1 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5 decl, use remap_type if we want to use the type.
7 2018-03-09 Martin Sebor <msebor@redhat.com>
9 PR tree-optimization/84526
10 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12 (builtin_access::generic_overlap): Be prepared to handle non-array
15 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
17 PR rtl-optimization/84682
18 * lra-constraints.c (process_address_1): Check is_address flag
19 for address constraints.
20 (process_alt_operands): Likewise.
21 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
22 preprocess_constraints.
23 * recog.h (preprocess_constraints): Add oploc parameter.
25 * recog.c (preprocess_constraints): Test address_operand for
26 CT_ADDRESS constraints.
28 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
31 * lra-assigns.c (assign_by_spills): Return a flag of reload
32 assignment failure. Do not process the reload assignment
33 failures. Do not spill other reload pseudos if they has the same
35 (lra_assign): Add a return arg. Set up from the result of
36 assign_by_spills call.
37 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
38 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
39 usage_insns if it is not NULL.
40 (spill_hard_reg_in_range): New function.
41 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
42 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
44 (lra_assign): Change prototype.
45 * lra.c (lra): Add code to deal with fails by splitting hard reg
48 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
51 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
52 Accept complain bool parameter. Only emit errors if it is true.
53 (arm_parse_cpu_option_name): Likewise.
54 (arm_target_thumb_only): Adjust callers of the above.
55 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
56 prototype to take a default true bool parameter.
57 (arm_parse_arch_option_name): Likewise.
59 2018-03-09 David Malcolm <dmalcolm@redhat.com>
60 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
64 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
65 * configure: Regenerate.
66 * configure.ac ("linker --version-script option"): New.
67 ("linker soname option"): New.
69 2018-03-09 Richard Biener <rguenther@suse.de>
71 PR tree-optimization/84775
72 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
73 immediate uses of predicate stmts and mark them modified.
76 PR tree-optimization/84178
77 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
79 (version_loop_for_if_conversion): Delay update_ssa call.
80 (tree_if_conversion): Delay update_ssa until after predicate
83 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
86 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
87 when the function accesses prior frames.
89 2018-03-08 Jakub Jelinek <jakub@redhat.com>
92 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
93 gen_llsym, otherwise call maybe_gen_llsym.
96 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
97 has ',' character inside of it.
99 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
102 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
103 as clobbering CC_REGNUM.
105 2018-03-08 Richard Biener <rguenther@suse.de>
108 * tree-scalar-evolution.c: Include tree-into-ssa.h.
109 (follow_copies_to_constant): Do not follow SSA names registered
112 2018-03-08 Richard Biener <rguenther@suse.de>
114 PR tree-optimization/84178
115 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
117 (version_loop_for_if_conversion): Delay update_ssa call.
118 (tree_if_conversion): Delay update_ssa until after predicate
121 2018-03-08 David Malcolm <dmalcolm@redhat.com>
123 PR tree-optimization/84178
124 * tree-if-conv.c (release_bb_predicate): Remove the
125 the assertion that the stmts have NULL use_ops.
126 Discard the statements, asserting that they haven't
127 yet been added to a BB.
129 2018-03-08 Richard Biener <rguenther@suse.de>
131 PR tree-optimization/84746
132 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
133 (phi_translate): Pass in destination ANTIC_OUT set.
134 (phi_translate_1): Likewise. For a simplified result lookup
135 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
136 (phi_translate_set): Adjust.
137 (do_pre_regular_insertion): Likewise.
138 (do_pre_partial_partial_insertion): Likewise.
140 2018-03-08 Martin Liska <mliska@suse.cz>
142 PR gcov-profile/84735
143 * doc/gcov.texi: Document usage of profile files.
144 * gcov-io.h: Document changes in the format.
146 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
150 * dwarf2out.c (struct dw_line_info_table): Update comments for
152 (FORCE_RESET_NEXT_VIEW): New.
153 (FORCE_RESETTING_VIEW_P): New.
154 (RESETTING_VIEW_P): Check for -1 too.
155 (ZERO_VIEW_P): Likewise.
156 (new_line_info_table): Force-reset next view.
157 (dwarf2out_begin_function): Likewise.
158 (dwarf2out_source_line): Simplify zero_view_p initialization.
159 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
160 view directly. Omit view when omitting .loc at line 0.
162 2018-03-08 Jakub Jelinek <jakub@redhat.com>
164 PR tree-optimization/84740
165 * tree-switch-conversion.c (process_switch): Call build_constructors
166 only if info.phi_count is non-zero.
168 PR tree-optimization/84739
169 * tree-tailcall.c (find_tail_calls): Check call arguments against
170 DECL_ARGUMENTS (current_function_decl) rather than
171 DECL_ARGUMENTS (func) when checking for tail recursion.
173 2018-03-07 Jakub Jelinek <jakub@redhat.com>
175 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
176 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
177 Volker Reichelt's entry and add entries for people that perform
178 GCC fuzzy testing and report numerous bugs.
180 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
183 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
184 readonly data in sdata, if that is disabled.
185 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
186 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
187 -mreadonly-in-sdata option.
189 2018-03-07 Martin Sebor <msebor@redhat.com>
191 PR tree-optimization/84468
192 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
193 basic block when looking for nul assignment.
195 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
198 * except.h (output_function_exception_table): Adjust prototype.
199 * except.c (output_function_exception_table): Remove FNNAME parameter
200 and add SECTION parameter. Ouput one part of the table at a time.
201 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
202 the first part of the exception table and emit unwind directives.
203 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
204 (i386_pe_seh_cold_init): Likewise.
205 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
206 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
207 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
208 (ix86_output_call_insn): Emit a nop in one more case for SEH.
209 * config/i386/winnt.c: Include except.h.
210 (struct seh_frame_state): Add reg_offset, after_prologue and
211 in_cold_section fields.
212 (i386_pe_seh_end_prologue): Set seh->after_prologue.
213 (i386_pe_seh_cold_init): New function.
214 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
215 to seh->in_cold_section.
216 (seh_emit_push): Record the offset of the push.
217 (seh_emit_save): Record the offet of the save.
218 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
219 Test seh->after_prologue to disregard the epilogue.
220 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
221 (i386_pe_end_cold_function): New function.
223 2018-03-07 Jakub Jelinek <jakub@redhat.com>
226 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
227 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
230 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
232 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
233 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
236 * multiple_target.c: Include tree-inline.h and intl.h.
237 (expand_target_clones): Diagnose and fail if node->definition and
238 !tree_versionable_function_p (node->decl).
240 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
242 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
244 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
245 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
246 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
248 2018-03-06 Jakub Jelinek <jakub@redhat.com>
251 * combine.c (try_combine): Use reg_or_subregno instead of handling
252 just paradoxical SUBREGs and REGs.
254 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
256 * config/arc/arc.c (arc_finalize_pic): Remove function.
257 (arc_must_save_register): We use single base PIC register, remove
258 checks to save/restore the PIC register.
259 (arc_expand_prologue): Likewise.
260 * config/arc/arc-protos.h (arc_set_default_type_attributes):
262 (arc_verify_short): Likewise.
263 (arc_attr_type): Likewise.
264 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
265 (walk_stores): Likewise.
266 (arc_address_cost): Make it static.
267 (arc_verify_short): Likewise.
268 (branch_dest): Likewise.
269 (arc_attr_type): Likewise.
270 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
271 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
272 (arc_final_prescan_insn): Remove inserting the nops due to
273 hardware hazards. It is done in reorg step.
274 (insn_length_variant_t): Remove.
275 (insn_length_parameters_t): Likewise.
276 (arc_insn_length_parameters): Likewise.
277 (arc_get_insn_variants): Likewise.
278 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
280 2018-03-06 Jakub Jelinek <jakub@redhat.com>
283 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
286 PR tree-optimization/84687
287 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
289 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
291 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
293 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
294 Rename to ppc_speculation_barrier.
295 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
296 __builtin_ppc_speculation_barrier.
298 2018-03-05 Jakub Jelinek <jakub@redhat.com>
301 * combine.c (combine_simplify_rtx): Don't try to simplify if
302 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
305 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
307 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
308 to 32 bytes when compiling for POWER9.
310 2018-03-05 Jakub Jelinek <jakub@redhat.com>
313 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
314 regparm >= 3 with no arg reg available also for calls with
315 flag_force_indirect_call. Pass decl to ix86_function_regparm.
318 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
320 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
322 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
325 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
327 2018-03-05 Richard Biener <rguenther@suse.de>
329 PR tree-optimization/84486
330 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
331 When inserting a __builtin_assume_aligned call set the LHS
332 SSA name alignment info accordingly.
334 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
336 PR tree-optimization/84114
337 * config/aarch64/aarch64.c (aarch64_reassociation_width)
338 Avoid reassociation of FLOAT_MODE addition.
340 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
342 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
343 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
344 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
345 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
346 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
347 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
348 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
350 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
351 __builtin_ia32_wbinvd): New builtins.
352 (SPECIAL_ARGS2): New.
353 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
354 (SPECIAL_ARGS2): New.
355 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
356 (ix86_valid_target_attribute_inner_p): Ditto.
357 (ix86_init_mmx_sse_builtins): Add special_args2.
358 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
359 TARGET_WBNOINVD_P): New.
360 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
361 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
362 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
363 * config/i386/immintrin.h (_wbinvd): New intrinsic.
364 * config/i386/pconfigintrin.h: New file.
365 * config/i386/wbnoinvdintrin.h: Ditto.
366 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
367 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
369 2018-03-05 Richard Biener <rguenther@suse.de>
371 PR tree-optimization/84670
372 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
374 (BB_VISITED_WITH_VISITED_SUCCS): New define.
375 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
376 (compute_antic_aux): Only assert the number of values in ANTIC_IN
377 doesn't grow if all successors (recursively) were visited at least
380 2018-03-05 Richard Biener <rguenther@suse.de>
382 PR tree-optimization/84650
383 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
384 if executed in the loop pipeline.
386 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
388 * doc/configfiles.texi (Configuration Files): Move info about
389 conditionalizing $target-protos.h to...
390 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
391 differs from $target-protos.h.
393 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
394 Chung-Ju Wu <jasonwucj@gmail.com>
396 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
397 * config/nds32/nds32-multiple.md (setmemsi): Define.
398 * config/nds32/nds32-memory-manipulation.c
399 (nds32_gen_dup_4_byte_to_word_value): New.
400 (emit_setmem_word_loop): New.
401 (emit_setmem_byte_loop): New.
402 (nds32_expand_setmem_loop): New.
403 (nds32_expand_setmem_loop_v3m): New.
404 (nds32_expand_setmem_unroll): New.
405 (nds32_expand_setmem): New.
407 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
408 Chung-Ju Wu <jasonwucj@gmail.com>
410 * config/nds32/nds32-memory-manipulation.c
411 (nds32_emit_load_store): New.
412 (nds32_emit_post_inc_load_store): New.
413 (nds32_emit_mem_move): New.
414 (nds32_emit_mem_move_block): New.
415 (nds32_expand_movmemsi_loop_unknown_size): New.
416 (nds32_expand_movmemsi_loop_known_size): New.
417 (nds32_expand_movmemsi_loop): New.
418 (nds32_expand_movmemsi_unroll): New.
419 (nds32_expand_movmemqi): Rename ...
420 (nds32_expand_movmemsi): ... to this.
421 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
422 (movmemsi): ... to this.
423 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
424 (nds32_expand_movmemsi): ... to this.
426 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
427 Monk Chiang <sh.chiang04@gmail.com>
428 Chung-Ju Wu <jasonwucj@gmail.com>
430 * config/nds32/nds32-protos.h
431 (nds32_expand_load_multiple): New arguments.
432 (nds32_expand_store_multiple): Ditto.
433 (nds32_valid_multiple_load_store): Rename ...
434 (nds32_valid_multiple_load_store_p): ... to this.
435 * config/nds32/nds32-memory-manipulation.c
436 (nds32_expand_load_multiple): Refine implementation.
437 (nds32_expand_store_multiple): Ditto.
438 * config/nds32/nds32-multiple.md
439 (load_multiple): Update nds32_expand_load_multiple interface.
440 (store_multiple): Update nds32_expand_store_multiple interface.
441 * config/nds32/nds32-predicates.c
442 (nds32_valid_multiple_load_store): Rename ...
443 (nds32_valid_multiple_load_store_p): ... to this and refine
445 * config/nds32/predicates.md
446 (nds32_load_multiple_and_update_address_operation): New predicate.
447 (nds32_store_multiple_and_update_address_operation): New predicate.
449 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
450 Chung-Ju Wu <jasonwucj@gmail.com>
452 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
453 (combo): New attribute.
454 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
456 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
458 * config/nds32/nds32.opt: Change -mcmodel= default value.
460 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
461 Monk Chiang <sh.chiang04@gmail.com>
462 Chung-Ju Wu <jasonwucj@gmail.com>
464 * config/nds32/constants.md (unspec_element): New enum.
465 * config/nds32/constraints.md (Umw): New constraint.
466 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
467 * config/nds32/nds32-intrinsic.md: Likewise.
468 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
469 (nds32_valid_smw_lwm_base_p): New.
470 (nds32_output_smw_single_word): New.
471 (nds32_output_lmw_single_word): New.
472 (nds32_expand_unaligned_load): New.
473 (nds32_expand_unaligned_store): New.
474 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
475 (nds32_output_smw_single_word): Declare.
476 (nds32_output_lmw_single_word): Declare.
477 (nds32_expand_unaligned_load): Declare.
478 (nds32_expand_unaligned_store): Declare.
479 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
480 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
481 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
482 NDS32_BUILTIN_UASTORE_DW.
483 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
486 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
487 Kito Cheng <kito.cheng@gmail.com>
488 Chung-Ju Wu <jasonwucj@gmail.com>
490 * config/nds32/nds32-intrinsic.c
491 (nds32_expand_builtin_null_ftype_reg): Delete.
492 (nds32_expand_builtin_reg_ftype_imm): Ditto.
493 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
494 (nds32_read_argument): New.
495 (nds32_legitimize_target): Ditto.
496 (nds32_legitimize_argument): Ditto.
497 (nds32_check_constant_argument): Ditto.
498 (nds32_expand_unop_builtin): Ditto.
499 (nds32_expand_unopimm_builtin): Ditto.
500 (nds32_expand_binop_builtin): Ditto.
501 (nds32_builtin_decl_impl): Ditto.
502 (builtin_description): Ditto.
503 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
504 (nds32_init_builtins_impl): Ditto.
505 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
506 (nds32_builtin_decl): New.
507 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
508 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
510 2018-03-02 Jeff Law <law@redhat.com>
512 * reorg.c (stop_search_p): Handle DEBUG_INSN.
513 (redundant_insn, fill_simple_delay_slots): Likewise.
514 (fill_slots_from_thread): Likewise.
515 * resource.c (mark_referenced_resources): Likewise.
516 (mark_set_resources, find_dead_or_set_registers): Likewise.
518 2018-03-02 Jakub Jelinek <jakub@redhat.com>
520 * substring-locations.h (format_warning_va): Formatting fix for
522 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
524 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
525 * substring-locations.c: Include intl.h.
526 (format_warning_va): Turned into small wrapper around
527 format_warning_n_va, renamed to ...
528 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
529 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
531 (format_warning_at_substring_n): New function.
532 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
533 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
534 format_warning_at_substring with just a shorter name instead of
535 const function pointer.
536 (fmtwarn_n): New function.
537 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
538 appropriate, get rid of all the fmtstr temporaries, move conditionals
539 with G_() wrapped string literals directly into fmtwarn arguments,
540 cast dir.len to (int), formatting fixes.
542 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
544 * doc/invoke.texi: Remove "Cilk Plus" references.
546 2018-03-02 Jakub Jelinek <jakub@redhat.com>
547 Richard Biener <rguenther@suse.de>
550 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
551 for error or warning attributes if CALL_FROM_THUNK_P is set.
554 2018-03-02 Jakub Jelinek <jakub@redhat.com>
557 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
558 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
561 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
562 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
564 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
565 instead of -1U in last predictors element's probability member.
567 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
570 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
571 arguments if they are comparable.
573 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
575 PR tree-optimization/84634
576 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
577 masks and masked_loop_p with a single loop_masks, making sure it's
578 null for bb vectorization.
580 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
582 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
583 (vect_analyze_data_ref_access): Use loop->safe_len rather than
584 loop->force_vectorize to check whether there is no alias.
586 2018-03-02 Jakub Jelinek <jakub@redhat.com>
589 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
591 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
593 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
594 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
595 instead of a loop around prev_real_insn.
596 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
600 * config/i386/i386.c (ix86_print_operand): Use conditional
601 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
604 2018-03-02 Richard Biener <rguenther@suse.de>
606 PR tree-optimization/84427
607 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
608 (bitmap_set_subtract_values): Rewrite to handle multiple
611 (prune_clobbered_mems): Likewise.
612 (phi_translate): Take edge instead of pred/phiblock.
613 (phi_translate_1): Likewise.
614 (phi_translate_set): Likewise. Insert all translated
615 exprs for a value into the set, keeping possibly multiple
616 expressions per value.
617 (compute_antic_aux): Adjust for phi_translate changes.
618 When intersecting union the expressions and prune those
619 not in the final value set, keeping possibly multiple
620 expressions per value. Do not use value-insertion
621 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
622 all expressions. Add verification that the value-sets
623 only shrink during iteration.
624 (compute_partial_antic_aux): Adjust for the phi_translate changes.
625 (do_pre_regular_insertion): Likewise.
626 (do_pre_partial_partial_insertion): Likewise.
628 2018-03-02 Richard Biener <rguenther@suse.de>
631 * config/darwin.c (saved_debug_info_level): New static global.
632 (darwin_asm_lto_start): Disable debug info generation for LTO out.
633 (darwin_asm_lto_end): Restore debug info generation settings.
635 2018-03-01 Martin Liska <mliska@suse.cz>
638 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
641 2018-03-01 Richard Biener <rguenther@suse.de>
644 * dwarf2out.c (gen_variable_die): Properly handle late VLA
645 type annotation with LTO when debug was disabled at compile-time.
647 2018-03-01 Matthew Fortune <mfortune@gmail.com>
649 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
651 (mips_final_postscan_insn): Likewise.
653 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
655 PR rtl-optimization/84528
656 * alias.c (init_alias_target): Add commentary.
657 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
658 a unique base value if the frame pointer is not eliminated
659 to the stack pointer.
661 2018-03-01 Tom de Vries <tom@codesourcery.com>
663 PR rtl-optimization/83327
664 * lra-int.h (hard_regs_spilled_into): Declare.
665 * lra.c (hard_regs_spilled_into): Define.
666 (init_reg_info): Init hard_regs_spilled_into.
667 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
668 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
669 (process_bb_lives): Handle hard_regs_spilled_into.
670 (lra_create_live_ranges_1): Before doing liveness propagation, clear
671 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
673 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
675 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
676 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
677 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
678 * config/rs6000/aix72.h: New file.
680 2018-02-28 Jakub Jelinek <jakub@redhat.com>
682 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
683 instead of warning_at with conditional singular and plural messages
687 * stor-layout.c (update_alignment_for_field): For
688 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
689 && !DECL_PACKED (field), do the alignment update, just use
690 only desired_align instead of MAX (type_align, desired_align)
692 (place_field): Don't do known_align < desired_align handling
693 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
694 is non-NULL, instead do it after rli->prev_field handling and
695 only if not within a bitfield word. For DECL_PACKED (field)
696 use type_align of BITS_PER_UNIT.
698 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
700 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
701 superfluous parentheses and trailing spaces.
703 2018-02-28 Richard Biener <rguenther@suse.de>
705 PR tree-optimization/84584
706 * graphite-scop-detection.c (scop_detection::add_scop): Discard
707 SCoPs with fake exit edge.
709 2018-02-28 Martin Liska <mliska@suse.cz>
712 * timevar.c (timer::print): Fix format to properly print 100%
715 2018-02-28 Richard Biener <rguenther@suse.de>
718 * genmatch.c (capture_info::walk_match): Do not mark
719 captured expressions without operands as expr_p given
720 they act more like predicates and should be subject to
721 "lost tail" side-effect preserving.
723 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
725 PR rtl-optimization/81611
726 * auto-inc-dec.c (attempt_change): Move dead note from
727 mem_insn if it's the next use of regno
728 (find_address): Take address use of reg holding
729 non-incremented value. Add parm to limit search to the named
731 (merge_in_block): Attempt to use a mem insn that is the next
732 use of the original regno.
734 2018-02-27 Martin Sebor <msebor@redhat.com>
737 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
738 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
740 2018-02-27 Martin Sebor <msebor@redhat.com>
743 * diagnostic-core.h (warning_n, error_n, inform_n): Change
744 n argument to unsigned HOST_WIDE_INT.
745 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
746 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
747 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
748 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
750 2018-02-27 Richard Biener <rguenther@suse.de>
752 PR tree-optimization/84512
753 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
754 Do not use the estimate returned from record_stmt_cost for
755 the scalar iteration cost but sum properly using add_stmt_cost.
757 2018-02-27 Richard Biener <rguenther@suse.de>
759 PR tree-optimization/84466
760 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
761 Adjust last change to less strictly validate use operands.
763 2018-02-27 Martin Liska <mliska@suse.cz>
765 PR gcov-profile/84548
766 * gcov.c (process_file): Allow partial overlap and consider it
767 also as group functions.
768 (output_lines): Properly calculate range of lines for a group.
770 2018-02-27 Martin Liska <mliska@suse.cz>
772 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
773 'ggc' suffixes. Change first column width.
774 (timer::print): Fix formatting of the column.
776 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
778 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
779 preserve inline entry blocks for the sake of debug inline
780 entry point markers alone.
781 (remove_unused_locals): Suggest in comments a better place to
782 force the preservation of inline entry blocks that are
783 otherwise unused, but do not preserve them.
785 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
787 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
789 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
792 * config/i386/constraints.md (Bs): Replace
793 ix86_indirect_branch_register with
794 TARGET_INDIRECT_BRANCH_REGISTER.
796 * config/i386/i386.md (indirect_jump): Likewise.
797 (tablejump): Likewise.
798 (*sibcall_memory): Likewise.
799 (*sibcall_value_memory): Likewise.
800 Peepholes of indirect call and jump via memory: Likewise.
801 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
802 (*sibcall_value_GOT_32): Likewise.
803 * config/i386/predicates.md (indirect_branch_operand): Likewise.
804 (GOT_memory_operand): Likewise.
805 (call_insn_operand): Likewise.
806 (sibcall_insn_operand): Likewise.
807 (GOT32_symbol_operand): Likewise.
808 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
810 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
812 PR rtl-optimization/83496
813 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
814 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
815 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
816 redundant insn, if any.
817 (relax_delay_slots): Likewise.
818 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
820 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
822 PR tree-optimization/83965
823 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
824 that grouped statements are part of a reduction chain. Return
825 true if the statement is not marked as a reduction itself but
827 (vect_recog_dot_prod_pattern): Don't check whether the statement
828 is part of a group here.
829 (vect_recog_sad_pattern): Likewise.
830 (vect_recog_widen_sum_pattern): Likewise.
832 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
835 * final.c (rest_of_clean_state): Also look for calls inside sequences.
837 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
840 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
842 (ix86_output_indirect_function_return): New prototype.
843 (ix86_split_simple_return_pop_internal): Likewise.
844 * config/i386/i386.c (indirect_return_via_cx): New.
845 (indirect_return_via_cx_bnd): Likewise.
846 (indirect_thunk_name): Handle return va CX_REG.
847 (output_indirect_thunk_function): Create alias for
848 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
849 (ix86_output_indirect_jmp): Remove the bool argument.
850 (ix86_output_indirect_function_return): New function.
851 (ix86_split_simple_return_pop_internal): Likewise.
852 * config/i386/i386.md (*indirect_jump): Don't pass false
853 to ix86_output_indirect_jmp.
854 (*tablejump_1): Likewise.
855 (simple_return_pop_internal): Change it to define_insn_and_split.
856 Call ix86_split_simple_return_pop_internal to split it for
858 (simple_return_indirect_internal): Call
859 ix86_output_indirect_function_return instead of
860 ix86_output_indirect_jmp.
862 2018-02-26 Jakub Jelinek <jakub@redhat.com>
865 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
866 memset and value initialization afterwards.
868 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
870 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
872 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
875 * common/config/aarch64/aarch64-common.c
876 (aarch_option_optimization_table[]): Switch
877 off fomit-frame-pointer
879 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
880 Chung-Ju Wu <jasonwucj@gmail.com>
882 * config/nds32/nds32-multiple.md (load_multiple): Disallow
884 (store_multiple): Ditto.
886 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
888 * config.gcc: Add --with-cpu support for nds32 target.
889 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
890 * config/nds32/nds32.opt: Add -mcpu= option.
892 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
894 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
895 isel=yes): Warn for these deprecated options.
897 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
899 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
900 ISA_2_5_MASKS_EMBEDDED.
902 2018-02-23 Jakub Jelinek <jakub@redhat.com>
904 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
905 p->max as pointers rather than using iterative_hash_expr.
907 2018-02-23 Carl Love <cel@us.ibm.com>
909 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
910 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
912 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
913 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
914 P8V_BUILTIN_VEC_VUNSIGNED2.
916 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
919 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
920 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
922 * lra-constraints.c (curr_insn_transform): Set up
923 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
925 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
926 * lra-eliminations.c (spill_pseudos): Ditto.
927 (process_insn_for_elimination): Ditto.
928 * lra-lives.c (reg_early_clobber_p): Use the new macros.
929 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
930 LRA_NON_CLOBBERED_ALT.
932 2018-02-22 Martin Sebor <msebor@redhat.com>
934 PR tree-optimization/84480
935 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
936 to maybe_diag_stxncpy_trunc. Call it.
937 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
938 from gimple_fold_builtin_strcpy. Print inlining stack.
939 (handle_builtin_stxncpy): Print inlining stack.
940 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
942 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
945 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
946 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
947 and -fcheck-pointer-bounds are used together.
948 (indirect_thunk_prefix): New enum.
949 (indirect_thunk_need_prefix): New function.
950 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
951 "_nt" instead of "_bnd" for NOTRACK prefix.
952 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
953 (output_indirect_thunk_function): Likewise.
955 (ix86_code_end): Update output_indirect_thunk_function calls.
956 (ix86_output_indirect_branch_via_reg): Replace
957 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
958 (ix86_output_indirect_branch_via_push): Likewise.
959 (ix86_output_function_return): Likewise.
960 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
961 incompatible with -fcf-protection=branch and
962 -fcheck-pointer-bounds.
964 2018-02-22 Steve Ellcey <sellcey@cavium.com>
967 * config/aarch64/aarch64.c (aarch64_print_address_internal):
968 Change gcc_assert call to output_operand_lossage.
970 2018-02-22 Steve Ellcey <sellcey@cavium.com>
972 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
974 2018-02-22 DJ Delorie <dj@redhat.com>
975 Sebastian Perta <sebastian.perta@renesas.com>
976 Oleg Endo <olegendo@gcc.gnu.org>
978 * config/rx/rx.c (rx_rtx_costs): New function.
979 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
981 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
983 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
985 2018-02-22 Martin Liska <mliska@suse.cz>
988 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
989 Add "native" as a possible value.
991 2018-02-22 Martin Liska <mliska@suse.cz>
994 * config/i386/i386.c (ix86_option_override_internal):
995 Add "native" as a possible value for -march and -mtune.
997 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1000 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1001 to all type variants.
1003 PR tree-optimization/84503
1004 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1005 width as info->bitpos + info->bitsize - start.
1006 (merged_store_group::merge_overlapping): Simplify width computation.
1007 (check_no_overlap): New function.
1008 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1009 start + width and last_order of the group, fail if check_no_overlap
1011 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1012 to group if check_no_overlap fails.
1014 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1016 * config/rs6000/altivec.md: Delete contraint arguments to
1017 define_expand, define_split, and define_peephole2, and in
1018 define_insn_and_split if always unused.
1019 * config/rs6000/darwin.md: Ditto.
1020 * config/rs6000/dfp.md: Ditto.
1021 * config/rs6000/rs6000.md: Ditto.
1022 * config/rs6000/sync.md: Ditto.
1023 * config/rs6000/vector.md: Ditto.
1024 * config/rs6000/vsx.md: Ditto.
1026 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1028 * config/rs6000/altivec.md: Write output control strings as braced
1029 blocks instead of double-quoted strings.
1030 * config/rs6000/darwin.md: Ditto.
1031 * config/rs6000/rs6000.md: Ditto.
1032 * config/rs6000/vector.md: Ditto.
1033 * config/rs6000/vsx.md: Ditto.
1035 2018-02-21 Jason Merrill <jason@redhat.com>
1037 PR c++/84314 - ICE with templates and fastcall attribute.
1038 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1040 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1042 * ipa-cp.c (determine_versionability): Fix comment typos.
