan Hayward <alan.hayward@arm.com>
7375 David Sherwood <david.sherwood@arm.com>
7377 * coretypes.h (fixed_size_mode): Declare.
7378 (fixed_size_mode_pod): New typedef.
7379 * builtins.h (target_builtins::x_apply_args_mode)
7380 (target_builtins::x_apply_result_mode): Change type to
7381 fixed_size_mode_pod.
7382 * builtins.c (apply_args_size, apply_result_size, result_vector)
7383 (expand_builtin_apply_args_1, expand_builtin_apply)
7384 (expand_builtin_return): Update accordingly.
7386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7388 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7389 * cselib.c (cselib_hash_rtx): Likewise.
7390 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7391 CONST_VECTOR encoding.
7393 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7394 Jeff Law <law@redhat.com>
7397 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7398 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7399 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7400 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7403 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7404 explicitly probe *sp in a noreturn function if there were any callee
7405 register saves or frame pointer is needed.
7407 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7410 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7411 BLKmode for ternary, binary or unary expressions.
7414 * var-tracking.c (delete_vta_debug_insn): New inline function.
7415 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7416 insns from get_insns () to NULL instead of each bb separately.
7417 Use delete_vta_debug_insn. No longer static.
7418 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7419 delete_vta_debug_insns callers.
7420 * rtl.h (delete_vta_debug_insns): Declare.
7421 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7422 instead of variable_tracking_main.
7424 2018-01-03 Martin Sebor <msebor@redhat.com>
7426 PR tree-optimization/83603
7427 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7428 arguments past the endof the argument list in functions declared
7429 without a prototype.
7430 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7431 Avoid checking when arguments are null.
7433 2018-01-03 Martin Sebor <msebor@redhat.com>
7436 * doc/extend.texi (attribute const): Fix a typo.
7437 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7438 issuing -Wsuggest-attribute for void functions.
7440 2018-01-03 Martin Sebor <msebor@redhat.com>
7442 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7443 offset_int::from instead of wide_int::to_shwi.
7444 (maybe_diag_overlap): Remove assertion.
7445 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7446 * gimple-ssa-sprintf.c (format_directive): Same.
7447 (parse_directive): Same.
7448 (sprintf_dom_walker::compute_format_length): Same.
7449 (try_substitute_return_value): Same.
7451 2018-01-03 Jeff Law <law@redhat.com>
7454 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7455 non-constant residual for zero at runtime and avoid probing in
7456 that case. Reorganize code for trailing problem to mirror handling
7459 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7461 PR tree-optimization/83501
7462 * tree-ssa-strlen.c (get_string_cst): New.
7463 (handle_char_store): Call get_string_cst.
7465 2018-01-03 Martin Liska <mliska@suse.cz>
7467 PR tree-optimization/83593
7468 * tree-ssa-strlen.c: Include tree-cfg.h.
7469 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7470 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7471 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7473 (strlen_dom_walker::before_dom_children): Call
7474 gimple_purge_dead_eh_edges. Dump tranformation with details
7476 (strlen_dom_walker::before_dom_children): Update call by adding
7477 new argument cleanup_eh.
7478 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7480 2018-01-03 Martin Liska <mliska@suse.cz>
7483 * cif-code.def (VARIADIC_THUNK): New enum value.
7484 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7487 2018-01-03 Jan Beulich <jbeulich@suse.com>
7489 * sse.md (mov<mode>_internal): Tighten condition for when to use
7490 vmovdqu<ssescalarsize> for TI and OI modes.
7492 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7494 Update copyright years.
7496 2018-01-03 Martin Liska <mliska@suse.cz>
7499 * ipa-visibility.c (function_and_variable_visibility): Skip
7500 functions with noipa attribure.
7502 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7504 * gcc.c (process_command): Update copyright notice dates.
7505 * gcov-dump.c (print_version): Ditto.
7506 * gcov.c (print_version): Ditto.
7507 * gcov-tool.c (print_version): Ditto.
7508 * gengtype.c (create_file): Ditto.
7509 * doc/cpp.texi: Bump @copying's copyright year.
7510 * doc/cppinternals.texi: Ditto.
7511 * doc/gcc.texi: Ditto.
7512 * doc/gccint.texi: Ditto.
7513 * doc/gcov.texi: Ditto.
7514 * doc/install.texi: Ditto.
7515 * doc/invoke.texi: Ditto.
7517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7519 * vector-builder.h (vector_builder::m_full_nelts): Change from
7520 unsigned int to poly_uint64.