1044 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1047 * ipa-cp.c (determine_versionability): Do not version functions caling
1050 2018-02-21 Martin Liska <mliska@suse.cz>
1053 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1054 Add "native" as a possible value.
1055 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1056 the macro when native cpu detection is available.
1058 2018-02-21 Martin Liska <mliska@suse.cz>
1061 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1062 Add "native" as a possible value.
1063 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1064 when native cpu detection is available.
1066 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1067 Martin Sebor <msebor@redhat.com>
1069 PR tree-optimization/84478
1070 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1072 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1073 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1074 support which is conservatively correct, for 2 only stay conservative
1075 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1076 argument to the 2 argument get_range_strlen, adjust 6 arg
1077 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1079 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1080 (gimple_fold_builtin_strlen): Pass true as last argument to
1083 2018-02-20 Martin Sebor <msebor@redhat.com>
1086 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1087 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1088 (builtin_memref::builtin_memref): Factor out parts into
1089 set_base_and_offset and call it.
1091 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1094 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1095 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1096 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1097 search at the associated MODE_INT.
1099 2018-02-20 Jeff Law <law@redhat.com>
1102 PR tree-optimization/81592
1104 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1105 for range data rather than using global data.
1106 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1107 range data rather than using global data.
1108 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1109 pass it to children as needed.
1110 (struct directive::fmtresult): Similarly.
1111 (struct directive::set_width): Similarly.
1112 (struct directive::set_precision): Similarly.
1113 (format_integer, format_directive, parse_directive): Similarly.
1114 (format_none): Accept unnamed vr_values parameter.
1115 (format_percent, format_floating, format_character): Similarly.
1116 (format_string, format_plain): Similarly.
1117 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1118 the EVRP range analyzer for range data rather than using global data.
1119 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1120 gimple-ssa-evrp-analyze.h
1121 (class sprintf_dom_walker): Add after_dom_children member function.
1122 Add evrp_range_analyzer member.
1123 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1124 range analyzer as needed.
1125 (sprintf_dom_walker::after_dom_children): New member function.
1126 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1128 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1129 (evrp_range_analyzer::pop_to_marker): Likewise.
1131 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1133 PR tree-optimization/84419
1134 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1135 with the required type if its current type is compatible but
1138 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1141 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1142 after vectorization.
1144 2018-02-20 Martin Liska <mliska@suse.cz>
1147 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1148 possible values if we don't have a hint.
1150 2018-02-20 Martin Liska <mliska@suse.cz>
1154 * final.c (shorten_branches): Build align_tab array with one
1156 * opts.c (finish_options): Add alignment option limit check.
1157 (MAX_CODE_ALIGN): Likewise.
1158 (MAX_CODE_ALIGN_VALUE): Likewise.
1159 * doc/invoke.texi: Document maximum allowed option value for
1160 all -falign-* options.
1162 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1165 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1166 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1167 * var-tracking.c (emit_note_insn_var_location): Remove all references
1168 to NOTE_INSN_CALL_ARG_LOCATION.
1169 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1170 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1171 Use copy_rtx_if_shared.
1172 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1173 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1174 (dwarf2out_var_location): Remove handling of
1175 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1177 * final.c (final_scan_insn): Remove all references to
1178 NOTE_INSN_CALL_ARG_LOCATION.
1179 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1180 before dumping final insns.
1181 * except.c (emit_note_eh_region_end): Remove all references to
1182 NOTE_INSN_CALL_ARG_LOCATION.
1183 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1184 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1185 * config/arc/arc.c (hwloop_optimize): Likewise.
1186 * config/arm/arm.c (create_fix_barrier): Likewise.
1187 * config/s390/s390.c (s390_chunkify_start): Likewise.
1188 * config/sh/sh.c (find_barrier): Likewise.
1189 * config/i386/i386.c (rest_of_insert_endbranch,
1190 ix86_seh_fixup_eh_fallthru): Likewise.
1191 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1192 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1193 * config/frv/frv.c (frv_function_prologue): Likewise.
1194 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1196 (note_outside_basic_block_p): Remove all references to
1197 NOTE_INSN_CALL_ARG_LOCATION.
1198 * gengtype.c (adjust_field_rtx_def): Likewise.
1199 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1201 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1202 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1205 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1208 PR tree-optimization/84452
1209 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1210 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1213 2018-02-19 Martin Liska <mliska@suse.cz>
1216 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1218 2018-02-19 Martin Liska <mliska@suse.cz>
1219 Richard Sandiford <richard.sandiford@linaro.org>
1221 PR tree-optimization/82491
1222 * gimple-fold.c (get_base_constructor): Make earlier bail out
1225 2018-02-19 Carl Love <cel@us.ibm.com>
1227 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1228 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1230 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1231 P8V_BUILTIN_VEC_NEG.
1233 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1235 * config/rl78/rl78.md (movdf): New define expand.
1237 2018-02-19 Martin Liska <mliska@suse.cz>
1240 * doc/invoke.texi: Fix typo.
1241 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1243 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1245 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1246 handle rs6000_single_float and rs6000_double_float specially for
1249 2018-02-16 Jeff Law <law@redhat.com>
1251 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1254 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1257 * ipa-inline.c (inline_small_functions): Fix a typo.
1259 2018-02-16 Nathan Sidwell <nathan@acm.org>
1261 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1263 2018-02-16 Carl Love <cel@us.ibm.com>
1265 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1266 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1267 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1268 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1269 expansion to P8V_BUILTIN_VEC_FLOAT2.
1271 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1273 PR rtl-optimization/70023
1274 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1275 src_regno into account.
1277 2018-02-16 Carl Love <cel@us.ibm.com>
1279 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1280 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1281 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1282 * config/rs6000/rs6000.c: Remove case statements for
1283 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1284 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1285 and P9V_BUILTIN_VEC_VINSERT4B.
1286 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1287 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1288 * config/rs6000/vsx.md:
1289 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1292 2018-02-16 Carl Love <cel@us.ibm.com>
1294 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1296 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1298 * config/rs6000/rs6000-c.c: Add the definitions for
1299 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1300 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1301 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1302 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1303 definition for insert4b and define insn *insert3b_internal.
1304 * doc/extend.texi: Add documentation for vec_extract4b.
1306 2018-02-16 Nathan Sidwell <nathan@acm.org>
1308 * doc/extend.texi (Backwards Compatibility): Mention friend
1309 injection. Note for-scope is deprecated.
1310 * doc/invoke.texi (-ffriend-injection): Deprecate.
1312 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1314 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1315 that moved to I2, also allow destinations that are a paradoxical
1316 subreg (instead of a normal reg).
1318 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1321 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1324 2018-02-16 Richard Biener <rguenther@suse.de>
1326 PR tree-optimization/84037
1327 PR tree-optimization/84016
1329 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1330 Adjust vec_construct for the fact we need additional higher latency
1331 128bit inserts for AVX256 and AVX512 vector builds.
1332 (ix86_add_stmt_cost): Scale vector construction cost for
1335 2018-02-16 Richard Biener <rguenther@suse.de>
1337 PR tree-optimization/84417
1338 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1339 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1340 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1342 2018-02-16 Martin Liska <mliska@suse.cz>
1345 * internal-fn.def (ASAN_CHECK): Set proper flags.
1346 (ASAN_MARK): Likewise.
1348 2018-02-16 Julia Koval <julia.koval@intel.com>
1350 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1351 from PTA_CANNONLAKE.
1353 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1356 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1357 Use ++iter rather than iter++ for std::list iterators.
1358 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1359 defer deleting them until all nodes in the forest are processed. Do
1360 free even leaf nodes. Change to_process into auto_vec.
1363 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1364 * vec.h (vec_default_construct): Use memset instead of placement new
1365 if BROKEN_VALUE_INITIALIZATION is defined.
1366 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1367 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1370 PR rtl-optimization/83723
1371 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1372 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1373 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1375 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1377 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1379 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1381 PR rtl-optimization/81443
1382 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1383 from inner REGs to paradoxical SUBREGs.
1385 2018-02-16 Richard Biener <rguenther@suse.de>
1387 PR tree-optimization/84399
1388 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1389 For operands we can analyze at their definition make sure we can
1390 analyze them at each use as well.
1392 2018-02-16 Richard Biener <rguenther@suse.de>
1394 PR tree-optimization/84190
1395 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1396 volatile accesses if the decl isn't volatile.
1398 2018-02-15 Jason Merrill <jason@redhat.com>
1400 PR c++/84314 - ICE with templates and fastcall attribute.
1401 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1402 TYPE_CANONICAL on an existing type.
1404 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1406 PR tree-optimization/84383
1407 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1408 dstoff nor call operand_equal_p if dstbase is NULL.
1410 PR tree-optimization/84334
1411 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1412 also a CONSTANT_CLASS_P, punt.
1414 2018-02-14 Jim Wilson <jimw@sifive.com>
1416 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1417 first SMALL_OPERAND check. New local min_second_step. Move assert
1418 to where locals are set. Add TARGET_RVC support.
1419 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1421 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1423 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1425 2018-02-14 Martin Sebor <msebor@redhat.com>
1427 PR tree-optimization/83698
1428 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1429 arrays constrain the offset range to their bounds.
1430 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1431 (builtin_access::overlap): Avoid setting the size of overlap if it's
1433 (maybe_diag_overlap): Also consider arrays when deciding what values
1434 of offsets to include in diagnostics.
1436 2018-02-14 Martin Sebor <msebor@redhat.com>
1439 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1440 that correspond to the kind of a declaration.
1442 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1445 * config/pa/pa.md: Load address of PIC label using the linkage table
1446 if the label is nonlocal.
1448 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1450 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1451 warning message if user requests -maltivec=be.
1452 * doc/invoke.texi: Document deprecation of -maltivec=be.
1454 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1457 * config/rs6000/rs6000-c.c: Update definitions for
1458 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1459 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1461 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1464 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1465 add _get_ssp intrinsics. Remove argument from
1466 __builtin_ia32_rdssp[d|q].
1467 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1468 * config/i386/i386-builtin.def: Remove argument from
1469 __builtin_ia32_rdssp[d|q].
1470 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1471 ix86_expand_special_args_builtin for _rdssp[d|q].
1472 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1473 Clear register before usage.
1474 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1475 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1477 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1479 PR tree-optimization/84357
1480 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1481 operand 1 of an ARRAY_REF too.
1483 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1486 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1487 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1489 (set_of_reg): New struct.
1490 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1491 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1492 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1494 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1495 Split into bitclr, bitset, bitinvert patterns if appropriate.
1496 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1497 use rx_fuse_in_memory_bitop.
1498 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1499 to named insn, correct maximum insn length.
1501 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1504 * machmode.def: Define a complex mode for PARTIAL_INT.
1505 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1507 * doc/rtl.texi: Document CSPImode.
1508 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1510 (msp430_hard_regno_nregs_with_padding): Likewise.
1512 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1515 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1517 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1519 PR rtl-optimization/84169
1520 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1521 we generated a parallel as new i3 and we split that to new i2 and i3
1522 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1523 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1524 those to i2, not i1. Partially rewrite this scan code.
1526 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1529 * stor-layout.c (place_field): For variable length fields, adjust
1530 offset_align afterwards not just based on the field's alignment,
1531 but also on the size.
1534 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1535 of exps and logs in the use_exp2 case.
1537 2018-02-13 Jeff Law <law@redhat.com>
1539 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1542 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1545 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1549 * common.opt (gas-loc-support, gas-locview-support): New.
1550 (ginline-points, ginternal-reset-location-views): New.
1551 * doc/invoke.texi: Document them. Use @itemx where intended.
1552 (gvariable-location-views): Adjust.
1553 * target.def (reset_location_view): New.
1554 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1555 (TARGET_RESET_LOCATION_VIEW): New.
1556 * doc/tm.texi: Rebuilt.
1557 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1558 (dwarf2out_default_as_locview_support): New.
1559 (output_asm_line_debug_info): Use option variables.
1560 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1561 (output_loc_list): Likewise.
1562 (add_high_low_attributes): Check option variables.
1563 Don't output entry view attribute in strict mode.
1564 (gen_inlined_subroutine_die): Check option variables.
1565 (dwarf2out_inline_entry): Likewise.
1566 (init_sections_and_labels): Likewise.
1567 (dwarf2out_early_finish): Likewise.
1568 (maybe_reset_location_view): New, from...
1569 (dwarf2out_var_location): ... here. Call it.
1570 * debug.h (dwarf2out_default_as_loc_support): Declare.
1571 (dwarf2out_default_as_locview_support): Declare.
1572 * hooks.c (hook_int_rtx_insn_0): New.
1573 * hooks.h (hook_int_rtx_insn_0): Declare.
1574 * toplev.c (process_options): Take -gas-loc-support and
1575 -gas-locview-support from dwarf2out. Enable
1576 -gvariable-location-views by default only with locview
1577 assembler support. Enable -ginternal-reset-location-views by
1578 default only if the target defines the corresponding hook.
1579 Enable -ginline-points by default if location views are
1580 enabled; force it disabled if statement frontiers are
1582 * tree-inline.c (expand_call_inline): Check option variables.
1583 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1585 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1587 PR tree-optimization/84321
1588 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1589 handling. Also check whether the anti-range contains any values
1590 that satisfy the mask; switch to a VR_RANGE if not.
1592 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1595 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1597 2018-02-13 Martin Jambor <mjambor@suse.cz>
1600 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1601 of call statements, also set location of a load to a temporary.
1603 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1605 * config/rl78/rl78.c (add_vector_labels): New function.
1606 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1607 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1608 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1609 which checks that no arguments are passed.
1610 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1611 * doc/extend.texi: Documentation for the new attribute.
1613 2018-02-13 Andreas Schwab <schwab@suse.de>
1615 * config/riscv/linux.h (CPP_SPEC): Define.
1617 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1620 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1621 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1622 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1623 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1624 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1625 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1626 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1627 temporarily for AES and PCLMUL builtins.
1629 PR tree-optimization/84339
1630 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1631 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1635 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1636 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1637 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1639 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1641 * omp-simd-clone.h: New file.
1642 * omp-simd-clone.c: Include omp-simd-clone.h.
1643 (expand_simd_clones): No longer static.
1644 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1645 cgraph.h and omp-simd-clone.h.
1646 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1647 (vect_recog_widen_shift_pattern): Formatting fix.
1648 (vect_pattern_recog_1): Don't check optab for calls.
1651 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1652 operands[2] into a REG before using gen_lowpart on it.
1654 2018-02-12 Jeff Law <law@redhat.com>
1657 * config/sh/sh.c (find_barrier): Consider a sibling call
1660 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1661 successfully back substituting a reg.
1663 2018-02-12 Richard Biener <rguenther@suse.de>
1665 PR tree-optimization/84037
1666 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1667 parameter, move visited init to caller.
1668 (vect_slp_analyze_operations): Separate cost from validity
1669 check, initialize visited once for all instances.
1670 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1672 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1674 (vect_model_promotion_demotion_cost): Likewise.
1675 (vectorizable_bswap): Guard cost modeling with !slp_node
1676 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1678 (vectorizable_call): Likewise.
1679 (vectorizable_conversion): Likewise.
1680 (vectorizable_assignment): Likewise.
1681 (vectorizable_shift): Likewise.
1682 (vectorizable_operation): Likewise.
1683 (vectorizable_store): Likewise.
1684 (vectorizable_load): Likewise.
1685 (vectorizable_condition): Likewise.
1686 (vectorizable_comparison): Likewise.
1688 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1691 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1692 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1693 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1695 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1698 * params.def (inline-min-speedup): Increase from 8 to 15.
1699 (max-inline-insns-auto): Decrease from 40 to 30.
1700 * ipa-split.c (consider_split): Add some buffer for function to
1701 be considered inlining candidate.
1702 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1705 2018-02-12 Richard Biener <rguenther@suse.de>
1707 PR tree-optimization/84037
1708 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1709 matched stmts if we cannot swap the non-matched ones.
1711 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1713 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1714 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1715 _mm_maskz_scalef_round_ss): New intrinsics.
1716 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1717 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1718 __builtin_ia32_scalefss_round): Remove.
1719 (__builtin_ia32_scalefsd_mask_round,
1720 __builtin_ia32_scalefss_mask_round): New intrinsics.
1721 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1722 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1723 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1724 "<round_constraint>")): Changed to ...
1725 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1726 "<round_scalar_constraint>")): ... this.
1727 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1728 %0, %1, %2<round_op3>}"): Changed to ...
1729 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1730 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1731 %2<round_scalar_mask_op3>}"): ... this.
1732 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1734 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1736 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1737 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1738 (_mm_maskz_sqrt_round_ss): New intrinsics.
1739 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1740 (__builtin_ia32_sqrtsd_mask_round)
1741 (__builtin_ia32_sqrtss_mask_round): New builtins.
1742 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1743 (__builtin_ia32_sqrtss_round): Remove.
1744 (__builtin_ia32_sqrtsd_mask_round)
1745 (__builtin_ia32_sqrtss_mask_round): New builtins.
1746 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1747 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1748 ((match_operand:VF_128 1 "vector_operand"
1749 "xBm,<round_constraint>")): Changed to ...
1750 ((match_operand:VF_128 1 "vector_operand"
1751 "xBm,<round_scalar_constraint>")): ... this.
1752 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1753 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1754 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1755 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1756 %<iptr>1<round_scalar_mask_op3>}): ... this.
1757 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1758 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1760 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1763 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1764 Cast vec_cmpeq result to correct type.
1765 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1766 Cast vec_cmpgt result to correct type.
1768 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1770 * final.c (final_scan_insn_1): Renamed from...
1771 (final_scan_insn): ... this. New wrapper, to recover
1772 seen from the outermost call in recursive ones.
1773 * config/sparc/sparc.c (output_return): Drop seen from call.
1774 (output_sibcall): Likewise.
1775 * config/visium/visium.c (output_branch): Likewise.
1777 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1779 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1782 2018-02-10 Alan Modra <amodra@gmail.com>
1785 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1786 Specify LR as an input.
1788 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1791 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1792 remove_member_access_dummy_vars): New functions.
1793 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1794 lower_omp_1, execute_lower_omp): Use them.
1796 PR rtl-optimization/84308
1797 * shrink-wrap.c (spread_components): Release todo vector.
1799 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1801 PR rtl-optimization/57193
1802 * ira-color.c (struct allocno_color_data): Add member
1803 conflict_allocno_hard_prefs.
1804 (update_conflict_allocno_hard_prefs): New.
1805 (bucket_allocno_compare_func): Add a preference based on
1806 conflict_allocno_hard_prefs.
1807 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1808 (color_allocnos): Remove a dead code. Initiate
1809 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1811 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1814 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1815 constraint from =wa to wa. Avoid a subreg on the output operand,
1816 instead use a pseudo and subreg it in a move.
1817 (p9_xxbrd_<mode>): Changed to ...
1818 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1819 (p9_xxbrd_v2df): New expander.
1820 (p9_xxbrw_<mode>): Changed to ...
1821 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1822 (p9_xxbrw_v4sf): New expander.
1824 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1826 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1828 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1831 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1832 multiply in 32-bit mode.
1833 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1834 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1837 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1839 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1840 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1841 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1842 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1844 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1847 * dwarf2out.c (is_trivial_indirect_ref): New function.
1848 (dwarf2out_late_global_decl): Do not generate a location
1849 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1850 and that are not defined in the current unit.
1852 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1854 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1855 instead of a libcall for UNORDERED.
1857 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1860 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1861 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1863 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1866 * config/s390/s390.c (s390_set_current_function): Invoke
1867 s390_indirect_branch_settings also if fndecl didn't change.
1869 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1871 * config/rs6000/rs6000.md (blockage): Set length to zero.
1873 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1875 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1877 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1880 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1881 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1885 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1886 PARALLEL incoming that failed vt_get_decl_and_offset check.
1889 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1890 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1892 (get_variable_section): For decls in named .bss* sections pass true as
1893 second argument to bss_initializer_p.
1895 2018-02-09 Marek Polacek <polacek@redhat.com>
1896 Jakub Jelinek <jakub@redhat.com>
1899 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1900 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1901 Sync some changes from cxx_fold_indirect_ref.
1903 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1905 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1907 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1908 (BLOCK_INLINE_ENTRY_LABEL): New.
1909 (dwarf2out_var_location): Disregard inline entry markers.
1910 (inline_entry_data): New struct.
1911 (inline_entry_data_hasher): New hashtable type.
1912 (inline_entry_data_hasher::hash): New.
1913 (inline_entry_data_hasher::equal): New.
1914 (inline_entry_data_table): New variable.
1915 (add_high_low_attributes): Add DW_AT_entry_pc and
1916 DW_AT_GNU_entry_view attributes if a pending entry is found
1917 in inline_entry_data_table. Add old entry_pc attribute only
1918 if debug nonbinding markers are disabled.
1919 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1920 markers are enabled.
1921 (block_within_block_p, dwarf2out_inline_entry): New.
1922 (dwarf2out_finish): Check that no entries remained in
1923 inline_entry_data_table.
1924 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1925 (final_scan_insn, notice_source_line): Likewise.
1926 (rest_of_clean_state): Skip inline entry markers.
1927 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1929 * gimple.c (gimple_build_debug_inline_entry): New.
1930 * gimple.h (enum gimple_debug_subcode): Add
1931 GIMPLE_DEBUG_INLINE_ENTRY.
1932 (gimple_build_debug_inline_entry): Declare.
1933 (gimple_debug_inline_entry_p): New.
1934 (gimple_debug_nonbind_marker_p): Adjust.
1935 * insn-notes.def (INLINE_ENTRY): New.
1936 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1937 inline entry marker notes.
1938 (print_insn): Likewise.
1939 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1940 (INSN_DEBUG_MARKER_KIND): Likewise.
1941 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1942 * tree-inline.c (expand_call_inline): Build and insert
1943 debug_inline_entry stmt.
1944 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1945 inline entry blocks early, if nonbind markers are enabled.
1946 (dump_scope_block): Dump fragment info.
1947 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1948 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1949 (gimple_build_debug_inline_entry): New.
1950 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1951 Enable/disable inline entry points too.
1952 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1953 (DEBUG_INSN): Describe inline entry markers.
1955 * common.opt (gvariable-location-views): New.
1956 (gvariable-location-views=incompat5): New.
1957 * config.in: Rebuilt.
1958 * configure: Rebuilt.
1959 * configure.ac: Test assembler for view support.
1960 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1961 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1962 * dwarf2out.c (var_loc_view): New typedef.
1963 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1964 (dwarf2out_locviews_in_attribute): New.
1965 (dwarf2out_locviews_in_loclist): New.
1966 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1967 (enum dw_line_info_opcode): Add LI_adv_address.
1968 (struct dw_line_info_table): Add view.
1969 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1970 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1971 (zero_view_p): New variable.
1972 (ZERO_VIEW_P): New macro.
1973 (output_asm_line_debug_info): New.
1974 (struct var_loc_node): Add view.
1975 (add_AT_view_list, AT_loc_list): New.
1976 (add_var_loc_to_decl): Add view param. Test it against last.
1977 (new_loc_list): Add view params. Record them.
1978 (AT_loc_list_ptr): Handle loc and view lists.
1979 (view_list_to_loc_list_val_node): New.
1980 (print_dw_val): Handle dw_val_class_view_list.
1981 (size_of_die): Likewise.
1982 (value_format): Likewise.
1983 (loc_list_has_views): New.
1984 (gen_llsym): Set vl_symbol too.
1985 (maybe_gen_llsym, skip_loc_list_entry): New.
1986 (dwarf2out_maybe_output_loclist_view_pair): New.
1987 (output_loc_list): Output view list or entries too.
1988 (output_view_list_offset): New.
1989 (output_die): Handle dw_val_class_view_list.
1990 (output_dwarf_version): New.
1991 (output_compilation_unit_header): Use it.
1992 (output_skeleton_debug_sections): Likewise.
1993 (output_rnglists, output_line_info): Likewise.
1994 (output_pubnames, output_aranges): Update version comments.
1995 (output_one_line_info_table): Output view numbers in asm comments.
1996 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1997 Call maybe_gen_llsym.
1998 (loc_list_from_tree_1): Adjust.
1999 (add_AT_location_description): Create view list attribute if
2000 needed, check it's absent otherwise.
2001 (convert_cfa_to_fb_loc_list): Adjust.
2002 (maybe_emit_file): Call output_asm_line_debug_info for test.
2003 (dwarf2out_var_location): Reset views as needed. Precompute
2004 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2005 attribute. Set view.
2006 (new_line_info_table): Reset next view.
2007 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2008 (dwarf2out_source_line): Likewise. Output view resets and labels to
2009 the assembler, or select appropriate line info opcodes.
2010 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2011 (optimize_string_length): Catch it. Adjust.
2012 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2013 dw_val_class_view_list, and remove it if no longer needed.
2014 (hash_loc_list): Hash view numbers.
2015 (loc_list_hasher::equal): Compare them.
2016 (optimize_location_lists): Check whether a view list symbol is
2017 needed, and whether the locview attribute is present, and
2018 whether they match. Remove the locview attribute if no longer
2020 (index_location_lists): Call skip_loc_list_entry for test.
2021 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2022 Use output_dwarf_version.
2023 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2024 (struct dw_val_node): Add val_view_list.
2025 * final.c (SEEN_NEXT_VIEW): New.
2026 (set_next_view_needed): New.
2027 (clear_next_view_needed): New.
2028 (maybe_output_next_view): New.
2029 (final_start_function): Rename to...
2030 (final_start_function_1): ... this. Take pointer to FIRST,
2031 add SEEN parameter. Emit param bindings in the initial view.
2032 (final_start_function): Reintroduce SEEN-less interface.
2033 (final): Rename to...
2034 (final_1): ... this. Take SEEN parameter. Output final pending
2035 next view at the end.
2036 (final): Reintroduce seen-less interface.
2037 (final_scan_insn): Output pending next view before switching
2038 sections or ending a block. Mark the next view as needed when
2039 outputting variable locations. Notify debug backend of section
2040 changes, and of location view changes.
2041 (rest_of_handle_final): Adjust.
2042 * toplev.c (process_options): Autodetect value for debug variable
2043 location views option. Warn on incompat5 without -gdwarf-5.
2044 * doc/invoke.texi (gvariable-location-views): New.
2045 (gvariable-location-views=incompat5): New.
2046 (gno-variable-location-views): New.
2048 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2050 PR tree-optimization/84136
2051 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2052 that the result of find_edge is non-NULL.
2054 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2057 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2058 storing integer register in SImode. Fix cost of 256 and 512
2059 byte aligned SSE register store.
2061 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2063 * config/i386/i386.c (ix86_multiplication_cost): Fix
2064 multiplication cost for TARGET_AVX512DQ.
2066 2018-02-08 Marek Polacek <polacek@redhat.com>
2068 PR tree-optimization/84238
2069 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2072 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2074 PR tree-optimization/84265
2075 * tree-vect-stmts.c (vectorizable_store): Don't treat
2076 VMAT_CONTIGUOUS accesses as grouped.
2077 (vectorizable_load): Likewise.
2079 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2081 PR tree-optimization/81635
2082 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2083 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2084 (test_round_for_mask): New functions.
2085 (wide_int_cc_tests): Call test_round_for_mask.
2086 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2087 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2088 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2089 range returned by get_range_info.
2091 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2094 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2095 * symtab.c: Include builtins.h
2096 (symtab_node::output_to_lto_symbol_table_p): Move here
2097 from lto-streamer-out.c:output_symbol_p.
2098 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2099 (output_symbol_p): Move all logic to symtab.c
2100 (produce_symtab): Update.
2102 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2104 * config/s390/s390-opts.h (enum indirect_branch): Define.
2105 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2106 (s390_indirect_branch_via_thunk)
2107 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2108 (enum s390_indirect_branch_type): Define.
2109 * config/s390/s390.c (struct s390_frame_layout, struct
2110 machine_function): Remove.
2111 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2112 (indirect_branch_table_label_no, indirect_branch_table_name):
2114 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2115 (enum s390_indirect_branch_option): Define.
2116 (s390_return_addr_from_memory): New function.
2117 (s390_handle_string_attribute): New function.
2118 (s390_attribute_table): Add new attribute handler.
2119 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2120 (s390_indirect_branch_via_thunk): New function.
2121 (s390_indirect_branch_via_inline_thunk): New function.
2122 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2123 sibling call optimization for non z10 compiles.
2124 (s390_emit_call): Force indirect branch target to be a single
2125 register. Add r1 clobber for non-z10 compiles.
2126 (s390_emit_epilogue): Emit return jump via return_use expander.
2127 (s390_reorg): Handle JUMP_INSNs as execute targets.
2128 (s390_option_override_internal): Perform validity checks for the
2129 new command line options.
2130 (s390_indirect_branch_attrvalue): New function.
2131 (s390_indirect_branch_settings): New function.