7521 (vector_builder::full_nelts): Update prototype accordingly.
7522 (vector_builder::new_vector): Likewise.
7523 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7524 (vector_builder::operator ==): Likewise.
7525 (vector_builder::finalize): Likewise.
7526 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7527 Take the number of elements as a poly_uint64 rather than an
7529 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7530 from unsigned int to poly_uint64.
7531 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7532 (vec_perm_indices::new_vector): Likewise.
7533 (vec_perm_indices::length): Likewise.
7534 (vec_perm_indices::nelts_per_input): Likewise.
7535 (vec_perm_indices::input_nelts): Likewise.
7536 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7537 number of elements per input as a poly_uint64 rather than an
7538 unsigned int. Use the original encoding for variable-length
7539 vectors, rather than clamping each individual element.
7540 For the second and subsequent elements in each pattern,
7541 clamp the step and base before clamping their sum.
7542 (vec_perm_indices::series_p): Handle polynomial element counts.
7543 (vec_perm_indices::all_in_range_p): Likewise.
7544 (vec_perm_indices_to_tree): Likewise.
7545 (vec_perm_indices_to_rtx): Likewise.
7546 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7547 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7548 (tree_vector_builder::new_binary_operation): Handle polynomial
7549 element counts. Return false if we need to know the number
7550 of elements at compile time.
7551 * fold-const.c (fold_vec_perm): Punt if the number of elements
7552 isn't known at compile time.
7554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7556 * vec-perm-indices.h (vec_perm_builder): Change element type
7557 from HOST_WIDE_INT to poly_int64.
7558 (vec_perm_indices::element_type): Update accordingly.
7559 (vec_perm_indices::clamp): Handle polynomial element_types.
7560 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7561 (vec_perm_indices::all_in_range_p): Likewise.
7562 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7564 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7565 polynomial vec_perm_indices element types.
7566 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7567 * fold-const.c (fold_vec_perm): Likewise.
7568 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7569 * tree-vect-generic.c (lower_vec_perm): Likewise.
7570 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7571 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7572 element type to HOST_WIDE_INT.
7574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7575 Alan Hayward <alan.hayward@arm.com>
7576 David Sherwood <david.sherwood@arm.com>
7578 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7579 rather than an int. Use plus_constant.
7580 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7581 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7584 Alan Hayward <alan.hayward@arm.com>
7585 David Sherwood <david.sherwood@arm.com>
7587 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7588 a HOST_WIDE_INT to a poly_int64.
7590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7591 Alan Hayward <alan.hayward@arm.com>
7592 David Sherwood <david.sherwood@arm.com>
7594 * calls.c (load_register_parameters): Cope with polynomial
7595 mode sizes. Require a constant size for BLKmode parameters
7596 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7597 forces a parameter to be padded at the lsb end in order to
7598 fill a complete number of words, require the parameter size
7599 to be ordered wrt UNITS_PER_WORD.
7601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7602 Alan Hayward <alan.hayward@arm.com>
7603 David Sherwood <david.sherwood@arm.com>
7605 * reload1.c (spill_stack_slot_width): Change element type
7606 from unsigned int to poly_uint64_pod.
7607 (alter_reg): Treat mode sizes as polynomial.
7609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7610 Alan Hayward <alan.hayward@arm.com>
7611 David Sherwood <david.sherwood@arm.com>
7613 * reload.c (complex_word_subreg_p): New function.
7614 (reload_inner_reg_of_subreg, push_reload): Use it.
7616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7617 Alan Hayward <alan.hayward@arm.com>
7618 David Sherwood <david.sherwood@arm.com>
7620 * lra-constraints.c (process_alt_operands): Reject matched
7621 operands whose sizes aren't ordered.
7622 (match_reload): Refer to this check here.
7624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7625 Alan Hayward <alan.hayward@arm.com>
7626 David Sherwood <david.sherwood@arm.com>
7628 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7629 that the mode size is in the set {1, 2, 4, 8, 16}.
7631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7632 Alan Hayward <alan.hayward@arm.com>
7633 David Sherwood <david.sherwood@arm.com>
7635 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7636 Use plus_constant instead of gen_rtx_PLUS.
7638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7639 Alan Hayward <alan.hayward@arm.com>
7640 David Sherwood <david.sherwood@arm.com>
7642 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7643 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7644 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7645 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7646 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7647 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7648 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7649 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7650 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7651 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7653 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7654 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7655 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7656 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7657 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7658 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7659 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7660 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7661 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7662 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7664 * expr.c (emit_move_resolve_push): Treat the input and result
7665 of PUSH_ROUNDING as a poly_int64.
7666 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7667 (emit_push_insn): Likewise.
7668 * lra-eliminations.c (mark_not_eliminable): Likewise.
7669 * recog.c (push_operand): Likewise.
7670 * reload1.c (elimination_effects): Likewise.
7671 * rtlanal.c (nonzero_bits1): Likewise.
7672 * calls.c (store_one_arg): Likewise. Require the padding to be
7673 known at compile time.
7675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7676 Alan Hayward <alan.hayward@arm.com>
7677 David Sherwood <david.sherwood@arm.com>
7679 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7680 Use plus_constant instead of gen_rtx_PLUS.
7682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7683 Alan Hayward <alan.hayward@arm.com>
7684 David Sherwood <david.sherwood@arm.com>
7686 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7690 Alan Hayward <alan.hayward@arm.com>
7691 David Sherwood <david.sherwood@arm.com>
7693 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7694 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7695 via stack temporaries. Treat the mode size as polynomial too.
7697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7698 Alan Hayward <alan.hayward@arm.com>
7699 David Sherwood <david.sherwood@arm.com>
7701 * expr.c (expand_expr_real_2): When handling conversions involving
7702 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7703 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7704 as a poly_uint64 too.
7706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7707 Alan Hayward <alan.hayward@arm.com>
7708 David Sherwood <david.sherwood@arm.com>
7710 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7713 Alan Hayward <alan.hayward@arm.com>
7714 David Sherwood <david.sherwood@arm.com>
7716 * combine.c (can_change_dest_mode): Handle polynomial
7717 REGMODE_NATURAL_SIZE.
7718 * expmed.c (store_bit_field_1): Likewise.
7719 * expr.c (store_constructor): Likewise.
7720 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7721 and polynomial REGMODE_NATURAL_SIZE.
7722 (gen_lowpart_common): Likewise.
7723 * reginfo.c (record_subregs_of_mode): Likewise.
7724 * rtlanal.c (read_modify_subreg_p): Likewise.
7726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7727 Alan Hayward <alan.hayward@arm.com>
7728 David Sherwood <david.sherwood@arm.com>
7730 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7731 numbers of elements.
7733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7734 Alan Hayward <alan.hayward@arm.com>
7735 David Sherwood <david.sherwood@arm.com>
7737 * match.pd: Cope with polynomial numbers of vector elements.
7739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7740 Alan Hayward <alan.hayward@arm.com>
7741 David Sherwood <david.sherwood@arm.com>
7743 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7744 in a POINTER_PLUS_EXPR.
7746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7747 Alan Hayward <alan.hayward@arm.com>
7748 David Sherwood <david.sherwood@arm.com>
7750 * omp-simd-clone.c (simd_clone_subparts): New function.
7751 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7752 (ipa_simd_modify_function_body): 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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7759 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7760 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7761 (expand_vector_condition, vector_element): Likewise.
7762 (subparts_gt): New function.
7763 (get_compute_type): Use subparts_gt.
7764 (count_type_subparts): Delete.
7765 (expand_vector_operations_1): Use subparts_gt instead of
7766 count_type_subparts.
7768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7769 Alan Hayward <alan.hayward@arm.com>
7770 David Sherwood <david.sherwood@arm.com>
7772 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7773 (vect_compile_time_alias): ...this new function. Do the calculation
7774 on poly_ints rather than trees.
7775 (vect_prune_runtime_alias_test_list): Update call accordingly.
7777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7778 Alan Hayward <alan.hayward@arm.com>
7779 David Sherwood <david.sherwood@arm.com>
7781 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7783 (vect_schedule_slp_instance): Likewise.
7785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7786 Alan Hayward <alan.hayward@arm.com>
7787 David Sherwood <david.sherwood@arm.com>
7789 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7790 constant and extern definitions for variable-length vectors.
7791 (vect_get_constant_vectors): Note that the number of units
7792 is known to be constant.
7794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7795 Alan Hayward <alan.hayward@arm.com>
7796 David Sherwood <david.sherwood@arm.com>
7798 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7799 of units as polynomial. Choose between WIDE and NARROW based
7802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7803 Alan Hayward <alan.hayward@arm.com>
7804 David Sherwood <david.sherwood@arm.com>
7806 * tree-vect-stmts.c (simd_clone_subparts): New function.