2132 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2133 (s390_output_indirect_thunk_function): New function.
2134 (s390_code_end): Implement target hook.
2135 (s390_case_values_threshold): Implement target hook.
2136 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2138 * config/s390/s390.h (struct s390_frame_layout)
2139 (struct machine_function): Move here from s390.c.
2140 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2141 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2142 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2143 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2144 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2145 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2146 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2147 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2148 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2149 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2150 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2151 (mnemonic attribute): Add values which aren't recognized
2153 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2154 pattern for branch conversion. Fix mnemonic attribute.
2155 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2156 indirect branch via thunk if requested.
2157 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2158 ("*indirect_jump"): Disable for branch conversion using out of
2160 ("indirect_jump_via_thunk<mode>_z10")
2161 ("indirect_jump_via_thunk<mode>")
2162 ("indirect_jump_via_inlinethunk<mode>_z10")
2163 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2164 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2165 ("casesi_jump_via_inlinethunk<mode>_z10")
2166 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2167 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2168 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2169 ("*indirect2_jump"): Disable for branch conversion.
2170 ("casesi_jump"): Turn into expander and expand patterns for branch
2172 ("return_use"): New expander.
2173 ("*return"): Emit return via thunk and rename it to ...
2174 ("*return<mode>"): ... this one.
2175 * config/s390/s390.opt: Add new options and and enum for the
2178 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2180 * lra-constraints.c (match_reload): Unconditionally use
2181 gen_lowpart_SUBREG, rather than selecting between that
2182 and equivalent gen_rtx_SUBREG code.
2184 2018-02-08 Richard Biener <rguenther@suse.de>
2186 PR tree-optimization/84233
2187 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2188 changed flag instead of boguously re-using phi_inserted.
2190 2018-02-08 Martin Jambor <mjambor@suse.cz>
2192 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2193 static local variables.
2195 2018-02-08 Richard Biener <rguenther@suse.de>
2197 PR tree-optimization/84278
2198 * tree-vect-stmts.c (vectorizable_store): When looking for
2199 smaller vector types to perform grouped strided loads/stores
2200 make sure the mode is supported by the target.
2201 (vectorizable_load): Likewise.
2203 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2205 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2206 Increase LDP/STP opportunities by adding adjacent callee-saves.
2208 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2210 PR rtl-optimization/84068
2211 PR rtl-optimization/83459
2212 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2214 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2216 PR tree-optimization/84224
2217 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2218 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2221 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2224 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2225 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2226 count, remove one USE.
2228 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2230 * doc/install.texi (Configuration): Document the
2231 --with-long-double-format={ibm,ieee} PowerPC configuration
2235 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2236 Convert from define_expand to be define_insn_and_split. Rework
2237 float/double/_Float128 conversions to QI/HI/SImode to work with
2238 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2239 conversions to QI/HImode types did a store and then a load to
2240 truncate the value. For conversions to VSX registers, don't split
2241 the insn, instead emit the code directly. Use the code iterator
2242 any_fix to combine signed and unsigned conversions.
2243 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2244 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2245 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2246 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2247 (fix_<mode>di2_hw): Likewise.
2248 (fixuns_<mode>di2_hw): Likewise.
2249 (fix_<mode>si2_hw): Likewise.
2250 (fixuns_<mode>si2_hw): Likewise.
2251 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2252 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2253 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2254 fix<uns>_trunc<SFDF:mode>si2_p8.
2255 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2257 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2258 (fix<uns>_<mode>_mem): Likewise.
2259 (fctiw<u>z_<mode>_mem): Likewise.
2260 (fix<uns>_<mode>_mem): Likewise.
2261 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2262 the register allocator from doing a direct move to the GPRs to do
2263 a store, and instead use the ISA 3.0 store byte/half-word from
2264 vector register instruction. For IEEE 128-bit floating point,
2265 also optimize stores of 32-bit ints.
2266 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2268 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2270 * genextract.c (push_pathstr_operand): New function to support
2272 (walk_rtx): Call push_pathstr_operand.
2273 (print_path): Support [a-zA-Z].
2275 2018-02-07 Richard Biener <rguenther@suse.de>
2277 PR tree-optimization/84037
2278 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2279 (cse_and_gimplify_to_preheader): Declare.
2280 (vect_get_place_in_interleaving_chain): Likewise.
2281 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2283 (_loop_vec_info::~_loop_vec_info): Delete it.
2284 (cse_and_gimplify_to_preheader): New function.
2285 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2286 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2287 (vectorizable_load): Likewise. For grouped stores always base
2288 the IV on the first element.
2289 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2290 condition before gimplifying.
2292 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2294 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2295 *DIV_EXPR and *MOD_EXPR.
2297 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2300 * config/i386/i386.c (ix86_option_override_internal): Mask out
2301 the CF_SET bit when checking -fcf-protection.
2303 2018-02-07 Tom de Vries <tom@codesourcery.com>
2306 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2309 2018-02-07 Richard Biener <rguenther@suse.de>
2311 PR tree-optimization/84204
2312 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2315 PR tree-optimization/84205
2316 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2317 special-case isl_ast_op_zdiv_r.
2319 PR tree-optimization/84223
2320 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2321 Only add conditions from within the region.
2322 (gather_bbs::after_dom_children): Adjust.
2324 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2327 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2328 * config/avr/avr.md: Only post-reload split REG-REG moves if
2329 either register is GENERAL_REG_P.
2331 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2333 PR tree-optimization/84235
2334 * tree-ssa-scopedtables.c
2335 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2336 if the subtraction is performed in floating point type where NaNs are
2337 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2338 build 1. Formatting fix.
2340 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2343 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2344 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2345 and skip it regardless of bb boundaries. Use CALL_P macro,
2346 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2347 unnecessarily, formatting fix.
2349 2018-02-06 Michael Collison <michael.collison@arm.com>
2351 * config/arm/thumb2.md:
2352 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2353 (*thumb_mov_notscc): Ditto.
2355 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2358 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2359 unsigned_fix, not "s".
2361 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2363 * configure.ac (gcc_fn_eh_frame_ro): New function.
2364 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2365 correct .eh_frame permissions.
2366 * configure: Regenerate.
2368 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2370 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2373 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2375 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2376 Display warning message for -mno-speculate-indirect-jumps.
2378 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2380 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2382 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2384 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2386 PR tree-optimization/84225
2387 * tree-eh.c (find_trapping_overflow): Only call
2388 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2390 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2393 * config/i386/i386.c: Reimplement the check of possible options
2394 -mibt/-mshstk conbination. Change error messages.
2395 * doc/invoke.texi: Fix a typo: remove extra '='.
2397 2018-02-06 Marek Polacek <polacek@redhat.com>
2399 PR tree-optimization/84228
2400 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2402 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2405 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2406 emitted arch directives.
2407 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2408 __ARM_FEATURE_COPROC before changing architectures.
2410 2018-02-06 Richard Biener <rguenther@suse.de>
2412 * config/i386/i386.c (print_reg): Fix typo.
2413 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2415 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2417 * configure: Regenerate.
2419 2018-02-05 Martin Sebor <msebor@redhat.com>
2421 PR tree-optimization/83369
2422 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2425 2018-02-05 Martin Liska <mliska@suse.cz>
2427 * doc/invoke.texi: Cherry-pick upstream r323995.
2429 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2431 * ira.c (ira_init_register_move_cost): Adjust comment.
2433 2018-02-05 Martin Liska <mliska@suse.cz>
2435 PR gcov-profile/84137
2436 * doc/gcov.texi: Fix typo in documentation.
2438 2018-02-05 Martin Liska <mliska@suse.cz>
2440 PR gcov-profile/83879
2441 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2442 using dlopen functionality.
2444 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2446 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2447 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2448 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2449 _mm_maskz_range_round_ss): New intrinsics.
2450 (__builtin_ia32_rangesd128_round)
2451 (__builtin_ia32_rangess128_round): Remove.
2452 (__builtin_ia32_rangesd128_mask_round,
2453 __builtin_ia32_rangess128_mask_round): New builtins.
2454 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2455 __builtin_ia32_rangess128_round): Remove.
2456 (__builtin_ia32_rangesd128_mask_round,
2457 __builtin_ia32_rangess128_mask_round): New builtins.
2458 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2459 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2460 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2461 "<round_saeonly_constraint>")): Changed to ...
2462 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2463 "<round_saeonly_scalar_constraint>")): ... this.
2464 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2465 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2466 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2467 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2468 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2470 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2472 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2474 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2475 Remove all values except native, 8540 and 8548.
2477 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2479 * config/i386/i386.c (ix86_output_function_return): Pass
2480 INVALID_REGNUM, instead of -1, as invalid register number to
2481 indirect_thunk_name and output_indirect_thunk.
2483 2018-02-02 Julia Koval <julia.koval@intel.com>
2485 * config.gcc: Add -march=icelake.
2486 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2487 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2488 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2489 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2490 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2491 (processor_target_table): Add icelake.
2492 (ix86_option_override_internal): Handle new PTAs.
2493 (get_builtin_code_for_version): Handle icelake.
2494 (M_INTEL_COREI7_ICELAKE): New.
2495 (fold_builtin_cpu): Handle icelake.
2496 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2497 * doc/invoke.texi: Add -march=icelake.
2499 2018-02-02 Julia Koval <julia.koval@intel.com>
2501 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2502 to wide_int_bitmask.
2503 * wide-int-bitmask.h: New.
2505 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2508 * config/i386/i386.md: Replace Pmode with word_mode in
2509 builtin_setjmp_setup and builtin_longjmp to support x32.
2511 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2515 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2517 (rs6000_supported_cpu_names): New static variable.
2518 (linux_cpu_translation_table): Likewise.
2519 (elf_platform) <cpu>: Define new static variable and use it.
2520 Translate kernel AT_PLATFORM name to canonical name if needed.
2521 Error if platform name is unknown.
2523 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2526 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2528 2018-02-01 Jeff Law <law@redhat.com>
2531 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2532 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2533 the scratch if RELEASE_VIA_POP is false.
2534 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2535 If we have to save a temporary register, decrement SIZE appropriately.
2536 Pass new arguments to release_scratch_register_on_entry.
2537 (ix86_adjust_stack_and_probe): Likewise.
2538 (ix86_emit_probe_stack_range): Pass new arguments to
2539 release_scratch_register_on_entry.
2541 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2543 PR rtl-optimization/84157
2544 * combine.c (change_zero_ext): Use REG_P predicate in
2545 front of HARD_REGISTER_P predicate.
2547 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2549 * config/avr/avr.c (avr_option_override): Move disabling of
2550 -fdelete-null-pointer-checks to...
2551 * common/config/avr/avr-common.c (avr_option_optimization_table):
2554 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2556 PR tree-optimization/81635
2557 * tree-data-ref.c (split_constant_offset_1): For types that
2558 wrap on overflow, try to use range info to prove that wrapping
2561 2018-02-01 Renlin Li <renlin.li@arm.com>
2564 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2566 (aarch64_register_move_cost): Likewise.
2567 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2569 (REG_CLASS_NAMES): Likewise.
2570 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2571 TAILCALL_ADDR_REGS. Remove IP registers.
2572 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2574 2018-02-01 Richard Biener <rguenther@suse.de>
2576 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2577 for specifying RPO order and allow NULL for that.
2578 * domwalk.c (dom_walker::dom_walker): Likewise.
2579 (dom_walker::walk): Handle NULL RPO order.
2580 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2582 (rewrite_update_dom_walker): Likewise.
2583 (mark_def_dom_walker): Likewise.
2585 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2587 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2588 (aarch64_maybe_expand_sve_subreg_move): Declare.
2589 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2590 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2592 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2593 that are semantically a reverse operation.
2594 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2595 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2596 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2598 (aarch64_can_change_mode_class): For big-endian, forbid changes
2599 between two SVE modes if they have different element sizes.
2601 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2603 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2604 the TImode handling for big-endian targets.
2606 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2608 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2609 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2611 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2612 Remove BSWAP handing for big-endian targets and use the form of
2613 LD1RQ appropariate for the mode.
2615 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2617 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2618 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2621 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2624 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2625 check for operands that need to go through aarch64_sve_reload_be.
2627 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2629 PR tree-optimization/81661
2630 PR tree-optimization/84117
2631 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2632 * tree-eh.c: Include gimplify.h.
2633 (find_trapping_overflow, replace_trapping_overflow,
2634 rewrite_to_non_trapping_overflow): New functions.
2635 * tree-vect-loop.c: Include tree-eh.h.
2636 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2637 * tree-data-ref.c: Include tree-eh.h.
2638 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2640 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2642 PR rtl-optimization/84123
2643 * combine.c (change_zero_ext): Check if hard register satisfies
2644 can_change_dest_mode before calling gen_lowpart_SUBREG.
2646 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2649 * ira.c (ira_init_register_move_cost): Remove assert.
2651 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2653 PR rtl-optimization/84071
2654 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2655 * doc/tm.texi: Regenerate.
2657 2018-01-31 Richard Biener <rguenther@suse.de>
2659 PR tree-optimization/84132
2660 * tree-data-ref.c (analyze_miv_subscript): Properly
2661 check whether evolution_function_is_affine_multivariate_p
2662 before calling gcd_of_steps_may_divide_p.
2664 2018-01-31 Julia Koval <julia.koval@intel.com>
2667 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2668 * config/i386/i386.md (rdpid_rex64) New.
2669 (rdpid): Make 32bit only.
2671 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2674 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2675 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2677 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2680 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2682 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2684 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2686 PR rtl-optimization/84071
2687 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2688 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2690 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2692 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2693 (arc_attribute_table): Add 'aux' attribute.
2694 (arc_in_small_data_p): Consider aux like variables.
2695 (arc_is_aux_reg_p): New function.
2696 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2697 (arc_get_aux_arg): New function.
2698 (prepare_move_operands): Handle aux-register access.
2699 (arc_handle_aux_attribute): New function.
2700 * doc/extend.texi (ARC Variable attributes): Add subsection.
2702 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2704 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2705 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2706 (arc_attribute_table): Add 'uncached' attribute.
2707 (arc_print_operand): Print '.di' flag for uncached memory
2709 (arc_in_small_data_p): Do not consider for small data the uncached
2711 (arc_is_uncached_mem_p): New function.
2712 * config/arc/predicates.md (compact_store_memory_operand): Check
2713 for uncached memory accesses.
2714 (nonvol_nonimm_operand): Likewise.
2715 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2717 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2720 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2721 falign-loops=): Add Optimization flag.
2723 2018-01-30 Jeff Law <law@redhat.com>
2726 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2727 INT_REGISTERS_SAVED. Check it prior to calling
2728 get_scratch_register_on_entry.
2729 (ix86_adjust_stack_and_probe): Similarly.
2730 (ix86_emit_probe_stack_range): Similarly.
2731 (ix86_expand_prologue): Corresponding changes.
2733 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2736 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2737 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2739 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2742 * lra-constraints.c (curr_insn_transform): Process AND in the
2745 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2747 PR rtl-optimization/83986
2748 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2749 dependence against last_pending_memory_flush in addition to
2752 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2754 PR tree-optimization/81611
2755 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2758 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2761 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2764 2018-01-30 Richard Biener <rguenther@suse.de>
2765 Jakub Jelinek <jakub@redhat.com>
2767 PR tree-optimization/84111
2768 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2769 inner loops added during recursion, as they don't have up-to-date
2772 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2775 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2776 (can_inline_edge_by_limits_p): ... here.
2777 (can_early_inline_edge_p, check_callers,
2778 update_caller_keys, update_callee_keys, recursive_inlining,
2779 add_new_edges_to_heap, speculation_useful_p,
2780 inline_small_functions,
2781 inline_small_functions, flatten_function,
2782 inline_to_all_callers_1): Update.
2784 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2786 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2789 2018-01-30 Richard Biener <rguenther@suse.de>
2791 PR tree-optimization/83008
2792 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2793 invariant and constant vector uses in stmts when they need
2796 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2799 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2800 * configure: Regenerate.
2802 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2804 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2806 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2807 Use gen_rtx_REG rather than gen_lowpart.
2809 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2811 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2812 rather than 0 when creating partial subregs.
2814 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2816 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2819 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2822 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2823 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2824 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2825 flags. This restores the settings used before the 2017-07-24.
2826 Turning off pre increment/decrement/modify allows IVOPTS to
2827 optimize DF/SF loops where the index is an int.
2829 2018-01-29 Richard Biener <rguenther@suse.de>
2830 Kelvin Nilsen <kelvin@gcc.gnu.org>
2833 * tree-vect-stmts.c (vectorizable_call): Don't call
2834 targetm.vectorize_builtin_md_vectorized_function if callee is
2837 2018-01-22 Carl Love <cel@us.ibm.com>
2839 * doc/extend.tex: Fix typo in second arg in
2840 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2842 2018-01-29 Richard Biener <rguenther@suse.de>
2844 PR tree-optimization/84086
2845 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2846 (flush_ssaname_freelist): When SSA names were released reset
2847 the SCEV hash table.
2849 2018-01-29 Richard Biener <rguenther@suse.de>
2851 PR tree-optimization/84057
2852 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2853 removed paths when removing edges.
2855 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2857 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2858 -mfunction-return=@var{choice}.
2860 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2863 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2865 (layout::print_source_line): Likewise.
2866 (test_get_line_width_without_trailing_whitespace): Add test cases.
2868 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2871 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2874 2018-01-26 Jim Wilson <jimw@sifive.com>
2876 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2878 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2881 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2883 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2884 and CMP + SUB-immediate -> SUBS.
2886 2018-01-26 Martin Sebor <msebor@redhat.com>
2888 PR tree-optimization/83896
2889 * tree-ssa-strlen.c (get_string_len): Rename...
2890 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2891 Avoid assuming length is constant.
2892 (handle_char_store): Use HOST_WIDE_INT for string length.
2894 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2897 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2898 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2900 2018-01-26 Richard Biener <rguenther@suse.de>
2902 PR rtl-optimization/84003
2903 * dse.c (record_store): Only record redundant stores when
2904 the earlier store aliases at least all accesses the later one does.
2906 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2908 PR rtl-optimization/83985
2909 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2910 REG_CFA_RESTORE insns.
2911 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2912 REG_CFA_RESTORE insns here.
2915 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2916 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2918 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2920 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2921 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2922 (arc_init): Likewise.
2923 (arc_override_options): Likewise.
2924 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2926 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2928 * config/arc/arc.h (TARGET_DBNZ): Define.
2929 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2930 properly set the tune attribute.
2931 (dbnz): Use TARGET_DBNZ guard.
2932 * config/arc/arc.opt (mtune): Add core3 option.
2934 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2936 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2937 recognize new pic like addresses.
2938 (arc_delegitimize_address): Clean up.
2940 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2942 * config/arc/arc-arches.def: Option mrf16 valid for all
2944 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2945 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2946 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2947 * config/arc/arc-tables.opt: Regenerate.
2948 * config/arc/arc.c (arc_conditional_register_usage): Handle
2949 reduced register file case.
2950 (arc_file_start): Set must have build attributes.
2951 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2953 * config/arc/arc.opt (mrf16): Add new option.
2954 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2955 * config/arc/genmultilib.awk: Handle new mrf16 option.
2956 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2957 * config/arc/t-multilib: Regenerate.
2958 * doc/invoke.texi (ARC Options): Document mrf16 option.
2960 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2962 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2963 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2964 (arc_attribute_table): Add 'secure_call' attribute.
2965 (arc_print_operand): Print secure call operand.
2966 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2968 (arc_is_secure_call_p): New function. * config/arc/arc.md
2969 (call_i): Add support for sjli instruction.
2970 (call_value_i): Likewise.
2971 * config/arc/constraints.md (Csc): New constraint.
2973 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2974 John Eric Martin <John.Martin@emmicro-us.com>
2976 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2977 * config/arc/arc.c (_arc_jli_section): New struct.
2978 (arc_jli_section): New type.
2979 (rc_jli_sections): New static variable.
2980 (arc_handle_jli_attribute): New function.
2981 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2982 (arc_file_end): New function.
2983 (TARGET_ASM_FILE_END): Define.
2984 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2985 (arc_add_jli_section): New function.
2986 (jli_call_scan): Likewise.
2987 (arc_reorg): Call jli_call_scan.
2988 (arc_output_addsi): Remove 'S' from printing asm operand.
2989 (arc_is_jli_call_p): New function.
2990 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2992 (movhi_insn): Likewise.
2993 (movsi_insn): Likewise.
2994 (movsi_set_cc_insn): Likewise.
2995 (loadqi_update): Likewise.
2996 (load_zeroextendqisi_update): Likewise.
2997 (load_signextendqisi_update): Likewise.
2998 (loadhi_update): Likewise.
2999 (load_zeroextendhisi_update): Likewise.
3000 (load_signextendhisi_update): Likewise.
3001 (loadsi_update): Likewise.
3002 (loadsf_update): Likewise.
3003 (movsicc_insn): Likewise.
3004 (bset_insn): Likewise.
3005 (bxor_insn): Likewise.
3006 (bclr_insn): Likewise.
3007 (bmsk_insn): Likewise.
3008 (bicsi3_insn): Likewise.
3009 (cmpsi_cc_c_insn): Likewise.
3010 (movsi_ne): Likewise.
3011 (movsi_cond_exec): Likewise.
3012 (clrsbsi2): Likewise.
3023 (call_i): Remove 'S' asm letter, add jli instruction.
3024 (call_value_i): Likewise.
3025 * config/arc/arc.op (mjli-always): New option.
3026 * config/arc/constraints.md (Cji): New constraint.
3027 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3029 (subsf3_fpx): Likewise.
3030 (mulsf3_fpx): Likewise.
3031 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3033 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3035 * doc/invoke.texi (ARC): Document mjli-always option.
3037 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3039 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3040 avoid addition with 0 and use incw and decw where possible.
3042 2018-01-26 Richard Biener <rguenther@suse.de>
3044 PR tree-optimization/81082
3045 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3046 association if it requires casting to unsigned.
3047 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3048 from fold_plusminus_mult_expr to catch important cases late when
3049 range info is available.
3051 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3053 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3054 * configure.ac (hidden_linkonce): New test.
3055 * configure: Regenerate.
3056 * config.in: Regenerate.
3058 2018-01-26 Julia Koval <julia.koval@intel.com>
3060 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3061 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3062 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3063 _mm_mask_bitshuffle_epi64_mask): Fix type.
3064 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3065 USI_FTYPE_V4DI_V4DI_USI): Remove.
3066 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3067 __builtin_ia32_vpshufbitqmb256_mask,
3068 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3069 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3070 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3072 2018-01-26 Alan Modra <amodra@gmail.com>
3075 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3076 UNSPEC_VBPERMQ. Sort other unspecs.
3078 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3080 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3082 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3085 * predict.c (drop_profile): Do not push/pop cfun; update also
3087 (handle_missing_profiles): Fix logic looking for zero profiles.
3089 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3092 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3093 on functions with #pragma omp declare simd or functions with simd
3095 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3096 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3097 Remove trailing \n from warning_at calls.
3099 2018-01-25 Tom de Vries <tom@codesourcery.com>
3102 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3103 for neutered workers.
3105 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3108 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3109 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3111 2018-01-24 Jeff Law <law@redhat.com>
3114 * i386.c (get_probe_interval): Move to earlier point.
3115 (ix86_compute_frame_layout): If -fstack-clash-protection and
3116 the frame is larger than the probe interval, then use pushes
3117 to save registers rather than reg->mem moves.
3118 (ix86_expand_prologue): Remove conditional for int_registers_saved
3121 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3124 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3125 min/max for never referenced object.
3127 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3130 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3132 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3133 attributes from DECL_ATTRIBUTES (decl) without affecting
3134 DECL_ATTRIBUTES (current_function_decl).
3135 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3136 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3138 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3140 PR tree-optimization/83979
3141 * fold-const.c (fold_comparison): Use constant_boolean_node
3142 instead of boolean_{true,false}_node.
3144 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3146 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3149 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3151 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3152 Simplify the clause that sets the length attribute.
3153 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3154 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3155 clause that sets the length attribute.
3156 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3158 2018-01-24 Tom de Vries <tom@codesourcery.com>
3161 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3162 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3163 Add strict parameter.
3164 (prevent_branch_around_nothing): Insert dummy insn between branch to
3165 label and label with no ptx insn inbetween.
3166 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3168 2018-01-24 Tom de Vries <tom@codesourcery.com>
3171 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3172 for neutered threads in warp.
3173 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3175 2018-01-24 Richard Biener <rguenther@suse.de>
3177 PR tree-optimization/83176
3178 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3181 2018-01-24 Richard Biener <rguenther@suse.de>
3183 PR tree-optimization/82819
3184 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3185 code generating pluses that are no-ops in the target precision.
3187 2018-01-24 Richard Biener <rguenther@suse.de>
3190 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3192 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3194 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3195 to merge probabilities.
3196 * predict.c (probably_never_executed): Also mark as cold functions
3197 with global 0 profile and guessed local profile.
3198 * profile-count.c (profile_probability::combine_with_count): New
3200 * profile-count.h (profile_probability::operator*,
3201 profile_probability::operator*=, profile_probability::operator/,
3202 profile_probability::operator/=): Reduce precision to adjusted
3203 and set value to guessed on contradictory divisions.
3204 (profile_probability::combine_with_freq): Remove.
3205 (profile_probability::combine_wiht_count): Declare.
3206 (profile_count::force_nonzero):: Set to adjusted.
3207 (profile_count::probability_in):: Set quality to adjusted.
3208 * tree-ssa-tail-merge.c (replace_block_by): Use
3211 2018-01-23 Andrew Waterman <andrew@sifive.com>
3212 Jim Wilson <jimw@sifive.com>
3214 * config/riscv/riscv.c (riscv_stack_boundary): New.
3215 (riscv_option_override): Set riscv_stack_boundary. Handle
3216 riscv_preferred_stack_boundary_arg.
3217 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3218 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3219 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3220 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3221 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3222 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3224 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3227 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3228 of struct ix86_frame.
3229 (ix86_expand_epilogue): Likewise. Add a local variable for
3230 the reg_save_offset field in struct ix86_frame.
3232 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3234 PR tree-optimization/82604
3235 * tree-loop-distribution.c (enum partition_kind): New enum item
3236 PKIND_PARTIAL_MEMSET.
3237 (partition_builtin_p): Support above new enum item.
3238 (generate_code_for_partition): Ditto.
3239 (compute_access_range): Differentiate cases that equality can be
3240 proven at all loops, the innermost loops or no loops.
3241 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3242 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3243 (finalize_partitions, distribute_loop): Don't fuse partition of
3244 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3245 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3248 2018-01-23 Martin Liska <mliska@suse.cz>
3250 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3251 order to ignore the predictor.
3252 (PRED_POLYMORPHIC_CALL): Likewise.
3253 (PRED_RECURSIVE_CALL): Likewise.
3255 2018-01-23 Martin Liska <mliska@suse.cz>
3257 * tree-profile.c (tree_profiling): Print function header to
3258 aware reader which function we are working on.
3259 * value-prof.c (gimple_find_values_to_profile): Do not print
3260 not interesting value histograms.
3262 2018-01-23 Martin Liska <mliska@suse.cz>
3264 * profile-count.h (enum profile_quality): Add
3265 profile_uninitialized as the first value. Do not number values
3266 as they are zero based.
3267 (profile_count::verify): Update sanity check.
3268 (profile_probability::verify): Likewise.
3270 2018-01-23 Nathan Sidwell <nathan@acm.org>
3272 * doc/invoke.texi (ffor-scope): Deprecate.
3274 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3276 PR tree-optimization/83510
3277 * domwalk.c (set_all_edges_as_executable): New function.
3278 (dom_walker::dom_walker): Convert bool param
3279 "skip_unreachable_blocks" to enum reachability. Move setup of
3280 edge flags to set_all_edges_as_executable and only do it when
3281 reachability is REACHABLE_BLOCKS.
3282 * domwalk.h (enum dom_walker::reachability): New enum.
3283 (dom_walker::dom_walker): Convert bool param
3284 "skip_unreachable_blocks" to enum reachability.
3285 (set_all_edges_as_executable): New decl.
3286 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3287 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3289 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3290 but converting true to REACHABLE_BLOCKS.
3291 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3293 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3294 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3295 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3297 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3298 if check_all_array_refs will be called.
3300 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3302 * tree.c (selftest::test_location_wrappers): Add more test
3305 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3307 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3308 (selftest::test_bit_in_range): Likewise.
3310 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3313 * doc/sourcebuild.texi (vect_float): Say that the selector
3314 only describes the situation when -funsafe-math-optimizations is on.
3315 (vect_float_strict): Document.
3317 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3319 PR tree-optimization/83965
3320 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3321 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3322 instead of checking only for a reduction.
3323 (vect_recog_widen_sum_pattern): Likewise.
3325 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3327 * predict.c (probably_never_executed): Only use precise profile info.