7807 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7810 Alan Hayward <alan.hayward@arm.com>
7811 David Sherwood <david.sherwood@arm.com>
7813 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7814 vectors as polynomial. Use build_index_vector for
7817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7818 Alan Hayward <alan.hayward@arm.com>
7819 David Sherwood <david.sherwood@arm.com>
7821 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7822 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7823 for variable-length vectors.
7824 (vectorizable_mask_load_store): Treat the number of units as
7825 polynomial, asserting that it is constant if the condition has
7826 already been enforced.
7827 (vectorizable_store, vectorizable_load): Likewise.
7829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7830 Alan Hayward <alan.hayward@arm.com>
7831 David Sherwood <david.sherwood@arm.com>
7833 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7834 of units as polynomial. Punt if we can't tell at compile time
7835 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
7842 of units as polynomial. Punt on SLP inductions. Use an integer
7843 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7844 cast of such a series for variable-length floating-point
7847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7848 Alan Hayward <alan.hayward@arm.com>
7849 David Sherwood <david.sherwood@arm.com>
7851 * tree.h (build_index_vector): Declare.
7852 * tree.c (build_index_vector): New function.
7853 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7854 of units as polynomial, forcibly converting it to a constant if
7855 vectorizable_reduction has already enforced the condition.
7856 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7857 to create a {1,2,3,...} vector.
7858 (vectorizable_reduction): Treat the number of units as polynomial.
7859 Choose vectype_in based on the largest scalar element size rather
7860 than the smallest number of units. Enforce the restrictions
7863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7864 Alan Hayward <alan.hayward@arm.com>
7865 David Sherwood <david.sherwood@arm.com>
7867 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7868 number of units as polynomial.
7870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7871 Alan Hayward <alan.hayward@arm.com>
7872 David Sherwood <david.sherwood@arm.com>
7874 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7875 * target.def (autovectorize_vector_sizes): Return the vector sizes
7876 by pointer, using vector_sizes rather than a bitmask.
7877 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7878 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7879 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7881 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7882 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7883 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7884 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7885 * omp-general.c (omp_max_vf): Likewise.
7886 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7887 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7888 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7889 * tree-vect-slp.c (vect_slp_bb): Likewise.
7890 * doc/tm.texi: Regenerate.
7891 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7893 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7894 the vector size as a poly_uint64 rather than an unsigned int.
7895 (current_vector_size): Change from an unsigned int to a poly_uint64.
7896 (get_vectype_for_scalar_type): Update accordingly.
7897 * tree.h (build_truth_vector_type): Take the size and number of
7898 units as a poly_uint64 rather than an unsigned int.
7899 (build_vector_type): Add a temporary overload that takes
7900 the number of units as a poly_uint64 rather than an unsigned int.
7901 * tree.c (make_vector_type): Likewise.
7902 (build_truth_vector_type): Take the number of units as a poly_uint64
7903 rather than an unsigned int.
7905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7906 Alan Hayward <alan.hayward@arm.com>
7907 David Sherwood <david.sherwood@arm.com>
7909 * target.def (get_mask_mode): Take the number of units and length
7910 as poly_uint64s rather than unsigned ints.
7911 * targhooks.h (default_get_mask_mode): Update accordingly.
7912 * targhooks.c (default_get_mask_mode): Likewise.
7913 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7914 * doc/tm.texi: Regenerate.
7916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7917 Alan Hayward <alan.hayward@arm.com>
7918 David Sherwood <david.sherwood@arm.com>
7920 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7921 * omp-general.c (omp_max_vf): Likewise.
7922 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7923 (expand_omp_simd): Handle polynomial safelen.
7924 * omp-low.c (omplow_simd_context): Add a default constructor.
7925 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7926 (lower_rec_simd_input_clauses): Update accordingly.
7927 (lower_rec_input_clauses): Likewise.
7929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7930 Alan Hayward <alan.hayward@arm.com>
7931 David Sherwood <david.sherwood@arm.com>
7933 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7934 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7935 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7936 (vect_analyze_slp_cost): Likewise.
7937 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7938 (vect_model_load_cost): Likewise.
7940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7941 Alan Hayward <alan.hayward@arm.com>
7942 David Sherwood <david.sherwood@arm.com>
7944 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7945 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7946 from an unsigned int * to a poly_uint64_pod *.
7947 (calculate_unrolling_factor): New function.