3328 (compute_function_frequency): Skip after inlining hack since we now
3329 have quality checking.
3331 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3333 * profile-count.h (profile_probability::very_unlikely,
3334 profile_probability::unlikely, profile_probability::even): Set
3335 precision to guessed.
3337 2018-01-23 Richard Biener <rguenther@suse.de>
3339 PR tree-optimization/83963
3340 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3341 Properly terminate dominator walk when crossing the exit edge not
3342 when visiting its source block.
3344 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3347 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3348 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3350 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3352 PR tree-optimization/83957
3353 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3354 semicolon after for body surrounded by braces.
3356 PR tree-optimization/83081
3357 * profile-count.h (profile_probability::split): New method.
3358 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3359 Use profile_probability::split.
3360 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3361 when splitting a single conditional jump into 2.
3363 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3365 PR tree-optimization/69452
3366 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3369 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3371 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3372 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3373 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3375 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3377 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3378 * config/rl78/rl78.md (movdi): New define_expand.
3379 * config/rl78/rl78.c (rl78_split_movdi): New function.
3381 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3384 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3386 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3387 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3388 128-bit to produce an UNSPEC move to get the double word with the
3389 signbit and then a shift directly to do signbit.
3390 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3391 implementation with a new version that just does either a direct
3392 move or a regular move. Move memory interface to separate insns.
3393 Move insns so they are next to the expander.
3394 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3395 with signbit move. Split big and little endian case.
3396 (signbit<mode>2_dm_mem_le): Likewise.
3397 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3398 (signbit<mode>2_dm2): Likewise.
3400 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3402 * config/rl78/rl78.md (anddi3): New define_expand.
3404 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3406 * config/rl78/rl78.md (umindi3): New define_expand.
3408 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3410 * config/rl78/rl78.md (smindi3): New define_expand.
3412 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3414 * config/rl78/rl78.md (smaxdi3): New define_expand.
3416 2018-01-22 Carl Love <cel@us.ibm.com>
3418 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3419 LVX_V1TI): Add macro expansion.
3420 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3421 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3422 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3423 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3424 Change check to determine if the instruction is a byte reversing
3425 entry. Fix typo in comment.
3426 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3427 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3428 Add def_builtin calls for new builtins.
3429 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3430 Add define_insn expansion.
3432 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3434 * config/rl78/rl78.md (umaxdi3): New define_expand.
3436 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3438 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3439 for non-QImode registers.
3441 2018-01-22 Richard Biener <rguenther@suse.de>
3443 PR tree-optimization/83963
3444 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3445 including the loop exit block.
3446 (scop_detection::merge_sese): Likewise.
3447 (scop_detection::add_scop): Do it here instead.
3449 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3451 * doc/sourcebuild.texi (arm_softfloat): Document.
3453 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3456 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3457 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3458 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3460 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3461 David Edelsohn <dje.gcc@gmail.com>
3464 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3465 Change "crset eq" to "crset 2".
3466 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3467 (*call_indirect_aix<mode>_nospec): Likewise.
3468 (*call_value_indirect_aix<mode>_nospec): Likewise.
3469 (*call_indirect_elfv2<mode>_nospec): Likewise.
3470 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3471 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3472 change assembly output from . to $.
3473 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3474 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3475 (*tablejump<mode>_internal1_nospec): Likewise.
3477 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3480 * config/sh/sh_optimize_sett_clrt.cc:
3481 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3483 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3485 PR tree-optimization/83940
3486 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3487 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3488 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3489 use it to pass back the definition type.
3490 (vect_check_store_rhs): Likewise rhs_dt_out.
3491 (vect_build_gather_load_calls): Add a mask_dt argument and use
3492 it instead of a call to vect_is_simple_use.
3493 (vectorizable_store): Update calls to vect_check_load_store_mask
3494 and vect_check_store_rhs. Use the dt returned by the latter instead
3495 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3496 instead of calls to vect_is_simple_use. Pass the scalar rather
3497 than the vector operand to vect_is_simple_use when handling
3498 second and subsequent copies of an rhs value.
3499 (vectorizable_load): Update calls to vect_check_load_store_mask
3500 and vect_build_gather_load_calls. Use the cached mask_dt and
3501 gs_info.offset_dt instead of calls to vect_is_simple_use.
3503 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3506 * tree-emutls.c: Include gimplify.h.
3507 (lower_emutls_2): New function.
3508 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3509 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3510 it before further processing.
3513 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3514 UINTVAL (trueop1) instead of INTVAL (op1).
3516 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3520 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3521 INCOMING_FRAME_SP_OFFSET if not defined.
3522 (scan_trace): Add ENTRY argument. If true and
3523 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3524 emit a note to adjust the CFA offset.
3525 (create_cfi_notes): Adjust scan_trace callers.
3526 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3527 INCOMING_FRAME_SP_OFFSET in the CIE.
3528 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3529 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3531 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3532 * doc/tm.texi: Regenerated.
3534 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3536 PR rtl-optimization/83147
3537 * lra-constraints.c (remove_inheritance_pseudos): Use
3538 lra_substitute_pseudo_within_insn.
3540 2018-01-19 Tom de Vries <tom@codesourcery.com>
3541 Cesar Philippidis <cesar@codesourcery.com>
3544 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3546 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3549 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3550 spaces for function labels.
3552 2018-01-19 Martin Liska <mliska@suse.cz>
3554 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3555 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3556 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3557 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3558 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3559 (PRED_CONST_RETURN): Change from 69 to 65.
3560 (PRED_NULL_RETURN): Change from 91 to 71.
3561 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3562 (PRED_LOOP_GUARD): Change from 66 to 73.
3564 2018-01-19 Martin Liska <mliska@suse.cz>
3566 * predict.c (predict_insn_def): Add new assert.
3567 (struct branch_predictor): Change type to signed integer.
3568 (test_prediction_value_range): Amend test to cover
3570 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3571 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3572 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3573 (PRED_LOOP_IV_COMPARE): Likewise.
3574 * predict.h (PROB_UNINITIALIZED): Define new constant.
3576 2018-01-19 Martin Liska <mliska@suse.cz>
3578 * predict.c (dump_prediction): Add new format for
3579 analyze_brprob.py script which is enabled with -details
3581 * profile-count.h (precise_p): New function.
3583 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3585 PR tree-optimization/83922
3586 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3587 there are no statements that need masking.
3588 (vect_active_double_reduction_p): New function.
3589 (vect_analyze_loop_operations): Use it when handling phis that
3590 are not in the loop header.
3592 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3594 PR tree-optimization/83914
3595 * tree-vect-loop.c (vectorizable_induction): Don't convert
3596 init_expr or apply the peeling adjustment for inductions
3597 that are nested within the vectorized loop.
3599 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3601 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3604 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3608 * function.h (gimplify_parameters): Add gimple_seq * argument.
3609 * function.c: Include gimple.h and options.h.
3610 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3611 for the added local temporaries if needed.
3612 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3613 if there are any parameter cleanups, wrap whole body into a
3614 try/finally with the cleanups.
3616 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3619 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3620 Use GET_MODE_CLASS for scalar floating point.
3622 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3626 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3627 Fix call of call_cgraph_insertion_hooks.
3629 2018-01-18 Martin Sebor <msebor@redhat.com>
3631 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3633 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3636 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3639 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3642 * common.opt: (-ffile-prefix-map): New option.
3643 * opts.c (common_handle_option): Defer it.
3644 * opts-global.c (handle_common_deferred_options): Handle it.
3645 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3646 * file-prefix-map.h: New file.
3647 (remap_debug_filename, add_debug_prefix_map): ...here.
3648 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3649 * final.c (debug_prefix_map, add_debug_prefix_map
3650 remap_debug_filename): Move to...
3651 * file-prefix-map.c: New file.
3652 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3653 generalize, get rid of alloca(), use strrchr() instead of strchr().
3654 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3655 Implement in terms of add_prefix_map().
3656 (remap_macro_filename, remap_debug_filename): Implement in term of
3658 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3659 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3660 * dbxout.c: Include file-prefix-map.h.
3661 * varasm.c: Likewise.
3662 * vmsdbgout.c: Likewise.
3663 * xcoffout.c: Likewise.
3664 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3665 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3666 * doc/invoke.texi (-ffile-prefix-map): Document.
3667 (-fdebug-prefix-map): Update description.
3669 2018-01-18 Martin Liska <mliska@suse.cz>
3671 * config/i386/i386.c (indirect_thunk_name): Document that also
3673 (output_indirect_thunk): Document why both instructions
3674 (pause and lfence) are generated.
3676 2018-01-18 Richard Biener <rguenther@suse.de>
3678 PR tree-optimization/83887
3679 * graphite-scop-detection.c
3680 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3681 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3682 (scop_detection::merge_sese): Re-implement with a flood-fill
3683 algorithm that properly finds a SESE region if it exists.
3685 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3688 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3689 pointer_diff optimizations use view_convert instead of convert.
3691 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3693 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3694 Generate different code for -mno-speculate-indirect-jumps.
3695 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3696 (*call_indirect_aix<mode>): Disable for
3697 -mno-speculate-indirect-jumps.
3698 (*call_indirect_aix<mode>_nospec): New define_insn.
3699 (*call_value_indirect_aix<mode>): Disable for
3700 -mno-speculate-indirect-jumps.
3701 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3702 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3703 -mno-speculate-indirect-jumps.
3704 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3706 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3708 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3709 long double type, set the flags for noting the default long double
3710 type, even if we don't pass or return a long double type.
3712 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3715 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3718 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3720 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3721 support for merge[hl].
3722 (fold_mergehl_helper): New helper function.
3723 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3724 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3725 (altivec_vmrglw_direct): Add xxmrglw insn.
3727 2018-01-17 Andrew Waterman <andrew@sifive.com>
3729 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3730 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3732 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3735 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3736 call the lto_location_cache before reading the
3737 DECL_SOURCE_LOCATION of the types.
3739 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3740 Richard Sandiford <richard.sandiford@linaro.org>
3742 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3743 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3744 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3745 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3746 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3748 * config/aarch64/constraints.md (aarch64_movti_operand):
3750 * config/aarch64/predicates.md (Uti): Add new constraint.
3752 2018-01-17 Carl Love <cel@us.ibm.com>
3754 * config/rs6000/vsx.md (define_expand xl_len_r,
3755 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3756 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3758 (define_expand, define_insn): Move the shift left from the
3759 define_insn to the define_expand for lxvl and stxvl instructions.
3760 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3761 and XL_LEN_R definitions to PURE.
3763 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3765 * config/i386/i386.c (indirect_thunk_name): Declare regno
3766 as unsigned int. Compare regno with INVALID_REGNUM.
3767 (output_indirect_thunk): Ditto.
3768 (output_indirect_thunk_function): Ditto.
3769 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3770 in the call to output_indirect_thunk_function.
3772 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3775 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3776 rather than the size of inner_type to determine the stack slot size
3777 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3779 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3782 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3785 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3787 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3788 endian Linux systems to optionally enable multilibs for selecting
3789 the long double type if the user configured an explicit type.
3790 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3791 have no long double multilibs if not defined.
3792 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3793 warn if the user used -mabi={ieee,ibm}longdouble and we built
3794 multilibs for long double.
3795 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3796 appropriate multilib option.
3797 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3799 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3800 for building long double multilibs.
3801 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3803 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3805 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3808 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3810 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3813 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3816 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3819 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3821 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3822 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3824 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3826 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3827 different rtl trees depending on TARGET_64BIT.
3828 (rs6000_gen_lvx): Likewise.
3830 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3832 * config/visium/visium.md (nop): Tweak comment.
3833 (hazard_nop): Likewise.
3835 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3837 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3838 -mspeculate-indirect-jumps.
3839 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3840 for -mno-speculate-indirect-jumps.
3841 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3842 (*call_value_indirect_elfv2<mode>): Disable for
3843 -mno-speculate-indirect-jumps.
3844 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3845 (indirect_jump): Emit different RTL for
3846 -mno-speculate-indirect-jumps.
3847 (*indirect_jump<mode>): Disable for
3848 -mno-speculate-indirect-jumps.
3849 (*indirect_jump<mode>_nospec): New define_insn.
3850 (tablejump): Emit different RTL for
3851 -mno-speculate-indirect-jumps.
3852 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3853 (tablejumpsi_nospec): New define_expand.
3854 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3855 (tablejumpdi_nospec): New define_expand.
3856 (*tablejump<mode>_internal1): Disable for
3857 -mno-speculate-indirect-jumps.
3858 (*tablejump<mode>_internal1_nospec): New define_insn.
3859 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3862 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3864 * caller-save.c (insert_save): Drop unnecessary parameter. All
3867 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3868 Richard Biener <rguenth@suse.de>
3871 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3872 return early, inline manually is_gimple_sizepos. Make sure if we
3873 call gimplify_expr we don't end up with a gimple constant.
3874 * tree.c (variably_modified_type_p): Don't return true for
3875 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3876 * gimplify.h (is_gimple_sizepos): Remove.
3878 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3880 PR tree-optimization/83857
3881 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3882 vectorizable_live_operation for pure SLP statements.
3883 (vectorizable_live_operation): Handle PHIs.
3885 2018-01-16 Richard Biener <rguenther@suse.de>
3887 PR tree-optimization/83867
3888 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3889 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3891 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3894 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3895 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3896 If off is not INTEGER_CST, issue a may not be aligned warning
3897 rather than isn't aligned. Use isn%'t rather than isn't.
3898 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3900 <case MULT_EXPR>: Improve the case when bottom and one of the
3901 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3902 operand, in that case check if the other operand is multiple of
3903 bottom divided by the INTEGER_CST operand.
3905 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3908 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3909 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3910 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3911 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3912 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3913 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3914 (pa_function_arg_size): New function.
3916 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3918 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3919 in a separate statement.
3921 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3923 PR tree-optimization/83847
3924 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3925 group gathers and scatters.
3927 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3929 PR rtl-optimization/86620
3930 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3932 PR rtl-optimization/83213
3933 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3934 to last if both are JUMP_INSNs.
3936 PR tree-optimization/83843
3937 * gimple-ssa-store-merging.c
3938 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3939 store_immediate_info for bswap/nop orig_stores.
3941 2018-01-15 Andrew Waterman <andrew@sifive.com>
3943 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3945 <UDIV>: Increase cost if !TARGET_DIV.
3947 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3949 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3950 (define_attr "cr_logical_3op"): New.
3951 (cceq_ior_compare): Adjust.
3952 (cceq_ior_compare_complement): Adjust.
3953 (*cceq_rev_compare): Adjust.
3954 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3955 (is_cracked_insn): Adjust.
3956 (insn_must_be_first_in_group): Adjust.
3957 * config/rs6000/40x.md: Adjust.
3958 * config/rs6000/440.md: Adjust.
3959 * config/rs6000/476.md: Adjust.
3960 * config/rs6000/601.md: Adjust.
3961 * config/rs6000/603.md: Adjust.
3962 * config/rs6000/6xx.md: Adjust.
3963 * config/rs6000/7450.md: Adjust.
3964 * config/rs6000/7xx.md: Adjust.
3965 * config/rs6000/8540.md: Adjust.
3966 * config/rs6000/cell.md: Adjust.
3967 * config/rs6000/e300c2c3.md: Adjust.
3968 * config/rs6000/e500mc.md: Adjust.
3969 * config/rs6000/e500mc64.md: Adjust.
3970 * config/rs6000/e5500.md: Adjust.
3971 * config/rs6000/e6500.md: Adjust.
3972 * config/rs6000/mpc.md: Adjust.
3973 * config/rs6000/power4.md: Adjust.
3974 * config/rs6000/power5.md: Adjust.
3975 * config/rs6000/power6.md: Adjust.
3976 * config/rs6000/power7.md: Adjust.
3977 * config/rs6000/power8.md: Adjust.
3978 * config/rs6000/power9.md: Adjust.
3979 * config/rs6000/rs64.md: Adjust.
3980 * config/rs6000/titan.md: Adjust.
3982 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3984 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3985 ix86_indirect_branch_register logic.
3987 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3989 * config/i386/constraints.md (Bs): Update
3990 ix86_indirect_branch_register check. Don't check
3991 ix86_indirect_branch_register with GOT_memory_operand.
3993 * config/i386/predicates.md (GOT_memory_operand): Don't check
3994 ix86_indirect_branch_register here.
3995 (GOT32_symbol_operand): Likewise.
3997 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3999 * config/i386/predicates.md (constant_call_address_operand):
4000 Rewrite ix86_indirect_branch_register logic.
4001 (sibcall_insn_operand): Likewise.
4003 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4005 * config/i386/constraints.md (Bs): Replace
4006 ix86_indirect_branch_thunk_register with
4007 ix86_indirect_branch_register.
4009 * config/i386/i386.md (indirect_jump): Likewise.
4010 (tablejump): Likewise.
4011 (*sibcall_memory): Likewise.
4012 (*sibcall_value_memory): Likewise.
4013 Peepholes of indirect call and jump via memory: Likewise.
4014 * config/i386/i386.opt: Likewise.
4015 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4016 (GOT_memory_operand): Likewise.
4017 (call_insn_operand): Likewise.
4018 (sibcall_insn_operand): Likewise.
4019 (GOT32_symbol_operand): Likewise.
4021 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4024 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4025 type rather than type addr's type points to.
4026 (expand_omp_atomic_mutex): Likewise.
4027 (expand_omp_atomic): Likewise.
4029 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4032 * config/i386/i386.c (output_indirect_thunk_function): Use
4033 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4034 for __x86_return_thunk.
4036 2018-01-15 Richard Biener <rguenther@suse.de>
4039 * expmed.c (extract_bit_field_1): Fix typo.
4041 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4044 * config/arm/iterators.md (VF): New mode iterator.
4045 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4046 Remove integer-related logic from pattern.
4047 (neon_vabd<mode>_3): Likewise.
4049 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4052 * common.opt (fstrict-overflow): No longer an alias.
4053 (fwrapv-pointer): New option.
4054 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4055 also for pointer types based on flag_wrapv_pointer.
4056 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4057 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4058 opts->x_flag_wrapv got set.
4059 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4060 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4061 POINTER_TYPE_OVERFLOW_UNDEFINED.
4062 * match.pd: Likewise in address comparison pattern.
4063 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4065 2018-01-15 Richard Biener <rguenther@suse.de>
4068 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4069 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4070 Reset type names to their identifier if their TYPE_DECL doesn't
4071 have linkage (and thus is used for ODR and devirt).
4072 (save_debug_info_for_decl): Remove.
4073 (save_debug_info_for_type): Likewise.
4074 (add_tree_to_fld_list): Adjust.
4075 * tree-pretty-print.c (dump_generic_node): Make dumping of
4076 type names more robust.
4078 2018-01-15 Richard Biener <rguenther@suse.de>
4080 * BASE-VER: Bump to 8.0.1.
4082 2018-01-14 Martin Sebor <msebor@redhat.com>
4085 * builtins.c (check_access): Avoid warning when the no-warning bit
4088 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4090 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4091 * ira-color (allocno_hard_regs_compare): Likewise.
4093 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4096 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4097 Use .pushsection/.popsection.
4099 2018-01-14 Martin Sebor <msebor@redhat.com>
4102 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4104 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4106 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4107 entry from extra_headers.
4108 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4109 extra_headers, make the list bitwise identical to the i?86-*-* one.
4111 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4113 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4114 -mcmodel=large with -mindirect-branch=thunk,
4115 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4116 -mfunction-return=thunk-extern.
4117 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4118 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4119 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4121 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4123 * config/i386/i386.c (print_reg): Print the name of the full
4124 integer register without '%'.
4125 (ix86_print_operand): Handle 'V'.
4126 * doc/extend.texi: Document 'V' modifier.
4128 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4130 * config/i386/constraints.md (Bs): Disallow memory operand for
4131 -mindirect-branch-register.
4133 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4134 (GOT_memory_operand): Likewise.
4135 (call_insn_operand): Likewise.
4136 (sibcall_insn_operand): Likewise.
4137 (GOT32_symbol_operand): Likewise.
4138 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4139 for -mindirect-branch-register.
4140 (tablejump): Likewise.
4141 (*sibcall_memory): Likewise.
4142 (*sibcall_value_memory): Likewise.
4143 Disallow peepholes of indirect call and jump via memory for
4144 -mindirect-branch-register.
4145 (*call_pop): Replace m with Bw.
4146 (*call_value_pop): Likewise.
4147 (*sibcall_pop_memory): Replace m with Bs.
4148 * config/i386/i386.opt (mindirect-branch-register): New option.
4149 * doc/invoke.texi: Document -mindirect-branch-register option.
4151 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4153 * config/i386/i386-protos.h (ix86_output_function_return): New.
4154 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4155 set function_return_type.
4156 (indirect_thunk_name): Add ret_p to indicate thunk for function
4158 (output_indirect_thunk_function): Pass false to
4159 indirect_thunk_name.
4160 (ix86_output_indirect_branch_via_reg): Likewise.
4161 (ix86_output_indirect_branch_via_push): Likewise.
4162 (output_indirect_thunk_function): Create alias for function
4163 return thunk if regno < 0.
4164 (ix86_output_function_return): New function.
4165 (ix86_handle_fndecl_attribute): Handle function_return.
4166 (ix86_attribute_table): Add function_return.
4167 * config/i386/i386.h (machine_function): Add
4168 function_return_type.
4169 * config/i386/i386.md (simple_return_internal): Use
4170 ix86_output_function_return.
4171 (simple_return_internal_long): Likewise.
4172 * config/i386/i386.opt (mfunction-return=): New option.
4173 (indirect_branch): Mention -mfunction-return=.
4174 * doc/extend.texi: Document function_return function attribute.
4175 * doc/invoke.texi: Document -mfunction-return= option.
4177 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4179 * config/i386/i386-opts.h (indirect_branch): New.
4180 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4181 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4182 with local indirect jump when converting indirect call and jump.
4183 (ix86_set_indirect_branch_type): New.
4184 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4185 (indirectlabelno): New.
4186 (indirect_thunk_needed): Likewise.
4187 (indirect_thunk_bnd_needed): Likewise.
4188 (indirect_thunks_used): Likewise.
4189 (indirect_thunks_bnd_used): Likewise.
4190 (INDIRECT_LABEL): Likewise.
4191 (indirect_thunk_name): Likewise.
4192 (output_indirect_thunk): Likewise.
4193 (output_indirect_thunk_function): Likewise.
4194 (ix86_output_indirect_branch_via_reg): Likewise.
4195 (ix86_output_indirect_branch_via_push): Likewise.
4196 (ix86_output_indirect_branch): Likewise.
4197 (ix86_output_indirect_jmp): Likewise.
4198 (ix86_code_end): Call output_indirect_thunk_function if needed.
4199 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4201 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4202 (ix86_attribute_table): Add indirect_branch.
4203 * config/i386/i386.h (machine_function): Add indirect_branch_type
4204 and has_local_indirect_jump.
4205 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4207 (tablejump): Likewise.
4208 (*indirect_jump): Use ix86_output_indirect_jmp.
4209 (*tablejump_1): Likewise.
4210 (simple_return_indirect_internal): Likewise.
4211 * config/i386/i386.opt (mindirect-branch=): New option.
4212 (indirect_branch): New.
4215 (thunk-inline): Likewise.
4216 (thunk-extern): Likewise.
4217 * doc/extend.texi: Document indirect_branch function attribute.
4218 * doc/invoke.texi: Document -mindirect-branch= option.
4220 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4223 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4225 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4227 * ipa-inline.c (want_inline_small_function_p): Return false if
4228 inlining has already failed with CIF_FINAL_ERROR.
4229 (update_caller_keys): Call want_inline_small_function_p before
4231 (update_callee_keys): Likewise.
4233 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4235 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4237 (rs6000_quadword_masked_address_p): Likewise.
4238 (quad_aligned_load_p): Likewise.
4239 (quad_aligned_store_p): Likewise.
4240 (const_load_sequence_p): Add comment to describe the outer-most loop.
4241 (mimic_memory_attributes_and_flags): New function.
4242 (rs6000_gen_stvx): Likewise.
4243 (replace_swapped_aligned_store): Likewise.
4244 (rs6000_gen_lvx): Likewise.
4245 (replace_swapped_aligned_load): Likewise.
4246 (replace_swapped_load_constant): Capitalize argument name in
4247 comment describing this function.
4248 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4249 and stores that access quad-word aligned addresses and replace
4250 with stvx or lvx instructions when appropriate.
4251 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4252 New function prototype.
4253 (rs6000_quadword_masked_address_p): Likewise.
4254 (rs6000_gen_lvx): Likewise.
4255 (rs6000_gen_stvx): Likewise.
4256 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4257 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4258 when memory address is aligned.
4259 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4260 this split to select lvx instruction when memory address is aligned.
4261 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4262 instruction when memory address is aligned.
4263 (*vsx_le_perm_load_v16qi): Likewise.
4264 (four unnamed splitters): Modify to select the stvx instruction
4265 when memory is aligned.
4267 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4269 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4270 which appears in the queue multiple times.
4272 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4273 Alan Hayward <alan.hayward@arm.com>
4274 David Sherwood <david.sherwood@arm.com>
4276 * tree-vectorizer.h (vec_lower_bound): New structure.
4277 (_loop_vec_info): Add check_nonzero and lower_bounds.
4278 (LOOP_VINFO_CHECK_NONZERO): New macro.
4279 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4280 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4281 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4282 fields. Make seg_len the distance travelled, not including the
4284 (dr_direction_indicator): Declare.
4285 (dr_zero_step_indicator): Likewise.
4286 (dr_known_forward_stride_p): Likewise.
4287 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4289 (runtime_alias_check_p): Allow runtime alias checks with
4291 (operator ==): Compare access_size and align.
4292 (prune_runtime_alias_test_list): Rework for new distinction between
4293 the access_size and seg_len.
4294 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4296 (get_segment_min_max): New function.
4297 (create_intersect_range_checks): Use it.
4298 (dr_step_indicator): New function.
4299 (dr_direction_indicator): Likewise.
4300 (dr_zero_step_indicator): Likewise.
4301 (dr_known_forward_stride_p): Likewise.
4302 * tree-loop-distribution.c (data_ref_segment_size): Return
4303 DR_STEP * (niters - 1).
4304 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4306 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4307 (vect_preserves_scalar_order_p): New function, split out from...
4308 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4309 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4310 (vect_vfa_access_size): New function.
4311 (vect_vfa_align): Likewise.
4312 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4313 (dump_lower_bound): New function.
4314 (vect_check_lower_bound): Likewise.
4315 (vect_small_gap_p): Likewise.
4316 (vectorizable_with_step_bound_p): Likewise.
4317 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4318 depencies if the vectorization factor is 1. Convert the checks
4319 for nonzero steps into checks on the bounds of DR_STEP. Try using
4320 a bunds check for variable steps if the minimum required step is
4321 relatively small. Update calls to the dr_with_seg_len
4322 constructor and to vect_compile_time_alias.
4323 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4325 (vect_loop_versioning): Call it.
4326 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4328 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4330 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4331 Alan Hayward <alan.hayward@arm.com>
4332 David Sherwood <david.sherwood@arm.com>
4334 * doc/sourcebuild.texi (vect_scatter_store): Document.
4335 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4337 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4339 * genopinit.c (main): Add supports_vec_scatter_store and
4340 supports_vec_scatter_store_cached to target_optabs.
4341 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4342 IFN_MASK_SCATTER_STORE.
4343 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4345 * internal-fn.h (internal_store_fn_p): Declare.
4346 (internal_fn_stored_value_index): Likewise.
4347 * internal-fn.c (scatter_store_direct): New macro.
4348 (expand_scatter_store_optab_fn): New function.
4349 (direct_scatter_store_optab_supported_p): New macro.
4350 (internal_store_fn_p): New function.
4351 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4352 IFN_MASK_SCATTER_STORE.
4353 (internal_fn_mask_index): Likewise.
4354 (internal_fn_stored_value_index): New function.
4355 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4357 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4358 * optabs-query.c (supports_vec_scatter_store_p): New function.
4359 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4360 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4361 true for scatter stores.
4362 (vect_gather_scatter_fn_p): Handle scatter stores too.
4363 (vect_check_gather_scatter): Consider using scatter stores if
4364 supports_vec_scatter_store_p.
4365 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4367 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4368 internal_fn_stored_value_index.
4369 (check_load_store_masking): Handle scatter stores too.
4370 (vect_get_store_rhs): Make public.
4371 (vectorizable_call): Use internal_store_fn_p.
4372 (vectorizable_store): Handle scatter store internal functions.
4373 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4374 when deciding whether the end of the group has been reached.
4375 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4376 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4377 (mask_scatter_store<mode>): New insns.
4379 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4380 Alan Hayward <alan.hayward@arm.com>
4381 David Sherwood <david.sherwood@arm.com>
4383 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4384 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4385 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4387 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4388 Use vect_truncate_gather_scatter_offset if we can't treat the
4389 operation as a normal gather load or scatter store.
4390 (get_group_load_store_type): Take the gather_scatter_info
4391 as argument. Try using a gather load or scatter store for
4392 single-element groups.
4393 (get_load_store_type): Update calls to get_group_load_store_type
4394 and vect_use_strided_gather_scatters_p.
4396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4397 Alan Hayward <alan.hayward@arm.com>
4398 David Sherwood <david.sherwood@arm.com>
4400 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4401 optional tree argument.
4402 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4404 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4405 but continue to use the current value as a fallback.
4406 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4407 to compare the updates.
4408 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4409 (get_load_store_type): Use it when handling a strided access.
4410 (vect_get_strided_load_store_ops): New function.
4411 (vect_get_data_ptr_increment): Likewise.
4412 (vectorizable_load): Handle strided gather loads. Always pass
4413 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4416 Alan Hayward <alan.hayward@arm.com>
4417 David Sherwood <david.sherwood@arm.com>
4419 * doc/md.texi (gather_load@var{m}): Document.
4420 (mask_gather_load@var{m}): Likewise.
4421 * genopinit.c (main): Add supports_vec_gather_load and
4422 supports_vec_gather_load_cached to target_optabs.
4423 * optabs-tree.c (init_tree_optimization_optabs): Use
4424 ggc_cleared_alloc to allocate target_optabs.
4425 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4426 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4428 * internal-fn.h (internal_load_fn_p): Declare.
4429 (internal_gather_scatter_fn_p): Likewise.
4430 (internal_fn_mask_index): Likewise.
4431 (internal_gather_scatter_fn_supported_p): Likewise.
4432 * internal-fn.c (gather_load_direct): New macro.
4433 (expand_gather_load_optab_fn): New function.
4434 (direct_gather_load_optab_supported_p): New macro.
4435 (direct_internal_fn_optab): New function.
4436 (internal_load_fn_p): Likewise.
4437 (internal_gather_scatter_fn_p): Likewise.
4438 (internal_fn_mask_index): Likewise.
4439 (internal_gather_scatter_fn_supported_p): Likewise.
4440 * optabs-query.c (supports_at_least_one_mode_p): New function.
4441 (supports_vec_gather_load_p): Likewise.
4442 * optabs-query.h (supports_vec_gather_load_p): Declare.
4443 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4444 and memory_type field.
4445 (NUM_PATTERNS): Bump to 15.
4446 * tree-vect-data-refs.c: Include internal-fn.h.
4447 (vect_gather_scatter_fn_p): New function.
4448 (vect_describe_gather_scatter_call): Likewise.
4449 (vect_check_gather_scatter): Try using internal functions for
4450 gather loads. Recognize existing calls to a gather load function.
4451 (vect_analyze_data_refs): Consider using gather loads if
4452 supports_vec_gather_load_p.
4453 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4454 (vect_get_gather_scatter_offset_type): Likewise.
4455 (vect_convert_mask_for_vectype): Likewise.
4456 (vect_add_conversion_to_patterm): Likewise.
4457 (vect_try_gather_scatter_pattern): Likewise.
4458 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4459 (vect_vect_recog_func_ptrs): Add it.
4460 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4461 internal_fn_mask_index and internal_gather_scatter_fn_p.
4462 (check_load_store_masking): Take the gather_scatter_info as an
4463 argument and handle gather loads.
4464 (vect_get_gather_scatter_ops): New function.
4465 (vectorizable_call): Check internal_load_fn_p.
4466 (vectorizable_load): Likewise. Handle gather load internal
4468 (vectorizable_store): Update call to check_load_store_masking.
4469 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4470 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4471 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4472 (aarch64_gather_scale_operand_d): New predicates.
4473 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4474 (mask_gather_load<mode>): New insns.
4476 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4477 Alan Hayward <alan.hayward@arm.com>
4478 David Sherwood <david.sherwood@arm.com>
4480 * optabs.def (fold_left_plus_optab): New optab.
4481 * doc/md.texi (fold_left_plus_@var{m}): Document.
4482 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4483 * internal-fn.c (fold_left_direct): Define.
4484 (expand_fold_left_optab_fn): Likewise.
4485 (direct_fold_left_optab_supported_p): Likewise.
4486 * fold-const-call.c (fold_const_fold_left): New function.
4487 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4488 * tree-parloops.c (valid_reduction_p): New function.
4489 (gather_scalar_reductions): Use it.
4490 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4491 (vect_finish_replace_stmt): Declare.
4492 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4493 (needs_fold_left_reduction_p): New function, split out from...
4494 (vect_is_simple_reduction): ...here. Accept reductions that
4495 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4496 (vect_force_simple_reduction): Also store the reduction type in
4497 the assignment's STMT_VINFO_REDUC_TYPE.
4498 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4499 (merge_with_identity): New function.
4500 (vect_expand_fold_left): Likewise.
4501 (vectorize_fold_left_reduction): Likewise.
4502 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4503 scalar phi in place for it. Check for target support and reject
4504 cases that would reassociate the operation. Defer the transform
4505 phase to vectorize_fold_left_reduction.
4506 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4507 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4508 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4510 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4512 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4513 call to ifc_temp_var.
4515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4516 Alan Hayward <alan.hayward@arm.com>
4517 David Sherwood <david.sherwood@arm.com>
4519 * target.def (legitimize_address_displacement): Take the original
4520 offset as a poly_int.
4521 * targhooks.h (default_legitimize_address_displacement): Update
4523 * targhooks.c (default_legitimize_address_displacement): Likewise.
4524 * doc/tm.texi: Regenerate.
4525 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4526 as an argument, moving assert of ad->disp == ad->disp_term to...
4527 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4528 Try calling targetm.legitimize_address_displacement before expanding
4529 the address rather than afterwards, and adjust for the new interface.
4530 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4531 Match the new hook interface. Handle SVE addresses.
4532 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4535 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4537 * Makefile.in (OBJS): Add early-remat.o.
4538 * target.def (select_early_remat_modes): New hook.
4539 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4540 * doc/tm.texi: Regenerate.
4541 * targhooks.h (default_select_early_remat_modes): Declare.
4542 * targhooks.c (default_select_early_remat_modes): New function.
4543 * timevar.def (TV_EARLY_REMAT): New timevar.
4544 * passes.def (pass_early_remat): New pass.
4545 * tree-pass.h (make_pass_early_remat): Declare.
4546 * early-remat.c: New file.
4547 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4549 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4551 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4552 Alan Hayward <alan.hayward@arm.com>
4553 David Sherwood <david.sherwood@arm.com>
4555 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4556 vfm1 with a bound_epilog parameter.
4557 (vect_do_peeling): Update calls accordingly, and move the prologue
4558 call earlier in the function. Treat the base bound_epilog as 0 for
4559 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4560 this base when peeling for gaps.
4561 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4562 with fully-masked loops.
4563 (vect_estimate_min_profitable_iters): Handle the single peeled
4564 iteration in that case.
4566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4567 Alan Hayward <alan.hayward@arm.com>
4568 David Sherwood <david.sherwood@arm.com>
4570 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4571 single-element interleaving even if the size is not a power of 2.
4572 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4573 accesses for single-element interleaving if the group size is
4576 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4577 Alan Hayward <alan.hayward@arm.com>
4578 David Sherwood <david.sherwood@arm.com>
4580 * doc/md.texi (fold_extract_last_@var{m}): Document.
4581 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4582 * optabs.def (fold_extract_last_optab): New optab.
4583 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4584 * internal-fn.c (fold_extract_direct): New macro.
4585 (expand_fold_extract_optab_fn): Likewise.
4586 (direct_fold_extract_optab_supported_p): Likewise.
4587 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4588 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4589 EXTRACT_LAST_REDUCTION.
4590 (get_initial_def_for_reduction): Do not create an initial vector
4591 for EXTRACT_LAST_REDUCTION reductions.
4592 (vectorizable_reduction): Leave the scalar phi in place for
4593 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4594 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4595 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4596 transform phase to vectorizable_condition.
4597 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4599 (vect_finish_stmt_generation): ...here.
4600 (vect_finish_replace_stmt): New function.
4601 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4602 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4604 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4606 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4607 Alan Hayward <alan.hayward@arm.com>
4608 David Sherwood <david.sherwood@arm.com>
4610 * doc/md.texi (extract_last_@var{m}): Document.
4611 * optabs.def (extract_last_optab): New optab.
4612 * internal-fn.def (EXTRACT_LAST): New internal function.
4613 * internal-fn.c (cond_unary_direct): New macro.
4614 (expand_cond_unary_optab_fn): Likewise.
4615 (direct_cond_unary_optab_supported_p): Likewise.
4616 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4617 loops using EXTRACT_LAST.
4618 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4619 (extract_last_<mode>): ...this optab.
4620 (vec_extract<mode><Vel>): Update accordingly.
4622 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4623 Alan Hayward <alan.hayward@arm.com>
4624 David Sherwood <david.sherwood@arm.com>
4626 * target.def (empty_mask_is_expensive): New hook.
4627 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4628 * doc/tm.texi: Regenerate.
4629 * targhooks.h (default_empty_mask_is_expensive): Declare.
4630 * targhooks.c (default_empty_mask_is_expensive): New function.
4631 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4632 if the target says that empty masks are expensive.
4633 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4635 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4637 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4638 Alan Hayward <alan.hayward@arm.com>
4639 David Sherwood <david.sherwood@arm.com>
4641 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4642 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4643 (vect_use_loop_mask_for_alignment_p): New function.
4644 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4645 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4646 niters_skip argument. Make sure that the first niters_skip elements
4647 of the first iteration are inactive.
4648 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4649 Update call to vect_set_loop_masks_directly.
4650 (get_misalign_in_elems): New function, split out from...
4651 (vect_gen_prolog_loop_niters): ...here.
4652 (vect_update_init_of_dr): Take a code argument that specifies whether
4653 the adjustment should be added or subtracted.
4654 (vect_update_init_of_drs): Likewise.
4655 (vect_prepare_for_masked_peels): New function.
4656 (vect_do_peeling): Skip prologue peeling if we're using a mask
4657 instead. Update call to vect_update_inits_of_drs.
4658 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4660 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4661 alignment. Do not include the number of peeled iterations in
4662 the minimum threshold in that case.
4663 (vectorizable_induction): Adjust the start value down by
4664 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4665 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4666 Take the number of skipped iterations into account when calculating
4668 * tree-vect-stmts.c (vect_gen_while_not): New function.
4670 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4671 Alan Hayward <alan.hayward@arm.com>
4672 David Sherwood <david.sherwood@arm.com>
4674 * doc/sourcebuild.texi (vect_fully_masked): Document.
4675 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4677 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4679 (vect_analyze_loop_2): ...here. Don't check the vectorization
4680 factor against the number of loop iterations if the loop is
4683 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4684 Alan Hayward <alan.hayward@arm.com>
4685 David Sherwood <david.sherwood@arm.com>
4687 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4688 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4689 (dump_groups): Update accordingly.
4690 (iv_use::mem_type): New member variable.
4691 (address_p): New function.
4692 (record_use): Add a mem_type argument and initialize the new
4694 (record_group_use): Add a mem_type argument. Use address_p.
4695 Remove obsolete null checks of base_object. Update call to record_use.
4696 (find_interesting_uses_op): Update call to record_group_use.
4697 (find_interesting_uses_cond): Likewise.
4698 (find_interesting_uses_address): Likewise.
4699 (get_mem_type_for_internal_fn): New function.
4700 (find_address_like_use): Likewise.
4701 (find_interesting_uses_stmt): Try find_address_like_use before
4702 calling find_interesting_uses_op.
4703 (addr_offset_valid_p): Use the iv mem_type field as the type
4704 of the addressed memory.
4705 (add_autoinc_candidates): Likewise.
4706 (get_address_cost): Likewise.
4707 (split_small_address_groups_p): Use address_p.
4708 (split_address_groups): Likewise.
4709 (add_iv_candidate_for_use): Likewise.
4710 (autoinc_possible_for_pair): Likewise.
4711 (rewrite_groups): Likewise.
4712 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4713 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4714 (get_alias_ptr_type_for_ptr_address): New function.
4715 (rewrite_use_address): Rewrite address uses in calls that were
4716 identified by find_address_like_use.
4718 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4719 Alan Hayward <alan.hayward@arm.com>
4720 David Sherwood <david.sherwood@arm.com>
4722 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4724 * gimple-expr.h (is_gimple_addressable: Likewise.
4725 * gimple-expr.c (is_gimple_address): Likewise.
4726 * internal-fn.c (expand_call_mem_ref): New function.
4727 (expand_mask_load_optab_fn): Use it.
4728 (expand_mask_store_optab_fn): Likewise.
4730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4731 Alan Hayward <alan.hayward@arm.com>
4732 David Sherwood <david.sherwood@arm.com>
4734 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4735 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4736 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4737 (cond_umax@var{mode}): Document.
4738 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4739 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4740 (cond_umin_optab, cond_umax_optab): New optabs.
4741 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4742 (COND_IOR, COND_XOR): New internal functions.
4743 * internal-fn.h (get_conditional_internal_fn): Declare.
4744 * internal-fn.c (cond_binary_direct): New macro.
4745 (expand_cond_binary_optab_fn): Likewise.
4746 (direct_cond_binary_optab_supported_p): Likewise.
4747 (get_conditional_internal_fn): New function.
4748 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4749 Cope with reduction statements that are vectorized as calls rather
4751 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4752 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4753 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4754 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4755 (UNSPEC_COND_EOR): New unspecs.
4756 (optab): Add mappings for them.
4757 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4758 (sve_int_op, sve_fp_op): New int attributes.
4760 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4761 Alan Hayward <alan.hayward@arm.com>
4762 David Sherwood <david.sherwood@arm.com>
4764 * optabs.def (while_ult_optab): New optab.
4765 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4766 * internal-fn.def (WHILE_ULT): New internal function.
4767 * internal-fn.h (direct_internal_fn_supported_p): New override
4768 that takes two types as argument.
4769 * internal-fn.c (while_direct): New macro.
4770 (expand_while_optab_fn): New function.
4771 (convert_optab_supported_p): Likewise.
4772 (direct_while_optab_supported_p): New macro.
4773 * wide-int.h (wi::udiv_ceil): New function.
4774 * tree-vectorizer.h (rgroup_masks): New structure.
4775 (vec_loop_masks): New typedef.
4776 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4778 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4779 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4780 (vect_max_vf): New function.
4781 (slpeel_make_loop_iterate_ntimes): Delete.
4782 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4783 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4784 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4785 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4786 internal-fn.h, stor-layout.h and optabs-query.h.
4787 (vect_set_loop_mask): New function.
4788 (add_preheader_seq): Likewise.
4789 (add_header_seq): Likewise.
4790 (interleave_supported_p): Likewise.
4791 (vect_maybe_permute_loop_masks): Likewise.
4792 (vect_set_loop_masks_directly): Likewise.
4793 (vect_set_loop_condition_masked): Likewise.
4794 (vect_set_loop_condition_unmasked): New function, split out from
4795 slpeel_make_loop_iterate_ntimes.
4796 (slpeel_make_loop_iterate_ntimes): Rename to..
4797 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4798 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4799 (vect_do_peeling): Update call accordingly.
4800 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4802 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4803 mask_compare_type, can_fully_mask_p and fully_masked_p.
4804 (release_vec_loop_masks): New function.
4805 (_loop_vec_info): Use it to free the loop masks.
4806 (can_produce_all_loop_masks_p): New function.
4807 (vect_get_max_nscalars_per_iter): Likewise.
4808 (vect_verify_full_masking): Likewise.
4809 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4810 retries, and free the mask rgroups before retrying. Check loop-wide
4811 reasons for disallowing fully-masked loops. Make the final decision
4812 about whether use a fully-masked loop or not.
4813 (vect_estimate_min_profitable_iters): Do not assume that peeling
4814 for the number of iterations will be needed for fully-masked loops.
4815 (vectorizable_reduction): Disable fully-masked loops.
4816 (vectorizable_live_operation): Likewise.
4817 (vect_halve_mask_nunits): New function.
4818 (vect_double_mask_nunits): Likewise.
4819 (vect_record_loop_mask): Likewise.
4820 (vect_get_loop_mask): Likewise.
4821 (vect_transform_loop): Handle the case in which the final loop
4822 iteration might handle a partial vector. Call vect_set_loop_condition
4823 instead of slpeel_make_loop_iterate_ntimes.
4824 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4825 (check_load_store_masking): New function.
4826 (prepare_load_store_mask): Likewise.
4827 (vectorizable_store): Handle fully-masked loops.
4828 (vectorizable_load): Likewise.
4829 (supportable_widening_operation): Use vect_halve_mask_nunits for
4831 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4832 (vect_gen_while): New function.
4833 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4834 (aarch64_uqdec<mode>): New insn.
4836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4837 Alan Hayward <alan.hayward@arm.com>
4838 David Sherwood <david.sherwood@arm.com>
4840 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4841 (reduc_xor_scal_optab): New optabs.
4842 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4843 (reduc_xor_scal_@var{m}): Document.
4844 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4845 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4847 * fold-const-call.c (fold_const_call): Handle them.
4848 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4849 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4850 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4851 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4852 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4853 (UNSPEC_XORV): New unspecs.
4854 (optab): Add entries for them.
4855 (BITWISEV): New int iterator.
4856 (bit_reduc_op): New int attributes.
4858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4859 Alan Hayward <alan.hayward@arm.com>
4860 David Sherwood <david.sherwood@arm.com>
4862 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4863 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4864 * optabs.def (vec_shl_insert_optab): New optab.
4865 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4866 (duplicate_and_interleave): Likewise.
4867 * tree-vect-loop.c: Include internal-fn.h.
4868 (neutral_op_for_slp_reduction): New function, split out from
4869 get_initial_defs_for_reduction.
4870 (get_initial_def_for_reduction): Handle option 2 for variable-length
4871 vectors by loading the neutral value into a vector and then shifting
4872 the initial value into element 0.
4873 (get_initial_defs_for_reduction): Replace the code argument with
4874 the neutral value calculated by neutral_op_for_slp_reduction.
4875 Use gimple_build_vector for constant-length vectors.
4876 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4877 but the first group_size elements have a neutral value.
4878 Use duplicate_and_interleave otherwise.
4879 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4880 Update call to get_initial_defs_for_reduction. Handle SLP
4881 reductions for variable-length vectors by creating one vector
4882 result for each scalar result, with the elements associated
4883 with other scalar results stubbed out with the neutral value.
4884 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4885 Require IFN_VEC_SHL_INSERT for double reductions on
4886 variable-length vectors, or SLP reductions that have
4887 a neutral value. Require can_duplicate_and_interleave_p
4888 support for variable-length unchained SLP reductions if there
4889 is no neutral value, such as for MIN/MAX reductions. Also require
4890 the number of vector elements to be a multiple of the number of
4891 SLP statements when doing variable-length unchained SLP reductions.
4892 Update call to vect_create_epilog_for_reduction.
4893 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4894 and remove initial values.
4895 (duplicate_and_interleave): Make public.
4896 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4897 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4899 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4900 Alan Hayward <alan.hayward@arm.com>
4901 David Sherwood <david.sherwood@arm.com>
4903 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4904 (can_duplicate_and_interleave_p): New function.
4905 (vect_get_and_check_slp_defs): Take the vector of statements
4906 rather than just the current one. Remove excess parentheses.
4907 Restriction rejectinon of vect_constant_def and vect_external_def
4908 for variable-length vectors to boolean types, or types for which
4909 can_duplicate_and_interleave_p is false.
4910 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4911 (duplicate_and_interleave): New function.
4912 (vect_get_constant_vectors): Use gimple_build_vector for
4913 constant-length vectors and suitable variable-length constant
4914 vectors. Use duplicate_and_interleave for other variable-length
4915 vectors. Don't defer the update when inserting new statements.
4917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4918 Alan Hayward <alan.hayward@arm.com>
4919 David Sherwood <david.sherwood@arm.com>
4921 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4922 min_profitable_iters doesn't go negative.
4924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4925 Alan Hayward <alan.hayward@arm.com>
4926 David Sherwood <david.sherwood@arm.com>
4928 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4929 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4930 * optabs.def (vec_mask_load_lanes_optab): New optab.
4931 (vec_mask_store_lanes_optab): Likewise.
4932 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4933 (MASK_STORE_LANES): Likewise.
4934 * internal-fn.c (mask_load_lanes_direct): New macro.
4935 (mask_store_lanes_direct): Likewise.
4936 (expand_mask_load_optab_fn): Handle masked operations.
4937 (expand_mask_load_lanes_optab_fn): New macro.
4938 (expand_mask_store_optab_fn): Handle masked operations.
4939 (expand_mask_store_lanes_optab_fn): New macro.
4940 (direct_mask_load_lanes_optab_supported_p): Likewise.
4941 (direct_mask_store_lanes_optab_supported_p): Likewise.
4942 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4944 (vect_load_lanes_supported): Likewise.
4945 * tree-vect-data-refs.c (strip_conversion): New function.
4946 (can_group_stmts_p): Likewise.
4947 (vect_analyze_data_ref_accesses): Use it instead of checking
4948 for a pair of assignments.
4949 (vect_store_lanes_supported): Take a masked_p parameter.
4950 (vect_load_lanes_supported): Likewise.
4951 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4952 vect_store_lanes_supported and vect_load_lanes_supported.
4953 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4954 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4955 parameter. Don't allow gaps for masked accesses.
4956 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4957 and vect_load_lanes_supported.
4958 (get_load_store_type): Take a masked_p parameter and update
4959 call to get_group_load_store_type.
4960 (vectorizable_store): Update call to get_load_store_type.
4961 Handle IFN_MASK_STORE_LANES.
4962 (vectorizable_load): Update call to get_load_store_type.
4963 Handle IFN_MASK_LOAD_LANES.
4965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4966 Alan Hayward <alan.hayward@arm.com>
4967 David Sherwood <david.sherwood@arm.com>
4969 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4971 * config/aarch64/aarch64-protos.h
4972 (aarch64_sve_struct_memory_operand_p): Declare.
4973 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4974 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4975 (VPRED, vpred): Handle SVE structure modes.
4976 * config/aarch64/constraints.md (Utx): New constraint.
4977 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4978 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4979 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4980 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4981 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4982 structure modes. Split into pieces after RA.
4983 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4984 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4986 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4987 SVE structure modes.
4988 (aarch64_classify_address): Likewise.
4989 (sizetochar): Move earlier in file.
4990 (aarch64_print_operand): Handle SVE register lists.
4991 (aarch64_array_mode): New function.
4992 (aarch64_sve_struct_memory_operand_p): Likewise.
4993 (TARGET_ARRAY_MODE): Redefine.
4995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4996 Alan Hayward <alan.hayward@arm.com>
4997 David Sherwood <david.sherwood@arm.com>
4999 * target.def (array_mode): New target hook.
5000 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5001 * doc/tm.texi: Regenerate.
5002 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5003 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5004 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5006 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5010 Alan Hayward <alan.hayward@arm.com>
5011 David Sherwood <david.sherwood@arm.com>
5013 * fold-const.c (fold_binary_loc): Check the argument types
5014 rather than the result type when testing for a vector operation.
5016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5018 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5019 * doc/tm.texi: Regenerate.
5021 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5022 Alan Hayward <alan.hayward@arm.com>
5023 David Sherwood <david.sherwood@arm.com>
5025 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5026 (sve): Document new AArch64 extension.
5027 * doc/md.texi (w): Extend the description of the AArch64
5028 constraint to include SVE vectors.
5029 (Upl, Upa): Document new AArch64 predicate constraints.
5030 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5032 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5033 (msve-vector-bits=): New option.
5034 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5035 SVE when these are disabled.
5036 (sve): New extension.
5037 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5038 modes. Adjust their number of units based on aarch64_sve_vg.
5039 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5040 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5041 aarch64_addr_query_type.
5042 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5043 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5044 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5045 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5046 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5047 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5048 (aarch64_simd_imm_zero_p): Delete.
5049 (aarch64_check_zero_based_sve_index_immediate): Declare.
5050 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5051 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5052 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5053 (aarch64_sve_float_mul_immediate_p): Likewise.
5054 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5056 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5057 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5058 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5059 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5060 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5061 (aarch64_regmode_natural_size): Likewise.
5062 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5063 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5065 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5066 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5067 for VG and the SVE predicate registers.
5068 (V_ALIASES): Add a "z"-prefixed alias.
5069 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5070 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5071 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5072 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5073 (REG_CLASS_NAMES): Add entries for them.
5074 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5075 and the predicate registers.
5076 (aarch64_sve_vg): Declare.
5077 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5078 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5079 (REGMODE_NATURAL_SIZE): Define.
5080 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5082 * config/aarch64/aarch64.c: Include cfgrtl.h.
5083 (simd_immediate_info): Add a constructor for series vectors,
5084 and an associated step field.
5085 (aarch64_sve_vg): New variable.
5086 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5087 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5088 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5089 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5090 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5091 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5092 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5093 (aarch64_get_mask_mode): New functions.
5094 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5095 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5096 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5097 predicate modes and predicate registers. Explicitly restrict
5098 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5099 to store a vector mode if it is recognized by
5100 aarch64_classify_vector_mode.
5101 (aarch64_regmode_natural_size): New function.
5102 (aarch64_hard_regno_caller_save_mode): Return the original mode
5104 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5105 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5106 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5107 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5109 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5110 does not overlap dest if the function is frame-related. Handle
5112 (aarch64_split_add_offset): New function.
5113 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5114 them aarch64_add_offset.
5115 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5116 and update call to aarch64_sub_sp.
5117 (aarch64_add_cfa_expression): New function.
5118 (aarch64_expand_prologue): Pass extra temporary registers to the
5119 functions above. Handle the case in which we need to emit new
5120 DW_CFA_expressions for registers that were originally saved
5121 relative to the stack pointer, but now have to be expressed
5122 relative to the frame pointer.
5123 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5125 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5126 IP0 and IP1 values for SVE frames.
5127 (aarch64_expand_vec_series): New function.
5128 (aarch64_expand_sve_widened_duplicate): Likewise.
5129 (aarch64_expand_sve_const_vector): Likewise.
5130 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5131 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5132 into the register, rather than emitting a SET directly.
5133 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5134 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5135 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5136 (offset_9bit_signed_scaled_p): New functions.
5137 (aarch64_replicate_bitmask_imm): New function.
5138 (aarch64_bitmask_imm): Use it.
5139 (aarch64_cannot_force_const_mem): Reject expressions involving
5140 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5141 (aarch64_classify_index): Handle SVE indices, by requiring
5142 a plain register index with a scale that matches the element size.
5143 (aarch64_classify_address): Handle SVE addresses. Assert that
5144 the mode of the address is VOIDmode or an integer mode.
5145 Update call to aarch64_classify_symbol.
5146 (aarch64_classify_symbolic_expression): Update call to
5147 aarch64_classify_symbol.
5148 (aarch64_const_vec_all_in_range_p): New function.
5149 (aarch64_print_vector_float_operand): Likewise.
5150 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5151 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5152 and the FP immediates 1.0 and 0.5.
5153 (aarch64_print_address_internal): Handle SVE addresses.
5154 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5155 (aarch64_regno_regclass): Handle predicate registers.
5156 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5158 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5159 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5160 (aarch64_convert_sve_vector_bits): New function.
5161 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5162 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5164 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5165 Handle SVE vector and predicate modes. Accept VL-based constants
5166 that need only one temporary register, and VL offsets that require
5167 no temporary registers.
5168 (aarch64_conditional_register_usage): Mark the predicate registers
5169 as fixed if SVE isn't available.
5170 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5171 Return true for SVE vector and predicate modes.
5172 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5173 rather than an unsigned int. Handle SVE modes.
5174 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5176 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5178 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5179 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5180 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5181 (aarch64_sve_float_mul_immediate_p): New functions.
5182 (aarch64_sve_valid_immediate): New function.
5183 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5184 Explicitly reject structure modes. Check for INDEX constants.
5185 Handle PTRUE and PFALSE constants.
5186 (aarch64_check_zero_based_sve_index_immediate): New function.
5187 (aarch64_simd_imm_zero_p): Delete.
5188 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5189 vector modes. Accept constants in the range of CNT[BHWD].
5190 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5191 ask for an Advanced SIMD mode.
5192 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5193 (aarch64_simd_vector_alignment): Handle SVE predicates.
5194 (aarch64_vectorize_preferred_vector_alignment): New function.
5195 (aarch64_simd_vector_alignment_reachable): Use it instead of
5197 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5198 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5200 (MAX_VECT_LEN): Delete.
5201 (expand_vec_perm_d): Add a vec_flags field.
5202 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5203 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5204 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5206 (aarch64_evpc_rev): Rename to...
5207 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5208 (aarch64_evpc_rev_global): New function.
5209 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5210 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5212 (aarch64_evpc_sve_tbl): New function.