7948 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7951 Alan Hayward <alan.hayward@arm.com>
7952 David Sherwood <david.sherwood@arm.com>
7954 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7955 from an unsigned int to a poly_uint64.
7956 (_loop_vec_info::slp_unrolling_factor): Likewise.
7957 (_loop_vec_info::vectorization_factor): Change from an int
7959 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7960 (vect_get_num_vectors): New function.
7961 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7962 (vect_get_num_copies): Use vect_get_num_vectors.
7963 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7964 to an unsigned int *.
7965 (vect_analyze_data_refs): Change min_vf from an int * to a
7967 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7968 than an unsigned HOST_WIDE_INT.
7969 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7970 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7971 to an unsigned int *.
7972 (vect_analyze_data_ref_dependences): Likewise.
7973 (vect_compute_data_ref_alignment): Handle polynomial vf.
7974 (vect_enhance_data_refs_alignment): Likewise.
7975 (vect_prune_runtime_alias_test_list): Likewise.
7976 (vect_shift_permute_load_chain): Likewise.
7977 (vect_supportable_dr_alignment): Likewise.
7978 (dependence_distance_ge_vf): Take the vectorization factor as a
7979 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7980 (vect_analyze_data_refs): Change min_vf from an int * to a
7982 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7983 vfm1 as a poly_uint64 rather than an int. Make the same change
7984 for the returned bound_scalar.
7985 (vect_gen_vector_loop_niters): Handle polynomial vf.
7986 (vect_do_peeling): Likewise. Update call to
7987 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7988 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7990 * tree-vect-loop.c (vect_determine_vectorization_factor)
7991 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7992 (vect_get_known_peeling_cost): Likewise.
7993 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7994 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7995 (vect_transform_loop): Likewise. Use the lowest possible VF when
7996 updating the upper bounds of the loop.
7997 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7999 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8000 polynomial unroll factors.
8001 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8002 (vect_make_slp_decision): Likewise.
8003 (vect_supported_load_permutation_p): Likewise, and polynomial
8005 (vect_analyze_slp_cost): Handle polynomial vf.
8006 (vect_slp_analyze_node_operations): Likewise.
8007 (vect_slp_analyze_bb_1): Likewise.
8008 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8009 than an unsigned HOST_WIDE_INT.
8010 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8011 (vectorizable_load): Handle polynomial vf.
8012 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8014 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8017 Alan Hayward <alan.hayward@arm.com>
8018 David Sherwood <david.sherwood@arm.com>
8020 * match.pd: Handle bit operations involving three constants
8021 and try to fold one pair.
8023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8025 * tree-vect-loop-manip.c: Include gimple-fold.h.
8026 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8027 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8028 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8029 Add a path that uses a step of VF instead of 1, but disable it
8031 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8032 and niters_no_overflow parameters. Update calls to
8033 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8034 Create a new SSA name if the latter choses to use a ste other
8035 than zero, and return it via niters_vector_mult_vf_var.
8036 * tree-vect-loop.c (vect_transform_loop): Update calls to
8037 vect_do_peeling, vect_gen_vector_loop_niters and
8038 slpeel_make_loop_iterate_ntimes.
8039 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8040 (vect_gen_vector_loop_niters): Update declarations after above changes.
8042 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8044 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8045 128-bit round to integer instructions.
8046 (ceil<mode>2): Likewise.
8047 (btrunc<mode>2): Likewise.
8048 (round<mode>2): Likewise.
8050 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8052 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8053 unaligned VSX load/store on P8/P9.
8054 (expand_block_clear): Allow the use of unaligned VSX
8055 load/store on P8/P9.
8057 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8059 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8061 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8062 swap associated with both a load and a store.
8064 2018-01-02 Andrew Waterman <andrew@sifive.com>
8066 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8067 * config/riscv/riscv.md (clear_cache): Use it.
8069 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8071 * web.c: Remove out-of-date comment.
8073 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8075 * expr.c (fixup_args_size_notes): Check that any existing
8076 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8077 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8078 (emit_single_push_insn): ...here.
8080 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8082 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8083 (const_vector_encoded_nelts): New function.
8084 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8085 (const_vector_int_elt, const_vector_elt): Declare.
8086 * emit-rtl.c (const_vector_int_elt_1): New function.
8087 (const_vector_elt): Likewise.
8088 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8089 of CONST_VECTOR_ELT.
8091 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8093 * expr.c: Include rtx-vector-builder.h.
8094 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8095 directly on the tree encoding.