5213 (aarch64_expand_vec_perm_const_1): Update after rename of
5214 aarch64_evpc_rev. Handle SVE permutes too, trying
5215 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5216 than aarch64_evpc_tbl.
5217 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5218 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5219 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5220 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5221 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5222 (aarch64_expand_sve_vcond): New functions.
5223 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5224 of aarch64_vector_mode_p.
5225 (aarch64_dwarf_poly_indeterminate_value): New function.
5226 (aarch64_compute_pressure_classes): Likewise.
5227 (aarch64_can_change_mode_class): Likewise.
5228 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5229 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5230 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5231 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5232 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5233 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5234 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5235 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5237 (Dn, Dl, Dr): Accept const as well as const_vector.
5238 (Dz): Likewise. Compare against CONST0_RTX.
5239 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5240 of "vector" where appropriate.
5241 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5242 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5243 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5244 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5245 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5246 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5247 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5248 (v_int_equiv): Extend to SVE modes.
5249 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5251 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5252 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5253 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5254 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5255 (SVE_COND_FP_CMP): New int iterators.
5256 (perm_hilo): Handle the new unpack unspecs.
5257 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5259 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5260 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5261 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5262 (aarch64_equality_operator, aarch64_constant_vector_operand)
5263 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5264 (aarch64_sve_nonimmediate_operand): Likewise.
5265 (aarch64_sve_general_operand): Likewise.
5266 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5267 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5268 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5269 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5270 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5271 (aarch64_sve_float_arith_immediate): Likewise.
5272 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5273 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5274 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5275 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5276 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5277 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5278 (aarch64_sve_float_arith_operand): Likewise.
5279 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5280 (aarch64_sve_float_mul_operand): Likewise.
5281 (aarch64_sve_vec_perm_operand): Likewise.
5282 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5283 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5284 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5285 as well as const_vector.
5286 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5287 in file. Use CONST0_RTX and CONSTM1_RTX.
5288 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5289 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5290 Use aarch64_simd_imm_zero.
5291 * config/aarch64/aarch64-sve.md: New file.
5292 * config/aarch64/aarch64.md: Include it.
5293 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5294 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5295 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5296 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5297 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5298 (sve): New attribute.
5299 (enabled): Disable instructions with the sve attribute unless
5301 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5302 aarch64_expand_mov_immediate.
5303 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5304 CNT[BHSD] immediates.
5305 (movti): Split CONST_POLY_INT moves into two halves.
5306 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5307 Split additions that need a temporary here if the destination
5308 is the stack pointer.
5309 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5310 (*add<mode>3_poly_1): New instruction.
5311 (set_clobber_cc): New expander.
5313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5315 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5316 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5317 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5318 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5319 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5320 Change innermode from fixed_mode_size to machine_mode.
5321 (simplify_subreg): Update call accordingly. Handle a constant-sized
5322 subreg of a variable-length CONST_VECTOR.
5324 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5325 Alan Hayward <alan.hayward@arm.com>
5326 David Sherwood <david.sherwood@arm.com>
5328 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5329 (add_offset_to_base): New function, split out from...
5330 (create_mem_ref): ...here. When handling a scale other than 1,
5331 check first whether the address is valid without the offset.
5332 Add it into the base if so, leaving the index and scale as-is.
5334 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5337 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5338 fold_for_warn before checking if arg2 is INTEGER_CST.
5340 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5342 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5343 (store_multiple_operation): Delete.
5344 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5345 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5346 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5347 guarded by TARGET_STRING.
5348 (rs6000_output_load_multiple): Delete.
5349 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5350 OPTION_MASK_STRING / TARGET_STRING handling.
5351 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5352 (const rs6000_opt_masks) <"string">: Change mask to 0.
5353 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5354 (MASK_STRING): Delete.
5355 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5357 (load_multiple): Delete.
5364 (store_multiple): Delete.
5371 (movmemsi_8reg): Delete.
5372 (corresponding unnamed define_insn): Delete.
5373 (movmemsi_6reg): Delete.
5374 (corresponding unnamed define_insn): Delete.
5375 (movmemsi_4reg): Delete.
5376 (corresponding unnamed define_insn): Delete.
5377 (movmemsi_2reg): Delete.
5378 (corresponding unnamed define_insn): Delete.
5379 (movmemsi_1reg): Delete.
5380 (corresponding unnamed define_insn): Delete.
5381 * config/rs6000/rs6000.opt (mno-string): New.
5382 (mstring): Replace by deprecation warning stub.
5383 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5385 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5387 * regrename.c (regrename_do_replace): If replacing the same
5388 reg multiple times, try to reuse last created gen_raw_REG.
5391 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5392 main to workaround a bug in GDB.
5394 2018-01-12 Tom de Vries <tom@codesourcery.com>
5397 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5399 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5401 PR rtl-optimization/80481
5402 * ira-color.c (get_cap_member): New function.
5403 (allocnos_conflict_by_live_ranges_p): Use it.
5404 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5405 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5407 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5410 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5411 (*saddl_se_1): Ditto.
5413 (*ssubl_se_1): Ditto.
5415 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5417 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5418 rather than wi::to_widest for DR_INITs.
5419 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5420 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5421 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5423 (vect_analyze_group_access_1): Note that here.
5425 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5427 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5428 polynomial type sizes.
5430 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5432 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5433 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5434 (gimple_add_tmp_var): Likewise.
5436 2018-01-12 Martin Liska <mliska@suse.cz>
5438 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5439 (gimple_alloc_sizes): Likewise.
5440 (dump_gimple_statistics): Use PRIu64 in printf format.
5441 * gimple.h: Change uint64_t to int.
5443 2018-01-12 Martin Liska <mliska@suse.cz>
5445 * tree-core.h: Use uint64_t instead of int.
5446 * tree.c (tree_node_counts): Likewise.
5447 (tree_node_sizes): Likewise.
5448 (dump_tree_statistics): Use PRIu64 in printf format.
5450 2018-01-12 Martin Liska <mliska@suse.cz>
5452 * Makefile.in: As qsort_chk is implemented in vec.c, add
5453 vec.o to linkage of gencfn-macros.
5454 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5455 passing the info to record_node_allocation_statistics.
5456 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5458 * ggc-common.c (struct ggc_usage): Add operator== and use
5459 it in operator< and compare function.
5460 * mem-stats.h (struct mem_usage): Likewise.
5461 * vec.c (struct vec_usage): Remove operator< and compare
5462 function. Can be simply inherited.
5464 2018-01-12 Martin Jambor <mjambor@suse.cz>
5467 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5468 * tree-ssa-math-opts.c: Include domwalk.h.
5469 (convert_mult_to_fma_1): New function.
5470 (fma_transformation_info): New type.
5471 (fma_deferring_state): Likewise.
5472 (cancel_fma_deferring): New function.
5473 (result_of_phi): Likewise.
5474 (last_fma_candidate_feeds_initial_phi): Likewise.
5475 (convert_mult_to_fma): Added deferring logic, split actual
5476 transformation to convert_mult_to_fma_1.
5477 (math_opts_dom_walker): New type.
5478 (math_opts_dom_walker::after_dom_children): New method, body moved
5479 here from pass_optimize_widening_mul::execute, added deferring logic
5481 (pass_optimize_widening_mul::execute): Moved most of code to
5482 math_opts_dom_walker::after_dom_children.
5483 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5484 * config/i386/i386.c (ix86_option_override_internal): Added
5485 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5487 2018-01-12 Richard Biener <rguenther@suse.de>
5490 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5491 inline instance vars.
5493 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5496 * config/rx/rx.c (rx_is_restricted_memory_address):
5499 2018-01-12 Richard Biener <rguenther@suse.de>
5501 PR tree-optimization/80846
5502 * target.def (split_reduction): New target hook.
5503 * targhooks.c (default_split_reduction): New function.
5504 * targhooks.h (default_split_reduction): Declare.
5505 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5506 target requests first reduce vectors by combining low and high
5508 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5509 (get_vectype_for_scalar_type_and_size): Export.
5510 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5511 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5512 * doc/tm.texi: Regenerate.
5513 * config/i386/i386.c (ix86_split_reduction): Implement
5514 TARGET_VECTORIZE_SPLIT_REDUCTION.
5516 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5519 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5520 in PIC mode except for TARGET_VXWORKS_RTP.
5521 * config/sparc/sparc.c: Include cfgrtl.h.
5522 (TARGET_INIT_PIC_REG): Define.
5523 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5524 (sparc_pic_register_p): New predicate.
5525 (sparc_legitimate_address_p): Use it.
5526 (sparc_legitimize_pic_address): Likewise.
5527 (sparc_delegitimize_address): Likewise.
5528 (sparc_mode_dependent_address_p): Likewise.
5529 (gen_load_pcrel_sym): Remove 4th parameter.
5530 (load_got_register): Adjust call to above. Remove obsolete stuff.
5531 (sparc_expand_prologue): Do not call load_got_register here.
5532 (sparc_flat_expand_prologue): Likewise.
5533 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5534 (sparc_use_pseudo_pic_reg): New function.
5535 (sparc_init_pic_reg): Likewise.
5536 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5537 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5539 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5541 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5542 Add item for branch_cost.
5544 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5546 PR rtl-optimization/83565
5547 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5548 not extend the result to a larger mode for rotate operations.
5549 (num_sign_bit_copies1): Likewise.
5551 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5554 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5556 Use values-Xc.o for -pedantic.
5557 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5559 2018-01-12 Martin Liska <mliska@suse.cz>
5562 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5564 (possible_polymorphic_call_targets): Use it.
5565 (ipa_devirt): Likewise.
5567 2018-01-12 Martin Liska <mliska@suse.cz>
5569 * profile-count.h (enum profile_quality): Use 0 as invalid
5570 enum value of profile_quality.
5572 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5574 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5575 -mext-string options.
5577 2018-01-12 Richard Biener <rguenther@suse.de>
5579 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5580 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5581 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5583 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5585 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5587 * configure.ac (--with-long-double-format): Add support for the
5588 configuration option to change the default long double format on
5590 * config.gcc (powerpc*-linux*-*): Likewise.
5591 * configure: Regenerate.
5592 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5593 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5594 used without modification.
5596 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5598 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5599 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5600 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5601 MISC_BUILTIN_SPEC_BARRIER.
5602 (rs6000_init_builtins): Likewise.
5603 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5605 (speculation_barrier): New define_insn.
5606 * doc/extend.texi: Document __builtin_speculation_barrier.
5608 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5611 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5612 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5613 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5615 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5616 integral modes instead of "ss" and "sd".
5617 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5618 vectors with 32-bit and 64-bit elements.
5619 (vecdupssescalarmodesuffix): New mode attribute.
5620 (vec_dup<mode>): Use it.
5622 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5625 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5626 frame if argument is passed on stack.
5628 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5631 * ree.c (combine_reaching_defs): Optimize also
5632 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5633 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5635 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5638 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5640 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5643 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5644 after they are computed.
5646 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5648 PR tree-optimization/83695
5649 * gimple-loop-linterchange.cc
5650 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5651 reset cached scev information after interchange.
5652 (pass_linterchange::execute): Remove call to scev_reset_htab.
5654 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5656 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5657 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5658 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5659 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5660 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5661 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5662 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5663 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5664 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5665 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5666 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5667 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5668 (V_lane_reg): Likewise.
5669 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5671 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5672 (vfmal_lane_low<mode>_intrinsic,
5673 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5674 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5675 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5676 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5677 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5678 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5680 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5682 * config/arm/arm-cpus.in (fp16fml): New feature.
5683 (ALL_SIMD): Add fp16fml.
5684 (armv8.2-a): Add fp16fml as an option.
5685 (armv8.3-a): Likewise.
5686 (armv8.4-a): Add fp16fml as part of fp16.
5687 * config/arm/arm.h (TARGET_FP16FML): Define.
5688 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5690 * config/arm/arm-modes.def (V2HF): Define.
5691 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5692 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5693 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5694 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5695 vfmsl_low, vfmsl_high): New set of builtins.
5696 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5697 (vfml_op): New code attribute.
5698 (VFMLHALVES): New int iterator.
5699 (VFML, VFMLSEL): New mode attributes.
5700 (V_reg): Define mapping for V2HF.
5701 (V_hi, V_lo): New mode attributes.
5702 (VF_constraint): Likewise.
5703 (vfml_half, vfml_half_selector): New int attributes.
5704 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5706 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5707 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5709 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5710 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5711 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5712 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5714 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5715 Document new effective target and option set.
5717 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5719 * config/arm/arm-cpus.in (armv8_4): New feature.
5720 (ARMv8_4a): New fgroup.
5721 (armv8.4-a): New arch.
5722 * config/arm/arm-tables.opt: Regenerate.
5723 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5724 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5725 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5726 Add matching rules for -march=armv8.4-a and extensions.
5727 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5729 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5732 * config/rx/rx.md (BW): New mode attribute.
5733 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5735 2018-01-11 Richard Biener <rguenther@suse.de>
5737 PR tree-optimization/83435
5738 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5739 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5740 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5742 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5743 Alan Hayward <alan.hayward@arm.com>
5744 David Sherwood <david.sherwood@arm.com>
5746 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5748 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5749 (aarch64_print_address_internal): Use it to check for a zero offset.
5751 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5752 Alan Hayward <alan.hayward@arm.com>
5753 David Sherwood <david.sherwood@arm.com>
5755 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5756 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5757 Return a poly_int64 rather than a HOST_WIDE_INT.
5758 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5759 rather than a HOST_WIDE_INT.
5760 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5761 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5762 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5763 final_offset from HOST_WIDE_INT to poly_int64.
5764 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5765 to_constant when getting the number of units in an Advanced SIMD
5767 (aarch64_builtin_vectorized_function): Check for a constant number
5769 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5771 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5772 attribute instead of GET_MODE_NUNITS.
5773 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5774 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5775 GET_MODE_SIZE for fixed-size registers.
5776 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5777 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5778 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5779 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5780 (aarch64_print_operand, aarch64_print_address_internal)
5781 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5782 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5783 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5784 Handle polynomial GET_MODE_SIZE.
5785 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5786 wider than SImode without modification.
5787 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5788 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5789 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5790 passing and returning SVE modes.
5791 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5792 rather than GEN_INT.
5793 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5794 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5795 (aarch64_allocate_and_probe_stack_space): Likewise.
5796 (aarch64_layout_frame): Cope with polynomial offsets.
5797 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5798 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5800 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5801 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5802 poly_int64 rather than a HOST_WIDE_INT.
5803 (aarch64_get_separate_components, aarch64_process_components)
5804 (aarch64_expand_prologue, aarch64_expand_epilogue)
5805 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5806 (aarch64_anchor_offset): New function, split out from...
5807 (aarch64_legitimize_address): ...here.
5808 (aarch64_builtin_vectorization_cost): Handle polynomial
5809 TYPE_VECTOR_SUBPARTS.
5810 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5812 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5813 number of elements from the PARALLEL rather than the mode.
5814 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5815 rather than GET_MODE_BITSIZE.
5816 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5817 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5818 (aarch64_expand_vec_perm_const_1): Handle polynomial
5819 d->perm.length () and d->perm elements.
5820 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5821 Apply to_constant to d->perm elements.
5822 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5823 polynomial CONST_VECTOR_NUNITS.
5824 (aarch64_move_pointer): Take amount as a poly_int64 rather
5826 (aarch64_progress_pointer): Avoid temporary variable.
5827 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5828 the mode attribute instead of GET_MODE.
5830 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5831 Alan Hayward <alan.hayward@arm.com>
5832 David Sherwood <david.sherwood@arm.com>
5834 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5835 x exists before using it.
5836 (aarch64_add_constant_internal): Rename to...
5837 (aarch64_add_offset_1): ...this. Replace regnum with separate
5838 src and dest rtxes. Handle the case in which they're different,
5839 including when the offset is zero. Replace scratchreg with an rtx.
5840 Use 2 additions if there is no spare register into which we can
5841 move a 16-bit constant.
5842 (aarch64_add_constant): Delete.
5843 (aarch64_add_offset): Replace reg with separate src and dest
5844 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5845 Use aarch64_add_offset_1.
5846 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5847 an rtx rather than an int. Take the delta as a poly_int64
5848 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5849 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5850 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5851 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5852 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5854 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5855 aarch64_add_constant.
5857 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5859 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5860 Use scalar_float_mode.
5862 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5864 * config/aarch64/aarch64-simd.md
5865 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5866 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5867 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5868 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5869 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5870 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5871 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5872 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5873 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5874 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5876 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5879 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5880 targ_options->x_arm_arch_string is non NULL.
5882 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5884 * config/aarch64/aarch64.h
5885 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5887 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5890 * expmed.c (emit_store_flag_force): Swap if const op0
5891 and change VOIDmode to mode of op0.
5893 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5895 PR rtl-optimization/83761
5896 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5897 than bytes to mode_for_size.
5899 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5902 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5903 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5906 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5909 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5910 when in layout mode.
5911 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5912 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5915 2018-01-10 Michael Collison <michael.collison@arm.com>
5917 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5918 * config/aarch64/aarch64-option-extension.def: Add
5919 AARCH64_OPT_EXTENSION of 'fp16fml'.
5920 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5921 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5922 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5923 * config/aarch64/constraints.md (Ui7): New constraint.
5924 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5925 (VFMLA_SEL_W): Ditto.
5928 (VFMLA16_LOW): New int iterator.
5929 (VFMLA16_HIGH): Ditto.
5930 (UNSPEC_FMLAL): New unspec.
5931 (UNSPEC_FMLSL): Ditto.
5932 (UNSPEC_FMLAL2): Ditto.
5933 (UNSPEC_FMLSL2): Ditto.
5934 (f16mac): New code attribute.
5935 * config/aarch64/aarch64-simd-builtins.def
5936 (aarch64_fmlal_lowv2sf): Ditto.
5937 (aarch64_fmlsl_lowv2sf): Ditto.
5938 (aarch64_fmlalq_lowv4sf): Ditto.
5939 (aarch64_fmlslq_lowv4sf): Ditto.
5940 (aarch64_fmlal_highv2sf): Ditto.
5941 (aarch64_fmlsl_highv2sf): Ditto.
5942 (aarch64_fmlalq_highv4sf): Ditto.
5943 (aarch64_fmlslq_highv4sf): Ditto.
5944 (aarch64_fmlal_lane_lowv2sf): Ditto.
5945 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5946 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5947 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5948 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5949 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5950 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5951 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5952 (aarch64_fmlal_lane_highv2sf): Ditto.
5953 (aarch64_fmlsl_lane_highv2sf): Ditto.
5954 (aarch64_fmlal_laneq_highv2sf): Ditto.
5955 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5956 (aarch64_fmlalq_lane_highv4sf): Ditto.
5957 (aarch64_fmlsl_lane_highv4sf): Ditto.
5958 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5959 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5960 * config/aarch64/aarch64-simd.md:
5961 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5962 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5963 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5964 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5965 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5966 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5967 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5968 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5969 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5970 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5971 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5972 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5973 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5974 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5975 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5976 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5977 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5978 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5979 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5980 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5981 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5982 (vfmlsl_low_u32): Ditto.
5983 (vfmlalq_low_u32): Ditto.
5984 (vfmlslq_low_u32): Ditto.
5985 (vfmlal_high_u32): Ditto.
5986 (vfmlsl_high_u32): Ditto.
5987 (vfmlalq_high_u32): Ditto.
5988 (vfmlslq_high_u32): Ditto.
5989 (vfmlal_lane_low_u32): Ditto.
5990 (vfmlsl_lane_low_u32): Ditto.
5991 (vfmlal_laneq_low_u32): Ditto.
5992 (vfmlsl_laneq_low_u32): Ditto.
5993 (vfmlalq_lane_low_u32): Ditto.
5994 (vfmlslq_lane_low_u32): Ditto.
5995 (vfmlalq_laneq_low_u32): Ditto.
5996 (vfmlslq_laneq_low_u32): Ditto.
5997 (vfmlal_lane_high_u32): Ditto.
5998 (vfmlsl_lane_high_u32): Ditto.
5999 (vfmlal_laneq_high_u32): Ditto.
6000 (vfmlsl_laneq_high_u32): Ditto.
6001 (vfmlalq_lane_high_u32): Ditto.
6002 (vfmlslq_lane_high_u32): Ditto.
6003 (vfmlalq_laneq_high_u32): Ditto.
6004 (vfmlslq_laneq_high_u32): Ditto.
6005 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6006 (AARCH64_FL_FOR_ARCH8_4): New.
6007 (AARCH64_ISA_F16FML): New ISA flag.
6008 (TARGET_F16FML): New feature flag for fp16fml.
6009 (doc/invoke.texi): Document new fp16fml option.
6011 2018-01-10 Michael Collison <michael.collison@arm.com>
6013 * config/aarch64/aarch64-builtins.c:
6014 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6015 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6016 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6017 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6018 (AARCH64_ISA_SHA3): New ISA flag.
6019 (TARGET_SHA3): New feature flag for sha3.
6020 * config/aarch64/iterators.md (sha512_op): New int attribute.
6021 (CRYPTO_SHA512): New int iterator.
6022 (UNSPEC_SHA512H): New unspec.
6023 (UNSPEC_SHA512H2): Ditto.
6024 (UNSPEC_SHA512SU0): Ditto.
6025 (UNSPEC_SHA512SU1): Ditto.
6026 * config/aarch64/aarch64-simd-builtins.def
6027 (aarch64_crypto_sha512hqv2di): New builtin.
6028 (aarch64_crypto_sha512h2qv2di): Ditto.
6029 (aarch64_crypto_sha512su0qv2di): Ditto.
6030 (aarch64_crypto_sha512su1qv2di): Ditto.
6031 (aarch64_eor3qv8hi): Ditto.
6032 (aarch64_rax1qv2di): Ditto.
6033 (aarch64_xarqv2di): Ditto.
6034 (aarch64_bcaxqv8hi): Ditto.
6035 * config/aarch64/aarch64-simd.md:
6036 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6037 (aarch64_crypto_sha512su0qv2di): Ditto.
6038 (aarch64_crypto_sha512su1qv2di): Ditto.
6039 (aarch64_eor3qv8hi): Ditto.
6040 (aarch64_rax1qv2di): Ditto.
6041 (aarch64_xarqv2di): Ditto.
6042 (aarch64_bcaxqv8hi): Ditto.
6043 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6044 (vsha512h2q_u64): Ditto.
6045 (vsha512su0q_u64): Ditto.
6046 (vsha512su1q_u64): Ditto.
6047 (veor3q_u16): Ditto.
6048 (vrax1q_u64): Ditto.
6050 (vbcaxq_u16): Ditto.
6051 * config/arm/types.md (crypto_sha512): New type attribute.
6052 (crypto_sha3): Ditto.
6053 (doc/invoke.texi): Document new sha3 option.
6055 2018-01-10 Michael Collison <michael.collison@arm.com>
6057 * config/aarch64/aarch64-builtins.c:
6058 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6059 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6060 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6061 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6062 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6063 (AARCH64_ISA_SM4): New ISA flag.
6064 (TARGET_SM4): New feature flag for sm4.
6065 * config/aarch64/aarch64-simd-builtins.def
6066 (aarch64_sm3ss1qv4si): Ditto.
6067 (aarch64_sm3tt1aq4si): Ditto.
6068 (aarch64_sm3tt1bq4si): Ditto.
6069 (aarch64_sm3tt2aq4si): Ditto.
6070 (aarch64_sm3tt2bq4si): Ditto.
6071 (aarch64_sm3partw1qv4si): Ditto.
6072 (aarch64_sm3partw2qv4si): Ditto.
6073 (aarch64_sm4eqv4si): Ditto.
6074 (aarch64_sm4ekeyqv4si): Ditto.
6075 * config/aarch64/aarch64-simd.md:
6076 (aarch64_sm3ss1qv4si): Ditto.
6077 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6078 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6079 (aarch64_sm4eqv4si): Ditto.
6080 (aarch64_sm4ekeyqv4si): Ditto.
6081 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6082 (sm3part_op): Ditto.
6083 (CRYPTO_SM3TT): Ditto.
6084 (CRYPTO_SM3PART): Ditto.
6085 (UNSPEC_SM3SS1): New unspec.
6086 (UNSPEC_SM3TT1A): Ditto.
6087 (UNSPEC_SM3TT1B): Ditto.
6088 (UNSPEC_SM3TT2A): Ditto.
6089 (UNSPEC_SM3TT2B): Ditto.
6090 (UNSPEC_SM3PARTW1): Ditto.
6091 (UNSPEC_SM3PARTW2): Ditto.
6092 (UNSPEC_SM4E): Ditto.
6093 (UNSPEC_SM4EKEY): Ditto.
6094 * config/aarch64/constraints.md (Ui2): New constraint.
6095 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6096 * config/arm/types.md (crypto_sm3): New type attribute.
6097 (crypto_sm4): Ditto.
6098 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6099 (vsm3tt1aq_u32): Ditto.
6100 (vsm3tt1bq_u32): Ditto.
6101 (vsm3tt2aq_u32): Ditto.
6102 (vsm3tt2bq_u32): Ditto.
6103 (vsm3partw1q_u32): Ditto.
6104 (vsm3partw2q_u32): Ditto.
6105 (vsm4eq_u32): Ditto.
6106 (vsm4ekeyq_u32): Ditto.
6107 (doc/invoke.texi): Document new sm4 option.
6109 2018-01-10 Michael Collison <michael.collison@arm.com>
6111 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6112 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6113 (AARCH64_FL_FOR_ARCH8_4): New.
6114 (AARCH64_FL_V8_4): New flag.
6115 (doc/invoke.texi): Document new armv8.4-a option.
6117 2018-01-10 Michael Collison <michael.collison@arm.com>
6119 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6120 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6121 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6122 * config/aarch64/aarch64-option-extension.def: Add
6123 AARCH64_OPT_EXTENSION of 'sha2'.
6124 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6125 (crypto): Disable sha2 and aes if crypto disabled.
6126 (crypto): Enable aes and sha2 if enabled.
6127 (simd): Disable sha2 and aes if simd disabled.
6128 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6130 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6131 (TARGET_SHA2): New feature flag for sha2.
6132 (TARGET_AES): New feature flag for aes.
6133 * config/aarch64/aarch64-simd.md:
6134 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6135 conditional on TARGET_AES.
6136 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6137 (aarch64_crypto_sha1hsi): Make pattern conditional
6139 (aarch64_crypto_sha1hv4si): Ditto.
6140 (aarch64_be_crypto_sha1hv4si): Ditto.
6141 (aarch64_crypto_sha1su1v4si): Ditto.
6142 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6143 (aarch64_crypto_sha1su0v4si): Ditto.
6144 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6145 (aarch64_crypto_sha256su0v4si): Ditto.
6146 (aarch64_crypto_sha256su1v4si): Ditto.
6147 (doc/invoke.texi): Document new aes and sha2 options.
6149 2018-01-10 Martin Sebor <msebor@redhat.com>
6151 PR tree-optimization/83781
6152 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6155 2018-01-11 Martin Sebor <msebor@gmail.com>
6156 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6158 PR tree-optimization/83501
6159 PR tree-optimization/81703
6161 * tree-ssa-strlen.c (get_string_cst): Rename...
6162 (get_string_len): ...to this. Handle global constants.
6163 (handle_char_store): Adjust.
6165 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6166 Jim Wilson <jimw@sifive.com>
6168 * config/riscv/riscv-protos.h (riscv_output_return): New.
6169 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6170 (riscv_attribute_table, riscv_output_return),
6171 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6172 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6173 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6175 (riscv_expand_prologue): Add early return for naked function.
6176 (riscv_expand_epilogue): Likewise.
6177 (riscv_function_ok_for_sibcall): Return false for naked function.
6178 (riscv_set_current_function): New.
6179 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6180 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6181 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6182 * doc/extend.texi (RISC-V Function Attributes): New.
6184 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6186 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6187 check for 128-bit long double before checking TCmode.
6188 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6189 128-bit long doubles before checking TFmode or TCmode.
6190 (FLOAT128_IBM_P): Likewise.
6192 2018-01-10 Martin Sebor <msebor@redhat.com>
6194 PR tree-optimization/83671
6195 * builtins.c (c_strlen): Unconditionally return zero for the empty
6197 Use -Warray-bounds for warnings.
6198 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6199 for non-constant array indices with COMPONENT_REF, arrays of
6200 arrays, and pointers to arrays.
6201 (gimple_fold_builtin_strlen): Determine and set length range for
6202 non-constant character arrays.
6204 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6207 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6210 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6212 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6214 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6217 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6218 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6219 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6220 indexed_or_indirect_operand predicate.
6221 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6222 (*vsx_le_perm_load_v8hi): Likewise.
6223 (*vsx_le_perm_load_v16qi): Likewise.
6224 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6225 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6226 (*vsx_le_perm_store_v8hi): Likewise.
6227 (*vsx_le_perm_store_v16qi): Likewise.
6228 (eight unnamed splitters): Likewise.
6230 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6232 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6233 * config/rs6000/emmintrin.h: Likewise.
6234 * config/rs6000/mmintrin.h: Likewise.
6235 * config/rs6000/xmmintrin.h: Likewise.
6237 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6240 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6242 * tree.c (tree_nop_conversion): Return true for location wrapper
6244 (maybe_wrap_with_location): New function.
6245 (selftest::check_strip_nops): New function.
6246 (selftest::test_location_wrappers): New function.
6247 (selftest::tree_c_tests): Call it.
6248 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6249 (maybe_wrap_with_location): New decl.
6250 (EXPR_LOCATION_WRAPPER_P): New macro.
6251 (location_wrapper_p): New inline function.
6252 (tree_strip_any_location_wrapper): New inline function.
6254 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6257 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6258 stack_realign_offset for the largest alignment of stack slot
6260 (ix86_find_max_used_stack_alignment): New function.
6261 (ix86_finalize_stack_frame_flags): Use it. Set
6262 max_used_stack_alignment if we don't realign stack.
6263 * config/i386/i386.h (machine_function): Add
6264 max_used_stack_alignment.
6266 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6268 * config/arm/arm.opt (-mbranch-cost): New option.
6269 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6272 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6275 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6276 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6278 2018-01-10 Richard Biener <rguenther@suse.de>
6281 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6282 early out so it also covers the case where we have a non-NULL
6285 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6287 PR tree-optimization/83753
6288 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6289 for non-strided grouped accesses if the number of elements is 1.
6291 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6294 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6295 * i386.h (TARGET_USE_GATHER): Define.
6296 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6298 2018-01-10 Martin Liska <mliska@suse.cz>
6301 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6302 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6304 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6305 CLEANUP_NO_PARTITIONING is not set.
6307 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6309 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6310 for vectors, as a partial revert of r254296.
6311 * rtl.h (const_vec_p): Delete.
6312 (const_vec_duplicate_p): Don't test for vector CONSTs.
6313 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6314 * expmed.c (make_tree): Likewise.
6317 * common.md (E, F): Use CONSTANT_P instead of checking for
6319 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6320 checking for CONST_VECTOR.
6322 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6325 * predict.c (force_edge_cold): Handle in more sane way edges
6328 2018-01-09 Carl Love <cel@us.ibm.com>
6330 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6332 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6333 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6334 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6335 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6336 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6337 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6338 * config/rs6000/rs6000-protos.h: Add extern defition for
6339 rs6000_generate_float2_double_code.
6340 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6342 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6343 (float2_v2df): Add define_expand.
6345 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6348 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6349 op_mode in the force_to_mode call.
6351 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6353 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6354 instead of checking each element individually.
6355 (aarch64_evpc_uzp): Likewise.
6356 (aarch64_evpc_zip): Likewise.
6357 (aarch64_evpc_ext): Likewise.
6358 (aarch64_evpc_rev): Likewise.
6359 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6360 instead of checking each element individually. Return true without
6362 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6363 whether all selected elements come from the same input, instead of
6364 checking each element individually. Remove calls to gen_rtx_REG,
6365 start_sequence and end_sequence and instead assert that no rtl is
6368 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6370 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6371 order of HIGH and CONST checks.
6373 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6375 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6376 if the destination isn't an SSA_NAME.
6378 2018-01-09 Richard Biener <rguenther@suse.de>
6380 PR tree-optimization/83668
6381 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6383 (canonicalize_loop_form): ... here, renamed from ...
6384 (canonicalize_loop_closed_ssa_form): ... this and amended to
6385 swap successor edges for loop exit blocks to make us use
6386 the RPO order we need for initial schedule generation.
6388 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6390 PR tree-optimization/64811
6391 * match.pd: When optimizing comparisons with Inf, avoid
6392 introducing or losing exceptions from comparisons with NaN.
6394 2018-01-09 Martin Liska <mliska@suse.cz>
6397 * asan.c (shadow_mem_size): Add gcc_assert.
6399 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6401 Don't save registers in main().
6404 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6405 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6406 * config/avr/avr.c (avr_set_current_function): Don't error if
6407 naked, OS_task or OS_main are specified at the same time.
6408 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6410 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6412 * common/config/avr/avr-common.c (avr_option_optimization_table):
6413 Switch on -mmain-is-OS_task for optimizing compilations.
6415 2018-01-09 Richard Biener <rguenther@suse.de>
6417 PR tree-optimization/83572
6418 * graphite.c: Include cfganal.h.
6419 (graphite_transform_loops): Connect infinite loops to exit
6420 and remove fake edges at the end.
6422 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6424 * ipa-inline.c (edge_badness): Revert accidental checkin.
6426 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6429 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6430 symbols; not inline clones.
6432 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6435 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6436 hard registers. Formatting fixes.
6438 PR preprocessor/83722
6439 * gcc.c (try_generate_repro): Pass
6440 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6441 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6444 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6445 Kito Cheng <kito.cheng@gmail.com>
6447 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6448 (riscv_leaf_function_p): Delete.
6449 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6451 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6453 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6455 (do_ifelse): New function.
6456 (do_isel): New function.
6457 (do_sub3): New function.
6458 (do_add3): New function.
6459 (do_load_mask_compare): New function.
6460 (do_overlap_load_compare): New function.
6461 (expand_compare_loop): New function.
6462 (expand_block_compare): Call expand_compare_loop() when appropriate.
6463 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6465 (-mblock-compare-inline-loop-limit): New option.
6467 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6470 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6471 Reverse order of second and third operands in first alternative.
6472 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6473 of first and second elements in UNSPEC_VPERMR vector.
6474 (altivec_expand_vec_perm_le): Likewise.
6476 2018-01-08 Jeff Law <law@redhat.com>
6478 PR rtl-optimizatin/81308
6479 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6480 (process_switch): If group_case_labels makes a change, then set
6482 (pass_convert_switch::execute): If a switch is converted, then
6483 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6485 PR rtl-optimization/81308
6486 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6489 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6491 PR target/83663 - Revert r255946
6492 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6493 generation for cases where splatting a value is not useful.
6494 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6495 across a vec_duplicate and a paradoxical subreg forming a vector
6496 mode to a vec_concat.
6498 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6500 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6501 -march=armv8.3-a variants.
6502 * config/arm/t-multilib: Likewise.
6503 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6505 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6507 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6509 (cceq_ior_compare_complement): Give it a name so I can use it, and
6510 change boolean_or_operator predicate to boolean_operator so it can
6511 be used to generate a crand.
6512 (eqne): New code iterator.
6513 (bd/bd_neg): New code_attrs.
6514 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6515 a single define_insn.
6516 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6517 decrement (bdnzt/bdnzf/bdzt/bdzf).
6518 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6519 with the new names of the branch decrement patterns, and added the
6520 names of the branch decrement conditional patterns.
6522 2018-01-08 Richard Biener <rguenther@suse.de>
6524 PR tree-optimization/83563
6525 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6528 2018-01-08 Richard Biener <rguenther@suse.de>
6531 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6533 2018-01-08 Richard Biener <rguenther@suse.de>
6535 PR tree-optimization/83685
6536 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6537 references to abnormals.
6539 2018-01-08 Richard Biener <rguenther@suse.de>
6542 * dwarf2out.c (output_indirect_strings): Handle empty
6543 skeleton_debug_str_hash.
6544 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6546 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6548 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6549 (emit_store_direct): Likewise.
6550 (arc_trampoline_adjust_address): Likewise.
6551 (arc_asm_trampoline_template): New function.
6552 (arc_initialize_trampoline): Use asm_trampoline_template.
6553 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6554 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6555 * config/arc/arc.md (flush_icache): Delete pattern.
6557 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6559 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6560 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6563 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6566 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6567 by not USED_FOR_TARGET.
6568 (make_pass_resolve_sw_modes): Likewise.
6570 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6572 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6575 2018-01-08 Richard Biener <rguenther@suse.de>
6578 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6580 2018-01-08 Richard Biener <rguenther@suse.de>
6583 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6585 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6588 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6589 basic blocks with a small number of successors.
6590 (convert_control_dep_chain_into_preds): Improve handling of
6592 (dump_predicates): Split apart into...
6593 (dump_pred_chain): ...here...
6594 (dump_pred_info): ...and here.
6595 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6596 (can_chain_union_be_invalidated_p): Improve check for invalidation
6598 (uninit_uses_cannot_happen): Avoid unnecessary if
6599 convert_control_dep_chain_into_preds yielded nothing.
6601 2018-01-06 Martin Sebor <msebor@redhat.com>
6603 PR tree-optimization/83640
6604 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6605 subtracting negative offset from size.
6606 (builtin_access::overlap): Adjust offset bounds of the access to fall
6607 within the size of the object if possible.
6609 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6611 PR rtl-optimization/83699
6612 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6613 extract_bit_field_as_subreg to cases in which the extracted
6614 value is also a vector.
6616 * lra-constraints.c (process_alt_operands): Test for the equivalence
6617 substitutions when detecting a possible reload cycle.
6619 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6622 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6623 by default if flag_selective_schedling{,2}. Formatting fixes.
6625 PR rtl-optimization/83682
6626 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6627 if it has non-VECTOR_MODE element mode.
6628 (vec_duplicate_p): Likewise.
6631 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6632 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6634 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6637 * config/i386/i386-builtin.def
6638 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6639 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6640 Require also OPTION_MASK_ISA_AVX512F in addition to
6641 OPTION_MASK_ISA_GFNI.
6642 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6643 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6644 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6645 to OPTION_MASK_ISA_GFNI.
6646 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6647 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6648 OPTION_MASK_ISA_AVX512BW.
6649 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6650 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6651 addition to OPTION_MASK_ISA_GFNI.
6652 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6653 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6654 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6655 to OPTION_MASK_ISA_GFNI.
6656 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6657 a requirement for all ISAs rather than any of them with a few
6659 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6661 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6662 bitmasks to be enabled with 3 exceptions, instead of requiring any
6663 enabled ISA with lots of exceptions.
6664 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6665 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6666 Change avx512bw in isa attribute to avx512f.
6667 * config/i386/sgxintrin.h: Add license boilerplate.
6668 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6669 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6670 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6671 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6673 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6674 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6675 temporarily sse2 rather than sse if not enabled already.
6678 * config/i386/sse.md (VI248_VLBW): Rename to ...
6679 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6680 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6681 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6682 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6683 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6684 mode iterator instead of VI248_VLBW.
6686 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6688 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6689 (record_modified): Skip clobbers; add debug output.
6690 (param_change_prob): Use sreal frequencies.
6692 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6694 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6695 punt for user-aligned variables.
6697 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6699 * tree-chrec.c (chrec_contains_symbols): Return true for
6702 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6705 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6706 of (x|y) == x for BICS pattern.
6708 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6710 PR tree-optimization/83605
6711 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6712 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6715 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6717 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6718 * config/epiphany/rtems.h: New file.
6720 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6721 Uros Bizjak <ubizjak@gmail.com>
6724 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6725 QIreg_operand instead of register_operand predicate.
6726 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6727 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6728 comments instead of -fmitigate[-_]rop.
6730 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6733 * cgraphunit.c (symbol_table::compile): Switch to text_section
6734 before calling assembly_start debug hook.
6735 * run-rtl-passes.c (run_rtl_passes): Likewise.
6738 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6740 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6741 range_int_cst_p rather than !symbolic_range_p before calling
6742 extract_range_from_multiplicative_op_1.
6744 2018-01-04 Jeff Law <law@redhat.com>
6746 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6747 redundant test in assertion.
6749 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6751 * doc/rtl.texi: Document machine_mode wrapper classes.
6753 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6755 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6758 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6760 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6761 the VEC_PERM_EXPR fold to fail.
6763 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6766 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6767 to switched_sections.
6769 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6772 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6775 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6778 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6779 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6781 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6784 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6785 is BLKmode and bitpos not zero or mode change is needed.
6787 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6790 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6793 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6796 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6797 instead of MULT rtx. Update all corresponding splitters.
6799 (*ssub<modesuffix>): Ditto.
6801 (*cmp_sadd_di): Update split patterns.
6802 (*cmp_sadd_si): Ditto.
6803 (*cmp_sadd_sidi): Ditto.
6804 (*cmp_ssub_di): Ditto.
6805 (*cmp_ssub_si): Ditto.
6806 (*cmp_ssub_sidi): Ditto.
6807 * config/alpha/predicates.md (const23_operand): New predicate.
6808 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6809 Look for ASHIFT, not MULT inner operand.
6810 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6812 2018-01-04 Martin Liska <mliska@suse.cz>
6814 PR gcov-profile/83669
6815 * gcov.c (output_intermediate_file): Add version to intermediate
6817 * doc/gcov.texi: Document new field 'version' in intermediate
6818 file format. Fix location of '-k' option of gcov command.
6820 2018-01-04 Martin Liska <mliska@suse.cz>
6823 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6825 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6827 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6829 2018-01-03 Martin Sebor <msebor@redhat.com>
6831 PR tree-optimization/83655
6832 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6833 checking calls with invalid arguments.
6835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6837 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6838 (vectorizable_mask_load_store): Delete.
6839 (vectorizable_call): Return false for masked loads and stores.
6840 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6841 instead of gimple_assign_rhs1.
6842 (vectorizable_load): Handle IFN_MASK_LOAD.
6843 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6847 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6849 (vectorizable_mask_load_store): ...here.
6850 (vectorizable_load): ...and here.
6852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6854 * tree-vect-stmts.c (vect_build_all_ones_mask)
6855 (vect_build_zero_merge_argument): New functions, split out from...
6856 (vectorizable_load): ...here.
6858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6860 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6862 (vectorizable_mask_load_store): ...here.
6863 (vectorizable_store): ...and here.
6865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6867 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6869 (vectorizable_mask_load_store): ...here.
6871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6873 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6874 (vect_model_store_cost): Take a vec_load_store_type instead of a
6876 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6877 (vect_model_store_cost): Take a vec_load_store_type instead of a
6879 (vectorizable_mask_load_store): Update accordingly.
6880 (vectorizable_store): Likewise.
6881 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6885 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6886 IFN_MASK_LOAD calls here rather than...
6887 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6890 Alan Hayward <alan.hayward@arm.com>
6891 David Sherwood <david.sherwood@arm.com>
6893 * expmed.c (extract_bit_field_1): For vector extracts,
6894 fall back to extract_bit_field_as_subreg if vec_extract
6897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6898 Alan Hayward <alan.hayward@arm.com>
6899 David Sherwood <david.sherwood@arm.com>
6901 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6902 they are variable or constant sized.
6903 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6904 slots for constant-sized data.
6906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6907 Alan Hayward <alan.hayward@arm.com>
6908 David Sherwood <david.sherwood@arm.com>
6910 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6911 handling COND_EXPRs with boolean comparisons, try to find a better
6912 basis for the mask type than the boolean itself.
6914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6916 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6917 is calculated and how it can be overridden.
6918 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6919 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6921 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6925 Alan Hayward <alan.hayward@arm.com>
6926 David Sherwood <david.sherwood@arm.com>
6928 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6929 Remove the mode argument.
6930 (aarch64_simd_valid_immediate): Remove the mode and inverse
6932 * config/aarch64/iterators.md (bitsize): New iterator.
6933 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6934 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6935 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6936 aarch64_simd_valid_immediate.
6937 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6938 (aarch64_reg_or_bic_imm): Likewise.
6939 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6940 with an insn_type enum and msl with a modifier_type enum.
6941 Replace element_width with a scalar_mode. Change the shift
6942 to unsigned int. Add constructors for scalar_float_mode and
6943 scalar_int_mode elements.
6944 (aarch64_vect_float_const_representable_p): Delete.
6945 (aarch64_can_const_movi_rtx_p)
6946 (aarch64_simd_scalar_immediate_valid_for_move)
6947 (aarch64_simd_make_constant): Update call to
6948 aarch64_simd_valid_immediate.
6949 (aarch64_advsimd_valid_immediate_hs): New function.
6950 (aarch64_advsimd_valid_immediate): Likewise.
6951 (aarch64_simd_valid_immediate): Remove mode and inverse
6952 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6953 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6954 and aarch64_float_const_representable_p on the result.
6955 (aarch64_output_simd_mov_immediate): Remove mode argument.
6956 Update call to aarch64_simd_valid_immediate and use of
6957 simd_immediate_info.
6958 (aarch64_output_scalar_simd_mov_immediate): Update call
6961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6962 Alan Hayward <alan.hayward@arm.com>
6963 David Sherwood <david.sherwood@arm.com>
6965 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6966 (mode_nunits): Likewise CONST_MODE_NUNITS.
6967 * machmode.def (ADJUST_NUNITS): Document.
6968 * genmodes.c (mode_data::need_nunits_adj): New field.
6969 (blank_mode): Update accordingly.
6970 (adj_nunits): New variable.
6971 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6973 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6974 listed in adj_nunits.
6975 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6976 listed in adj_nunits. Don't emit case statements for such modes.
6977 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6978 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6979 nothing if adj_nunits is nonnull.
6980 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6981 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6982 (emit_mode_fbit): Update use of print_maybe_const_decl.
6983 (emit_move_size): Likewise. Treat the array as non-const
6985 (emit_mode_adjustments): Handle adj_nunits.
6987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6989 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6990 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6991 (VECTOR_MODES): Use it.
6992 (make_vector_modes): Take the prefix as an argument.
6994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6995 Alan Hayward <alan.hayward@arm.com>
6996 David Sherwood <david.sherwood@arm.com>
6998 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6999 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7000 for MODE_VECTOR_BOOL.
7001 * machmode.def (VECTOR_BOOL_MODE): Document.
7002 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7003 (make_vector_bool_mode): New function.
7004 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7006 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7007 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7009 * stor-layout.c (int_mode_for_mode): Likewise.
7010 * tree.c (build_vector_type_for_mode): Likewise.
7011 * varasm.c (output_constant_pool_2): Likewise.
7012 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7013 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7014 for MODE_VECTOR_BOOL.
7015 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7016 of mode class checks.
7017 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7018 instead of a list of mode class checks.
7019 (expand_vector_scalar_condition): Likewise.
7020 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7023 Alan Hayward <alan.hayward@arm.com>
7024 David Sherwood <david.sherwood@arm.com>
7026 * machmode.h (mode_size): Change from unsigned short to
7028 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7029 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7030 or if measurement_type is not polynomial.
7031 (fixed_size_mode::includes_p): Check for constant-sized modes.
7032 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7033 return a poly_uint16 rather than an unsigned short.
7034 (emit_mode_size): Change the type of mode_size from unsigned short
7035 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7036 (emit_mode_adjustments): Cope with polynomial vector sizes.
7037 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7039 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7041 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7042 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7043 * caller-save.c (setup_save_areas): Likewise.
7044 (replace_reg_with_saved_mem): Likewise.
7045 * calls.c (emit_library_call_value_1): Likewise.
7046 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7047 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7048 (gen_lowpart_for_combine): Likewise.
7049 * convert.c (convert_to_integer_1): Likewise.
7050 * cse.c (equiv_constant, cse_insn): Likewise.
7051 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7052 (cselib_subst_to_values): Likewise.
7053 * dce.c (word_dce_process_block): Likewise.
7054 * df-problems.c (df_word_lr_mark_ref): Likewise.
7055 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7056 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7057 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7058 (rtl_for_decl_location): Likewise.
7059 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7060 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7061 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7062 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7063 (expand_expr_real_1): Likewise.
7064 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7065 (pad_below): Likewise.
7066 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7067 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7068 * ira.c (get_subreg_tracking_sizes): Likewise.
7069 * ira-build.c (ira_create_allocno_objects): Likewise.
7070 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7071 (ira_sort_regnos_for_alter_reg): Likewise.
7072 * ira-costs.c (record_operand_costs): Likewise.
7073 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7074 (resolve_simple_move): Likewise.
7075 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7076 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7077 (lra_constraints): Likewise.
7078 (CONST_POOL_OK_P): Reject variable-sized modes.
7079 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7080 (add_pseudo_to_slot, lra_spill): Likewise.
7081 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7082 * optabs-query.c (get_best_extraction_insn): Likewise.
7083 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7084 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7085 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7086 * recog.c (offsettable_address_addr_space_p): Likewise.
7087 * regcprop.c (maybe_mode_change): Likewise.
7088 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7089 * regrename.c (build_def_use): Likewise.
7090 * regstat.c (dump_reg_info): Likewise.
7091 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7092 (find_reloads, find_reloads_subreg_address): Likewise.
7093 * reload1.c (eliminate_regs_1): Likewise.
7094 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7095 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7096 (simplify_binary_operation_1, simplify_subreg): Likewise.
7097 * targhooks.c (default_function_arg_padding): Likewise.
7098 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7099 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7100 (verify_gimple_assign_ternary): Likewise.
7101 * tree-inline.c (estimate_move_cost): Likewise.
7102 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7103 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7104 (get_address_cost_ainc): Likewise.
7105 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7106 (vect_supportable_dr_alignment): Likewise.
7107 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7108 (vectorizable_reduction): Likewise.
7109 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7110 (vectorizable_operation, vectorizable_load): Likewise.
7111 * tree.c (build_same_sized_truth_vector_type): Likewise.
7112 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7113 * var-tracking.c (emit_note_insn_var_location): Likewise.
7114 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7115 (ADDR_VEC_ALIGN): Likewise.
7117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7118 Alan Hayward <alan.hayward@arm.com>
7119 David Sherwood <david.sherwood@arm.com>
7121 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7123 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7124 or if measurement_type is polynomial.
7125 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7126 * combine.c (make_extraction): Likewise.
7127 * dse.c (find_shift_sequence): Likewise.
7128 * dwarf2out.c (mem_loc_descriptor): Likewise.
7129 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7130 (extract_bit_field, extract_low_bits): Likewise.
7131 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7132 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7133 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7134 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7135 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7136 * reload.c (find_reloads): Likewise.
7137 * reload1.c (alter_reg): Likewise.
7138 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7139 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7140 * tree-if-conv.c (predicate_mem_writes): Likewise.
7141 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7142 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7143 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7144 * valtrack.c (dead_debug_insert_temp): Likewise.
7145 * varasm.c (mergeable_constant_section): Likewise.
7146 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7149 Alan Hayward <alan.hayward@arm.com>
7150 David Sherwood <david.sherwood@arm.com>
7152 * expr.c (expand_assignment): Cope with polynomial mode sizes
7153 when assigning to a CONCAT.
7155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7156 Alan Hayward <alan.hayward@arm.com>
7157 David Sherwood <david.sherwood@arm.com>
7159 * machmode.h (mode_precision): Change from unsigned short to
7161 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7163 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7164 or if measurement_type is not polynomial.
7165 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7166 in which the mode is already known to be a scalar_int_mode.
7167 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7168 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7170 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7171 for GET_MODE_PRECISION.
7172 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7173 for GET_MODE_PRECISION.
7174 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7176 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7177 (expand_field_assignment, make_extraction): Likewise.
7178 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7179 (get_last_value): Likewise.
7180 * convert.c (convert_to_integer_1): Likewise.
7181 * cse.c (cse_insn): Likewise.
7182 * expr.c (expand_expr_real_1): Likewise.
7183 * lra-constraints.c (simplify_operand_subreg): Likewise.
7184 * optabs-query.c (can_atomic_load_p): Likewise.
7185 * optabs.c (expand_atomic_load): Likewise.
7186 (expand_atomic_store): Likewise.
7187 * ree.c (combine_reaching_defs): Likewise.
7188 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7189 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7190 * tree.h (type_has_mode_precision_p): Likewise.
7191 * ubsan.c (instrument_si_overflow): Likewise.
7193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7194 Alan Hayward <alan.hayward@arm.com>
7195 David Sherwood <david.sherwood@arm.com>
7197 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7198 polynomial numbers of units.
7199 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7200 (valid_vector_subparts_p): New function.
7201 (build_vector_type): Remove temporary shim and take the number
7202 of units as a poly_uint64 rather than an int.
7203 (build_opaque_vector_type): Take the number of units as a
7204 poly_uint64 rather than an int.
7205 * tree.c (build_vector_from_ctor): Handle polynomial
7206 TYPE_VECTOR_SUBPARTS.
7207 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7208 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7209 (build_vector_from_val): If the number of units is variable,
7210 use build_vec_duplicate_cst for constant operands and
7211 VEC_DUPLICATE_EXPR otherwise.
7212 (make_vector_type): Remove temporary is_constant ().
7213 (build_vector_type, build_opaque_vector_type): Take the number of
7214 units as a poly_uint64 rather than an int.
7215 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7217 * cfgexpand.c (expand_debug_expr): Likewise.
7218 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7219 (store_constructor, expand_expr_real_1): Likewise.
7220 (const_scalar_mask_from_tree): Likewise.
7221 * fold-const-call.c (fold_const_reduction): Likewise.
7222 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7223 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7224 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7225 (fold_relational_const): Likewise.
7226 (native_interpret_vector): Likewise. Change the size from an
7227 int to an unsigned int.
7228 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7229 TYPE_VECTOR_SUBPARTS.
7230 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7231 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7232 duplicating a non-constant operand into a variable-length vector.
7233 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7234 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7235 * ipa-icf.c (sem_variable::equals): Likewise.
7236 * match.pd: Likewise.
7237 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7238 * print-tree.c (print_node): Likewise.
7239 * stor-layout.c (layout_type): Likewise.
7240 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7241 * tree-cfg.c (verify_gimple_comparison): Likewise.
7242 (verify_gimple_assign_binary): Likewise.
7243 (verify_gimple_assign_ternary): Likewise.
7244 (verify_gimple_assign_single): Likewise.
7245 * tree-pretty-print.c (dump_generic_node): Likewise.
7246 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7247 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7248 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7249 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7250 (vect_shift_permute_load_chain): Likewise.
7251 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7252 (expand_vector_condition, optimize_vector_constructor): Likewise.
7253 (lower_vec_perm, get_compute_type): Likewise.
7254 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7255 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7256 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7257 (vect_recog_mask_conversion_pattern): Likewise.
7258 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7259 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7260 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7261 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7262 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7263 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7264 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7265 (supportable_widening_operation): Likewise.
7266 (supportable_narrowing_operation): Likewise.
7267 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7269 * varasm.c (output_constant): Likewise.
7271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7272 Alan Hayward <alan.hayward@arm.com>
7273 David Sherwood <david.sherwood@arm.com>
7275 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7276 so that both the length == 3 and length != 3 cases set up their
7277 own permute vectors. Add comments explaining why we know the
7278 number of elements is constant.
7279 (vect_permute_load_chain): Likewise.
7281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7282 Alan Hayward <alan.hayward@arm.com>
7283 David Sherwood <david.sherwood@arm.com>
7285 * machmode.h (mode_nunits): Change from unsigned char to
7287 (ONLY_FIXED_SIZE_MODES): New macro.
7288 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7289 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7290 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7292 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7293 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7294 or if measurement_type is not polynomial.
7295 * genmodes.c (ZERO_COEFFS): New macro.
7296 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7298 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7299 Use ZERO_COEFFS when emitting initializers.
7300 * data-streamer.h (bp_pack_poly_value): New function.
7301 (bp_unpack_poly_value): Likewise.
7302 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7303 for GET_MODE_NUNITS.
7304 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7305 for GET_MODE_NUNITS.
7306 * tree.c (make_vector_type): Remove temporary shim and make
7307 the real function take the number of units as a poly_uint64
7309 (build_vector_type_for_mode): Handle polynomial nunits.
7310 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7311 * emit-rtl.c (const_vec_series_p_1): Likewise.
7312 (gen_rtx_CONST_VECTOR): Likewise.
7313 * fold-const.c (test_vec_duplicate_folding): Likewise.
7314 * genrecog.c (validate_pattern): Likewise.
7315 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7316 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7317 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7318 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7319 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7320 * rtlanal.c (subreg_get_info): Likewise.
7321 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7322 (vect_grouped_load_supported): Likewise.
7323 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7324 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7325 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7326 (simplify_const_unary_operation, simplify_binary_operation_1)
7327 (simplify_const_binary_operation, simplify_ternary_operation)
7328 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7329 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7330 instead of CONST_VECTOR_NUNITS.
7331 * varasm.c (output_constant_pool_2): Likewise.
7332 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7333 explicit-encoded elements in the XVEC for variable-length vectors.
7335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7337 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7340 Alan Hayward <alan.hayward@arm.com>
7341 David Sherwood <david.sherwood@arm.com>
7343 * coretypes.h (fixed_size_mode): Declare.
7344 (fixed_size_mode_pod): New typedef.
7345 * builtins.h (target_builtins::x_apply_args_mode)
7346 (target_builtins::x_apply_result_mode): Change type to
7347 fixed_size_mode_pod.