8096 (const_vector_from_tree): Likewise.
8097 * optabs.c: Include rtx-vector-builder.h.
8098 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8099 sequence of "u" values.
8100 * vec-perm-indices.c: Include rtx-vector-builder.h.
8101 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8102 directly on the vec_perm_indices encoding.
8104 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8106 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8107 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8108 * rtx-vector-builder.h: New file.
8109 * rtx-vector-builder.c: Likewise.
8110 * rtl.h (rtx_def::u2): Add a const_vector field.
8111 (CONST_VECTOR_NPATTERNS): New macro.
8112 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8113 (CONST_VECTOR_DUPLICATE_P): Likewise.
8114 (CONST_VECTOR_STEPPED_P): Likewise.
8115 (CONST_VECTOR_ENCODED_ELT): Likewise.
8116 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8117 (unwrap_const_vec_duplicate): Likewise.
8118 (const_vec_series_p): Check for a non-duplicated vector encoding.
8119 Say that the function only returns true for integer vectors.
8120 * emit-rtl.c: Include rtx-vector-builder.h.
8121 (gen_const_vec_duplicate_1): Delete.
8122 (gen_const_vector): Call gen_const_vec_duplicate instead of
8123 gen_const_vec_duplicate_1.
8124 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8125 (gen_const_vec_duplicate): Use rtx_vector_builder.
8126 (gen_const_vec_series): Likewise.
8127 (gen_rtx_CONST_VECTOR): Likewise.
8128 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8129 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8130 Build a new vector rather than modifying a CONST_VECTOR in-place.
8131 (handle_special_swappables): Update call accordingly.
8132 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8133 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8134 Build a new vector rather than modifying a CONST_VECTOR in-place.
8135 (handle_special_swappables): Update call accordingly.
8137 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8139 * simplify-rtx.c (simplify_const_binary_operation): Use
8140 CONST_VECTOR_ELT instead of XVECEXP.
8142 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8144 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8145 the selector elements to be different from the data elements
8146 if the selector is a VECTOR_CST.
8147 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8148 ssizetype for the selector.
8150 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8152 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8153 before testing each element individually.
8154 * tree-vect-generic.c (lower_vec_perm): Likewise.
8156 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8158 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8159 * selftest-run-tests.c (selftest::run_tests): Call it.
8160 * vector-builder.h (vector_builder::operator ==): New function.
8161 (vector_builder::operator !=): Likewise.
8162 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8163 (vec_perm_indices::all_from_input_p): New function.
8164 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8165 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8166 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8167 instead of reading the VECTOR_CST directly. Detect whether both
8168 vector inputs are the same before constructing the vec_perm_indices,
8169 and update the number of inputs argument accordingly. Use the
8170 utility functions added above. Only construct sel2 if we need to.
8172 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8174 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8175 the broadcast of the low byte.
8176 (expand_mult_highpart): Use an explicit encoding for the permutes.
8177 * optabs-query.c (can_mult_highpart_p): Likewise.
8178 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8179 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8180 (vectorizable_bswap): Likewise.
8181 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8182 explicit encoding for the power-of-2 permutes.
8183 (vect_permute_store_chain): Likewise.
8184 (vect_grouped_load_supported): Likewise.
8185 (vect_permute_load_chain): Likewise.
8187 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8189 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8190 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8191 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8192 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8193 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8194 (vect_gen_perm_mask_any): Likewise.
8196 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8198 * int-vector-builder.h: New file.
8199 * vec-perm-indices.h: Include int-vector-builder.h.
8200 (vec_perm_indices): Redefine as an int_vector_builder.
8201 (auto_vec_perm_indices): Delete.
8202 (vec_perm_builder): Redefine as a stand-alone class.
8203 (vec_perm_indices::vec_perm_indices): New function.
8204 (vec_perm_indices::clamp): Likewise.
8205 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8206 (vec_perm_indices::new_vector): New function.
8207 (vec_perm_indices::new_expanded_vector): Update for new
8208 vec_perm_indices class.
8209 (vec_perm_indices::rotate_inputs): New function.
8210 (vec_perm_indices::all_in_range_p): Operate directly on the
8211 encoded form, without computing elided elements.
8212 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8213 encoding. Update for new vec_perm_indices class.
8214 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8215 the given vec_perm_builder.
8216 (expand_vec_perm_var): Update vec_perm_builder constructor.
8217 (expand_mult_highpart): Use vec_perm_builder instead of
8218 auto_vec_perm_indices.
8219 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8220 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8221 or double series encoding as appropriate.
8222 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8223 vec_perm_indices instead of auto_vec_perm_indices.
8224 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8225 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8226 (vect_permute_store_chain): Likewise.
8227 (vect_grouped_load_supported): Likewise.
8228 (vect_permute_load_chain): Likewise.
8229 (vect_shift_permute_load_chain): Likewise.
8230 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8231 (vect_transform_slp_perm_load): Likewise.
8232 (vect_schedule_slp_instance): Likewise.
8233 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8234 (vectorizable_mask_load_store): Likewise.
8235 (vectorizable_bswap): Likewise.
8236 (vectorizable_store): Likewise.
8237 (vectorizable_load): Likewise.
8238 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8239 vec_perm_indices instead of auto_vec_perm_indices. Use
8240 tree_to_vec_perm_builder to read the vector from a tree.
8241 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8242 vec_perm_builder instead of a vec_perm_indices.
8243 (have_whole_vector_shift): Use vec_perm_builder and
8244 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8245 truncation to calc_vec_perm_mask_for_shift.
8246 (vect_create_epilog_for_reduction): Likewise.
8247 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8248 from auto_vec_perm_indices to vec_perm_indices.
8249 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8250 instead of changing individual elements.
8251 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8252 the vector in d.perm.
8253 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8254 from auto_vec_perm_indices to vec_perm_indices.
8255 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8256 instead of changing individual elements.
8257 (arm_vectorize_vec_perm_const): Use new_vector to install
8258 the vector in d.perm.
8259 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8260 Update vec_perm_builder constructor.
8261 (rs6000_expand_interleave): Likewise.
8262 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8263 (rs6000_expand_interleave): Likewise.
8265 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8267 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8268 to qimode could truncate the indices.
8269 * optabs.c (expand_vec_perm_var): Likewise.
8271 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8273 * Makefile.in (OBJS): Add vec-perm-indices.o.
8274 * vec-perm-indices.h: New file.
8275 * vec-perm-indices.c: Likewise.
8276 * target.h (vec_perm_indices): Replace with a forward class
8278 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8279 * optabs.h: Include vec-perm-indices.h.
8280 (expand_vec_perm): Delete.
8281 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8282 (expand_vec_perm_const): Declare.
8283 * target.def (vec_perm_const_ok): Replace with...
8284 (vec_perm_const): ...this new hook.
8285 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8286 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8287 * doc/tm.texi: Regenerate.
8288 * optabs.def (vec_perm_const): Delete.
8289 * doc/md.texi (vec_perm_const): Likewise.
8290 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8291 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8292 expand_vec_perm for constant permutation vectors. Assert that
8293 the mode of variable permutation vectors is the integer equivalent
8294 of the mode that is being permuted.
8295 * optabs-query.h (selector_fits_mode_p): Declare.
8296 * optabs-query.c: Include vec-perm-indices.h.
8297 (selector_fits_mode_p): New function.
8298 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8299 is defined, instead of checking whether the vec_perm_const_optab
8300 exists. Use targetm.vectorize.vec_perm_const instead of
8301 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8302 fit in the vector mode before using a variable permute.
8303 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8304 vec_perm_indices instead of an rtx.
8305 (expand_vec_perm): Replace with...
8306 (expand_vec_perm_const): ...this new function. Take the selector
8307 as a vec_perm_indices rather than an rtx. Also take the mode of
8308 the selector. Update call to shift_amt_for_vec_perm_mask.
8309 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8310 Use vec_perm_indices::new_expanded_vector to expand the original
8311 selector into bytes. Check whether the indices fit in the vector
8312 mode before using a variable permute.
8313 (expand_vec_perm_var): Make global.
8314 (expand_mult_highpart): Use expand_vec_perm_const.
8315 * fold-const.c: Includes vec-perm-indices.h.
8316 * tree-ssa-forwprop.c: Likewise.
8317 * tree-vect-data-refs.c: Likewise.
8318 * tree-vect-generic.c: Likewise.
8319 * tree-vect-loop.c: Likewise.
8320 * tree-vect-slp.c: Likewise.
8321 * tree-vect-stmts.c: Likewise.
8322 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8324 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8325 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8326 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8327 (aarch64_vectorize_vec_perm_const): ...this new function.
8328 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8329 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8330 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8331 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8332 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8333 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8334 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8336 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8337 check for NEON modes.
8338 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8339 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8340 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8341 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8343 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8344 the old VEC_PERM_CONST conditions.
8345 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8346 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8347 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8348 (ia64_vectorize_vec_perm_const_ok): Merge into...
8349 (ia64_vectorize_vec_perm_const): ...this new function.
8350 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8351 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8352 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8353 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8354 * config/mips/mips.c (mips_expand_vec_perm_const)
8355 (mips_vectorize_vec_perm_const_ok): Merge into...
8356 (mips_vectorize_vec_perm_const): ...this new function.
8357 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8358 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8359 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8360 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8361 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8362 (rs6000_expand_vec_perm_const): Delete.
8363 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8365 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8366 (altivec_expand_vec_perm_const_le): Take each operand individually.
8367 Operate on constant selectors rather than rtxes.
8368 (altivec_expand_vec_perm_const): Likewise. Update call to
8369 altivec_expand_vec_perm_const_le.
8370 (rs6000_expand_vec_perm_const): Delete.
8371 (rs6000_vectorize_vec_perm_const_ok): Delete.
8372 (rs6000_vectorize_vec_perm_const): New function.
8373 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8374 an element count and rtx array.
8375 (rs6000_expand_extract_even): Update call accordingly.
8376 (rs6000_expand_interleave): Likewise.
8377 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8378 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8379 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8380 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8381 (rs6000_expand_vec_perm_const): Delete.
8382 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8383 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8384 (altivec_expand_vec_perm_const_le): Take each operand individually.
8385 Operate on constant selectors rather than rtxes.
8386 (altivec_expand_vec_perm_const): Likewise. Update call to
8387 altivec_expand_vec_perm_const_le.
8388 (rs6000_expand_vec_perm_const): Delete.
8389 (rs6000_vectorize_vec_perm_const_ok): Delete.
8390 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8391 reference to the SPE evmerge intructions.
8392 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8393 an element count and rtx array.
8394 (rs6000_expand_extract_even): Update call accordingly.
8395 (rs6000_expand_interleave): Likewise.
8396 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8397 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8399 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8401 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8403 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8404 vector mode and that that mode matches the mode of the data
8406 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8407 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8408 directly using expand_vec_perm_1 when forcing selectors into
8410 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8412 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8414 * optabs-query.h (can_vec_perm_p): Delete.
8415 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8416 * optabs-query.c (can_vec_perm_p): Split into...
8417 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8418 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8419 particular selector is valid.
8420 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8421 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8422 (vect_grouped_load_supported): Likewise.
8423 (vect_shift_permute_load_chain): Likewise.
8424 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8425 (vect_transform_slp_perm_load): Likewise.
8426 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8427 (vectorizable_bswap): Likewise.
8428 (vect_gen_perm_mask_checked): Likewise.
8429 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8430 implementations of variable permutation vectors into account
8431 when deciding which selector to use.
8432 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8433 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8434 with a false third argument.
8435 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8436 to test whether the constant selector is valid and can_vec_perm_var_p
8437 to test whether a variable selector is valid.
8439 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8441 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8442 * optabs-query.c (can_vec_perm_p): Likewise.
8443 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8444 instead of vec_perm_indices.
8445 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8446 (vect_gen_perm_mask_checked): Likewise,
8447 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8448 (vect_gen_perm_mask_checked): Likewise,
8450 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8452 * optabs-query.h (qimode_for_vec_perm): Declare.
8453 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8454 (qimode_for_vec_perm): ...this new function.
8455 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8457 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8459 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8460 does not have a conditional at the top.
8462 2018-01-02 Richard Biener <rguenther@suse.de>
8464 * ipa-inline.c (big_speedup_p): Fix expression.
8466 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8469 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8472 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8476 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8477 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8478 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8479 cond_taken_branch_cost 3->4.
8481 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8483 PR tree-optimization/83581
8484 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8485 TODO_cleanup_cfg if any changes have been made.
8488 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8489 convert_modes if target mode has the right side, but different mode
8493 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8494 last argument when extracting from CONCAT. If either from_real or
8495 from_imag is NULL, use expansion through memory. If result is not
8496 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8497 the parts directly to inner mode, if even that fails, use expansion
8501 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8502 check for bswap in mode rather than HImode and use that in expand_unop
8505 Copyright (C) 2018 Free Software Foundation, Inc.
8507 Copying and distribution of this file, with or without modification,
8508 are permitted in any medium without royalty provided the copyright
8509 notice and this notice are preserved.