7348 * builtins.c (apply_args_size, apply_result_size, result_vector)
7349 (expand_builtin_apply_args_1, expand_builtin_apply)
7350 (expand_builtin_return): Update accordingly.
7352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7354 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7355 * cselib.c (cselib_hash_rtx): Likewise.
7356 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7357 CONST_VECTOR encoding.
7359 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7360 Jeff Law <law@redhat.com>
7363 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7364 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7365 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7366 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7369 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7370 explicitly probe *sp in a noreturn function if there were any callee
7371 register saves or frame pointer is needed.
7373 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7376 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7377 BLKmode for ternary, binary or unary expressions.
7380 * var-tracking.c (delete_vta_debug_insn): New inline function.
7381 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7382 insns from get_insns () to NULL instead of each bb separately.
7383 Use delete_vta_debug_insn. No longer static.
7384 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7385 delete_vta_debug_insns callers.
7386 * rtl.h (delete_vta_debug_insns): Declare.
7387 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7388 instead of variable_tracking_main.
7390 2018-01-03 Martin Sebor <msebor@redhat.com>
7392 PR tree-optimization/83603
7393 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7394 arguments past the endof the argument list in functions declared
7395 without a prototype.
7396 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7397 Avoid checking when arguments are null.
7399 2018-01-03 Martin Sebor <msebor@redhat.com>
7402 * doc/extend.texi (attribute const): Fix a typo.
7403 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7404 issuing -Wsuggest-attribute for void functions.
7406 2018-01-03 Martin Sebor <msebor@redhat.com>
7408 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7409 offset_int::from instead of wide_int::to_shwi.
7410 (maybe_diag_overlap): Remove assertion.
7411 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7412 * gimple-ssa-sprintf.c (format_directive): Same.
7413 (parse_directive): Same.
7414 (sprintf_dom_walker::compute_format_length): Same.
7415 (try_substitute_return_value): Same.
7417 2018-01-03 Jeff Law <law@redhat.com>
7420 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7421 non-constant residual for zero at runtime and avoid probing in
7422 that case. Reorganize code for trailing problem to mirror handling
7425 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7427 PR tree-optimization/83501
7428 * tree-ssa-strlen.c (get_string_cst): New.
7429 (handle_char_store): Call get_string_cst.
7431 2018-01-03 Martin Liska <mliska@suse.cz>
7433 PR tree-optimization/83593
7434 * tree-ssa-strlen.c: Include tree-cfg.h.
7435 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7436 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7437 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7439 (strlen_dom_walker::before_dom_children): Call
7440 gimple_purge_dead_eh_edges. Dump tranformation with details
7442 (strlen_dom_walker::before_dom_children): Update call by adding
7443 new argument cleanup_eh.
7444 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7446 2018-01-03 Martin Liska <mliska@suse.cz>
7449 * cif-code.def (VARIADIC_THUNK): New enum value.
7450 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7453 2018-01-03 Jan Beulich <jbeulich@suse.com>
7455 * sse.md (mov<mode>_internal): Tighten condition for when to use
7456 vmovdqu<ssescalarsize> for TI and OI modes.
7458 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7460 Update copyright years.
7462 2018-01-03 Martin Liska <mliska@suse.cz>
7465 * ipa-visibility.c (function_and_variable_visibility): Skip
7466 functions with noipa attribure.
7468 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7470 * gcc.c (process_command): Update copyright notice dates.
7471 * gcov-dump.c (print_version): Ditto.
7472 * gcov.c (print_version): Ditto.
7473 * gcov-tool.c (print_version): Ditto.
7474 * gengtype.c (create_file): Ditto.
7475 * doc/cpp.texi: Bump @copying's copyright year.
7476 * doc/cppinternals.texi: Ditto.
7477 * doc/gcc.texi: Ditto.
7478 * doc/gccint.texi: Ditto.
7479 * doc/gcov.texi: Ditto.
7480 * doc/install.texi: Ditto.
7481 * doc/invoke.texi: Ditto.
7483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7485 * vector-builder.h (vector_builder::m_full_nelts): Change from
7486 unsigned int to poly_uint64.
7487 (vector_builder::full_nelts): Update prototype accordingly.
7488 (vector_builder::new_vector): Likewise.
7489 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7490 (vector_builder::operator ==): Likewise.
7491 (vector_builder::finalize): Likewise.
7492 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7493 Take the number of elements as a poly_uint64 rather than an
7495 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7496 from unsigned int to poly_uint64.
7497 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7498 (vec_perm_indices::new_vector): Likewise.
7499 (vec_perm_indices::length): Likewise.
7500 (vec_perm_indices::nelts_per_input): Likewise.
7501 (vec_perm_indices::input_nelts): Likewise.
7502 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7503 number of elements per input as a poly_uint64 rather than an
7504 unsigned int. Use the original encoding for variable-length
7505 vectors, rather than clamping each individual element.
7506 For the second and subsequent elements in each pattern,
7507 clamp the step and base before clamping their sum.
7508 (vec_perm_indices::series_p): Handle polynomial element counts.
7509 (vec_perm_indices::all_in_range_p): Likewise.
7510 (vec_perm_indices_to_tree): Likewise.
7511 (vec_perm_indices_to_rtx): Likewise.
7512 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7513 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7514 (tree_vector_builder::new_binary_operation): Handle polynomial
7515 element counts. Return false if we need to know the number
7516 of elements at compile time.
7517 * fold-const.c (fold_vec_perm): Punt if the number of elements
7518 isn't known at compile time.
7520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7522 * vec-perm-indices.h (vec_perm_builder): Change element type
7523 from HOST_WIDE_INT to poly_int64.
7524 (vec_perm_indices::element_type): Update accordingly.
7525 (vec_perm_indices::clamp): Handle polynomial element_types.
7526 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7527 (vec_perm_indices::all_in_range_p): Likewise.
7528 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7530 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7531 polynomial vec_perm_indices element types.
7532 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7533 * fold-const.c (fold_vec_perm): Likewise.
7534 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7535 * tree-vect-generic.c (lower_vec_perm): Likewise.
7536 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7537 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7538 element type to HOST_WIDE_INT.
7540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7541 Alan Hayward <alan.hayward@arm.com>
7542 David Sherwood <david.sherwood@arm.com>
7544 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7545 rather than an int. Use plus_constant.
7546 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7547 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7550 Alan Hayward <alan.hayward@arm.com>
7551 David Sherwood <david.sherwood@arm.com>
7553 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7554 a HOST_WIDE_INT to a poly_int64.
7556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7557 Alan Hayward <alan.hayward@arm.com>
7558 David Sherwood <david.sherwood@arm.com>
7560 * calls.c (load_register_parameters): Cope with polynomial
7561 mode sizes. Require a constant size for BLKmode parameters
7562 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7563 forces a parameter to be padded at the lsb end in order to
7564 fill a complete number of words, require the parameter size
7565 to be ordered wrt UNITS_PER_WORD.
7567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7568 Alan Hayward <alan.hayward@arm.com>
7569 David Sherwood <david.sherwood@arm.com>
7571 * reload1.c (spill_stack_slot_width): Change element type
7572 from unsigned int to poly_uint64_pod.
7573 (alter_reg): Treat mode sizes as polynomial.
7575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7576 Alan Hayward <alan.hayward@arm.com>
7577 David Sherwood <david.sherwood@arm.com>
7579 * reload.c (complex_word_subreg_p): New function.
7580 (reload_inner_reg_of_subreg, push_reload): Use it.
7582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7583 Alan Hayward <alan.hayward@arm.com>
7584 David Sherwood <david.sherwood@arm.com>
7586 * lra-constraints.c (process_alt_operands): Reject matched
7587 operands whose sizes aren't ordered.
7588 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7595 that the mode size is in the set {1, 2, 4, 8, 16}.
7597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7598 Alan Hayward <alan.hayward@arm.com>
7599 David Sherwood <david.sherwood@arm.com>
7601 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7602 Use plus_constant instead of gen_rtx_PLUS.
7604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7605 Alan Hayward <alan.hayward@arm.com>
7606 David Sherwood <david.sherwood@arm.com>
7608 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7609 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7610 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7611 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7612 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7613 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7614 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7615 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7616 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7617 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7619 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7620 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7621 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7622 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7623 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7624 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7625 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7626 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7627 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7628 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7630 * expr.c (emit_move_resolve_push): Treat the input and result
7631 of PUSH_ROUNDING as a poly_int64.
7632 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7633 (emit_push_insn): Likewise.
7634 * lra-eliminations.c (mark_not_eliminable): Likewise.
7635 * recog.c (push_operand): Likewise.
7636 * reload1.c (elimination_effects): Likewise.
7637 * rtlanal.c (nonzero_bits1): Likewise.
7638 * calls.c (store_one_arg): Likewise. Require the padding to be
7639 known at compile time.
7641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7642 Alan Hayward <alan.hayward@arm.com>
7643 David Sherwood <david.sherwood@arm.com>
7645 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7646 Use plus_constant instead of gen_rtx_PLUS.
7648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7649 Alan Hayward <alan.hayward@arm.com>
7650 David Sherwood <david.sherwood@arm.com>
7652 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7656 Alan Hayward <alan.hayward@arm.com>
7657 David Sherwood <david.sherwood@arm.com>
7659 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7660 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7661 via stack temporaries. Treat the mode size as polynomial too.
7663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7664 Alan Hayward <alan.hayward@arm.com>
7665 David Sherwood <david.sherwood@arm.com>
7667 * expr.c (expand_expr_real_2): When handling conversions involving
7668 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7669 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7670 as a poly_uint64 too.
7672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7673 Alan Hayward <alan.hayward@arm.com>
7674 David Sherwood <david.sherwood@arm.com>
7676 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7679 Alan Hayward <alan.hayward@arm.com>
7680 David Sherwood <david.sherwood@arm.com>
7682 * combine.c (can_change_dest_mode): Handle polynomial
7683 REGMODE_NATURAL_SIZE.
7684 * expmed.c (store_bit_field_1): Likewise.
7685 * expr.c (store_constructor): Likewise.
7686 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7687 and polynomial REGMODE_NATURAL_SIZE.
7688 (gen_lowpart_common): Likewise.
7689 * reginfo.c (record_subregs_of_mode): Likewise.
7690 * rtlanal.c (read_modify_subreg_p): Likewise.
7692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7693 Alan Hayward <alan.hayward@arm.com>
7694 David Sherwood <david.sherwood@arm.com>
7696 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7697 numbers of elements.
7699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7700 Alan Hayward <alan.hayward@arm.com>
7701 David Sherwood <david.sherwood@arm.com>
7703 * match.pd: Cope with polynomial numbers of vector elements.
7705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7706 Alan Hayward <alan.hayward@arm.com>
7707 David Sherwood <david.sherwood@arm.com>
7709 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7710 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
7717 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7718 (ipa_simd_modify_function_body): Likewise.
7720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7721 Alan Hayward <alan.hayward@arm.com>
7722 David Sherwood <david.sherwood@arm.com>
7724 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7725 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7726 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7727 (expand_vector_condition, vector_element): Likewise.
7728 (subparts_gt): New function.
7729 (get_compute_type): Use subparts_gt.
7730 (count_type_subparts): Delete.
7731 (expand_vector_operations_1): Use subparts_gt instead of
7732 count_type_subparts.
7734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7735 Alan Hayward <alan.hayward@arm.com>
7736 David Sherwood <david.sherwood@arm.com>
7738 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7739 (vect_compile_time_alias): ...this new function. Do the calculation
7740 on poly_ints rather than trees.
7741 (vect_prune_runtime_alias_test_list): Update call accordingly.
7743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7744 Alan Hayward <alan.hayward@arm.com>
7745 David Sherwood <david.sherwood@arm.com>
7747 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7749 (vect_schedule_slp_instance): Likewise.
7751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7752 Alan Hayward <alan.hayward@arm.com>
7753 David Sherwood <david.sherwood@arm.com>
7755 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7756 constant and extern definitions for variable-length vectors.
7757 (vect_get_constant_vectors): Note that the number of units
7758 is known to be constant.
7760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7761 Alan Hayward <alan.hayward@arm.com>
7762 David Sherwood <david.sherwood@arm.com>
7764 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7765 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
7773 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7776 Alan Hayward <alan.hayward@arm.com>
7777 David Sherwood <david.sherwood@arm.com>
7779 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7780 vectors as polynomial. Use build_index_vector for
7783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7784 Alan Hayward <alan.hayward@arm.com>
7785 David Sherwood <david.sherwood@arm.com>
7787 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7788 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7789 for variable-length vectors.
7790 (vectorizable_mask_load_store): Treat the number of units as
7791 polynomial, asserting that it is constant if the condition has
7792 already been enforced.
7793 (vectorizable_store, vectorizable_load): Likewise.
7795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7796 Alan Hayward <alan.hayward@arm.com>
7797 David Sherwood <david.sherwood@arm.com>
7799 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7800 of units as polynomial. Punt if we can't tell at compile time
7801 which vector contains the final result.
7803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7804 Alan Hayward <alan.hayward@arm.com>
7805 David Sherwood <david.sherwood@arm.com>
7807 * tree-vect-loop.c (vectorizable_induction): Treat the number
7808 of units as polynomial. Punt on SLP inductions. Use an integer
7809 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7810 cast of such a series for variable-length floating-point
7813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7814 Alan Hayward <alan.hayward@arm.com>
7815 David Sherwood <david.sherwood@arm.com>
7817 * tree.h (build_index_vector): Declare.
7818 * tree.c (build_index_vector): New function.
7819 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7820 of units as polynomial, forcibly converting it to a constant if
7821 vectorizable_reduction has already enforced the condition.
7822 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7823 to create a {1,2,3,...} vector.
7824 (vectorizable_reduction): Treat the number of units as polynomial.
7825 Choose vectype_in based on the largest scalar element size rather
7826 than the smallest number of units. Enforce the restrictions
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-data-refs.c (vector_alignment_reachable_p): Treat the
7834 number of units as polynomial.
7836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7837 Alan Hayward <alan.hayward@arm.com>
7838 David Sherwood <david.sherwood@arm.com>
7840 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7841 * target.def (autovectorize_vector_sizes): Return the vector sizes
7842 by pointer, using vector_sizes rather than a bitmask.
7843 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7844 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7845 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7847 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7848 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7849 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7850 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7851 * omp-general.c (omp_max_vf): Likewise.
7852 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7853 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7854 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7855 * tree-vect-slp.c (vect_slp_bb): Likewise.
7856 * doc/tm.texi: Regenerate.
7857 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7859 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7860 the vector size as a poly_uint64 rather than an unsigned int.
7861 (current_vector_size): Change from an unsigned int to a poly_uint64.
7862 (get_vectype_for_scalar_type): Update accordingly.
7863 * tree.h (build_truth_vector_type): Take the size and number of
7864 units as a poly_uint64 rather than an unsigned int.
7865 (build_vector_type): Add a temporary overload that takes
7866 the number of units as a poly_uint64 rather than an unsigned int.
7867 * tree.c (make_vector_type): Likewise.
7868 (build_truth_vector_type): Take the number of units as a poly_uint64
7869 rather than an unsigned int.
7871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7872 Alan Hayward <alan.hayward@arm.com>
7873 David Sherwood <david.sherwood@arm.com>
7875 * target.def (get_mask_mode): Take the number of units and length
7876 as poly_uint64s rather than unsigned ints.
7877 * targhooks.h (default_get_mask_mode): Update accordingly.
7878 * targhooks.c (default_get_mask_mode): Likewise.
7879 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7880 * doc/tm.texi: Regenerate.
7882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7883 Alan Hayward <alan.hayward@arm.com>
7884 David Sherwood <david.sherwood@arm.com>
7886 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7887 * omp-general.c (omp_max_vf): Likewise.
7888 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7889 (expand_omp_simd): Handle polynomial safelen.
7890 * omp-low.c (omplow_simd_context): Add a default constructor.
7891 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7892 (lower_rec_simd_input_clauses): Update accordingly.
7893 (lower_rec_input_clauses): Likewise.
7895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7896 Alan Hayward <alan.hayward@arm.com>
7897 David Sherwood <david.sherwood@arm.com>
7899 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7900 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7901 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7902 (vect_analyze_slp_cost): Likewise.
7903 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7904 (vect_model_load_cost): Likewise.
7906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7907 Alan Hayward <alan.hayward@arm.com>
7908 David Sherwood <david.sherwood@arm.com>
7910 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7911 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7912 from an unsigned int * to a poly_uint64_pod *.
7913 (calculate_unrolling_factor): New function.
7914 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7921 from an unsigned int to a poly_uint64.
7922 (_loop_vec_info::slp_unrolling_factor): Likewise.
7923 (_loop_vec_info::vectorization_factor): Change from an int
7925 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7926 (vect_get_num_vectors): New function.
7927 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7928 (vect_get_num_copies): Use vect_get_num_vectors.
7929 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7930 to an unsigned int *.
7931 (vect_analyze_data_refs): Change min_vf from an int * to a
7933 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7934 than an unsigned HOST_WIDE_INT.
7935 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7936 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7937 to an unsigned int *.
7938 (vect_analyze_data_ref_dependences): Likewise.
7939 (vect_compute_data_ref_alignment): Handle polynomial vf.
7940 (vect_enhance_data_refs_alignment): Likewise.
7941 (vect_prune_runtime_alias_test_list): Likewise.
7942 (vect_shift_permute_load_chain): Likewise.
7943 (vect_supportable_dr_alignment): Likewise.
7944 (dependence_distance_ge_vf): Take the vectorization factor as a
7945 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7946 (vect_analyze_data_refs): Change min_vf from an int * to a
7948 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7949 vfm1 as a poly_uint64 rather than an int. Make the same change
7950 for the returned bound_scalar.
7951 (vect_gen_vector_loop_niters): Handle polynomial vf.
7952 (vect_do_peeling): Likewise. Update call to
7953 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7954 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7956 * tree-vect-loop.c (vect_determine_vectorization_factor)
7957 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7958 (vect_get_known_peeling_cost): Likewise.
7959 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7960 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7961 (vect_transform_loop): Likewise. Use the lowest possible VF when
7962 updating the upper bounds of the loop.
7963 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7965 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7966 polynomial unroll factors.
7967 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7968 (vect_make_slp_decision): Likewise.
7969 (vect_supported_load_permutation_p): Likewise, and polynomial
7971 (vect_analyze_slp_cost): Handle polynomial vf.
7972 (vect_slp_analyze_node_operations): Likewise.
7973 (vect_slp_analyze_bb_1): Likewise.
7974 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7975 than an unsigned HOST_WIDE_INT.
7976 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7977 (vectorizable_load): Handle polynomial vf.
7978 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7980 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7983 Alan Hayward <alan.hayward@arm.com>
7984 David Sherwood <david.sherwood@arm.com>
7986 * match.pd: Handle bit operations involving three constants
7987 and try to fold one pair.
7989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7991 * tree-vect-loop-manip.c: Include gimple-fold.h.
7992 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7993 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7994 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7995 Add a path that uses a step of VF instead of 1, but disable it
7997 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7998 and niters_no_overflow parameters. Update calls to
7999 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8000 Create a new SSA name if the latter choses to use a ste other
8001 than zero, and return it via niters_vector_mult_vf_var.
8002 * tree-vect-loop.c (vect_transform_loop): Update calls to
8003 vect_do_peeling, vect_gen_vector_loop_niters and
8004 slpeel_make_loop_iterate_ntimes.
8005 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8006 (vect_gen_vector_loop_niters): Update declarations after above changes.
8008 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8010 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8011 128-bit round to integer instructions.
8012 (ceil<mode>2): Likewise.
8013 (btrunc<mode>2): Likewise.
8014 (round<mode>2): Likewise.
8016 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8018 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8019 unaligned VSX load/store on P8/P9.
8020 (expand_block_clear): Allow the use of unaligned VSX
8021 load/store on P8/P9.
8023 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8025 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8027 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8028 swap associated with both a load and a store.
8030 2018-01-02 Andrew Waterman <andrew@sifive.com>
8032 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8033 * config/riscv/riscv.md (clear_cache): Use it.
8035 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8037 * web.c: Remove out-of-date comment.
8039 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8041 * expr.c (fixup_args_size_notes): Check that any existing
8042 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8043 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8044 (emit_single_push_insn): ...here.
8046 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8048 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8049 (const_vector_encoded_nelts): New function.
8050 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8051 (const_vector_int_elt, const_vector_elt): Declare.
8052 * emit-rtl.c (const_vector_int_elt_1): New function.
8053 (const_vector_elt): Likewise.
8054 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8055 of CONST_VECTOR_ELT.
8057 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8059 * expr.c: Include rtx-vector-builder.h.
8060 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8061 directly on the tree encoding.
8062 (const_vector_from_tree): Likewise.
8063 * optabs.c: Include rtx-vector-builder.h.
8064 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8065 sequence of "u" values.
8066 * vec-perm-indices.c: Include rtx-vector-builder.h.
8067 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8068 directly on the vec_perm_indices encoding.
8070 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8072 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8073 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8074 * rtx-vector-builder.h: New file.
8075 * rtx-vector-builder.c: Likewise.
8076 * rtl.h (rtx_def::u2): Add a const_vector field.
8077 (CONST_VECTOR_NPATTERNS): New macro.
8078 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8079 (CONST_VECTOR_DUPLICATE_P): Likewise.
8080 (CONST_VECTOR_STEPPED_P): Likewise.
8081 (CONST_VECTOR_ENCODED_ELT): Likewise.
8082 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8083 (unwrap_const_vec_duplicate): Likewise.
8084 (const_vec_series_p): Check for a non-duplicated vector encoding.
8085 Say that the function only returns true for integer vectors.
8086 * emit-rtl.c: Include rtx-vector-builder.h.
8087 (gen_const_vec_duplicate_1): Delete.
8088 (gen_const_vector): Call gen_const_vec_duplicate instead of
8089 gen_const_vec_duplicate_1.
8090 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8091 (gen_const_vec_duplicate): Use rtx_vector_builder.
8092 (gen_const_vec_series): Likewise.
8093 (gen_rtx_CONST_VECTOR): Likewise.
8094 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8095 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8096 Build a new vector rather than modifying a CONST_VECTOR in-place.
8097 (handle_special_swappables): Update call accordingly.
8098 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8099 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8100 Build a new vector rather than modifying a CONST_VECTOR in-place.
8101 (handle_special_swappables): Update call accordingly.
8103 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8105 * simplify-rtx.c (simplify_const_binary_operation): Use
8106 CONST_VECTOR_ELT instead of XVECEXP.
8108 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8110 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8111 the selector elements to be different from the data elements
8112 if the selector is a VECTOR_CST.
8113 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8114 ssizetype for the selector.
8116 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8118 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8119 before testing each element individually.
8120 * tree-vect-generic.c (lower_vec_perm): Likewise.
8122 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8124 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8125 * selftest-run-tests.c (selftest::run_tests): Call it.
8126 * vector-builder.h (vector_builder::operator ==): New function.
8127 (vector_builder::operator !=): Likewise.
8128 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8129 (vec_perm_indices::all_from_input_p): New function.
8130 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8131 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8132 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8133 instead of reading the VECTOR_CST directly. Detect whether both
8134 vector inputs are the same before constructing the vec_perm_indices,
8135 and update the number of inputs argument accordingly. Use the
8136 utility functions added above. Only construct sel2 if we need to.
8138 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8140 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8141 the broadcast of the low byte.
8142 (expand_mult_highpart): Use an explicit encoding for the permutes.
8143 * optabs-query.c (can_mult_highpart_p): Likewise.
8144 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8145 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8146 (vectorizable_bswap): Likewise.
8147 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8148 explicit encoding for the power-of-2 permutes.
8149 (vect_permute_store_chain): Likewise.
8150 (vect_grouped_load_supported): Likewise.
8151 (vect_permute_load_chain): Likewise.
8153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8155 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8156 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8157 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8158 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8159 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8160 (vect_gen_perm_mask_any): Likewise.
8162 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8164 * int-vector-builder.h: New file.
8165 * vec-perm-indices.h: Include int-vector-builder.h.
8166 (vec_perm_indices): Redefine as an int_vector_builder.
8167 (auto_vec_perm_indices): Delete.
8168 (vec_perm_builder): Redefine as a stand-alone class.
8169 (vec_perm_indices::vec_perm_indices): New function.
8170 (vec_perm_indices::clamp): Likewise.
8171 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8172 (vec_perm_indices::new_vector): New function.
8173 (vec_perm_indices::new_expanded_vector): Update for new
8174 vec_perm_indices class.
8175 (vec_perm_indices::rotate_inputs): New function.
8176 (vec_perm_indices::all_in_range_p): Operate directly on the
8177 encoded form, without computing elided elements.
8178 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8179 encoding. Update for new vec_perm_indices class.
8180 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8181 the given vec_perm_builder.
8182 (expand_vec_perm_var): Update vec_perm_builder constructor.
8183 (expand_mult_highpart): Use vec_perm_builder instead of
8184 auto_vec_perm_indices.
8185 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8186 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8187 or double series encoding as appropriate.
8188 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8189 vec_perm_indices instead of auto_vec_perm_indices.
8190 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8191 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8192 (vect_permute_store_chain): Likewise.
8193 (vect_grouped_load_supported): Likewise.
8194 (vect_permute_load_chain): Likewise.
8195 (vect_shift_permute_load_chain): Likewise.
8196 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8197 (vect_transform_slp_perm_load): Likewise.
8198 (vect_schedule_slp_instance): Likewise.
8199 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8200 (vectorizable_mask_load_store): Likewise.
8201 (vectorizable_bswap): Likewise.
8202 (vectorizable_store): Likewise.
8203 (vectorizable_load): Likewise.
8204 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8205 vec_perm_indices instead of auto_vec_perm_indices. Use
8206 tree_to_vec_perm_builder to read the vector from a tree.
8207 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8208 vec_perm_builder instead of a vec_perm_indices.
8209 (have_whole_vector_shift): Use vec_perm_builder and
8210 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8211 truncation to calc_vec_perm_mask_for_shift.
8212 (vect_create_epilog_for_reduction): Likewise.
8213 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8214 from auto_vec_perm_indices to vec_perm_indices.
8215 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8216 instead of changing individual elements.
8217 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8218 the vector in d.perm.
8219 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8220 from auto_vec_perm_indices to vec_perm_indices.
8221 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8222 instead of changing individual elements.
8223 (arm_vectorize_vec_perm_const): Use new_vector to install
8224 the vector in d.perm.
8225 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8226 Update vec_perm_builder constructor.
8227 (rs6000_expand_interleave): Likewise.
8228 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8229 (rs6000_expand_interleave): Likewise.
8231 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8233 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8234 to qimode could truncate the indices.
8235 * optabs.c (expand_vec_perm_var): Likewise.
8237 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8239 * Makefile.in (OBJS): Add vec-perm-indices.o.
8240 * vec-perm-indices.h: New file.
8241 * vec-perm-indices.c: Likewise.
8242 * target.h (vec_perm_indices): Replace with a forward class
8244 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8245 * optabs.h: Include vec-perm-indices.h.
8246 (expand_vec_perm): Delete.
8247 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8248 (expand_vec_perm_const): Declare.
8249 * target.def (vec_perm_const_ok): Replace with...
8250 (vec_perm_const): ...this new hook.
8251 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8252 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8253 * doc/tm.texi: Regenerate.
8254 * optabs.def (vec_perm_const): Delete.
8255 * doc/md.texi (vec_perm_const): Likewise.
8256 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8257 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8258 expand_vec_perm for constant permutation vectors. Assert that
8259 the mode of variable permutation vectors is the integer equivalent
8260 of the mode that is being permuted.
8261 * optabs-query.h (selector_fits_mode_p): Declare.
8262 * optabs-query.c: Include vec-perm-indices.h.
8263 (selector_fits_mode_p): New function.
8264 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8265 is defined, instead of checking whether the vec_perm_const_optab
8266 exists. Use targetm.vectorize.vec_perm_const instead of
8267 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8268 fit in the vector mode before using a variable permute.
8269 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8270 vec_perm_indices instead of an rtx.
8271 (expand_vec_perm): Replace with...
8272 (expand_vec_perm_const): ...this new function. Take the selector
8273 as a vec_perm_indices rather than an rtx. Also take the mode of
8274 the selector. Update call to shift_amt_for_vec_perm_mask.
8275 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8276 Use vec_perm_indices::new_expanded_vector to expand the original
8277 selector into bytes. Check whether the indices fit in the vector
8278 mode before using a variable permute.
8279 (expand_vec_perm_var): Make global.
8280 (expand_mult_highpart): Use expand_vec_perm_const.
8281 * fold-const.c: Includes vec-perm-indices.h.
8282 * tree-ssa-forwprop.c: Likewise.
8283 * tree-vect-data-refs.c: Likewise.
8284 * tree-vect-generic.c: Likewise.
8285 * tree-vect-loop.c: Likewise.
8286 * tree-vect-slp.c: Likewise.
8287 * tree-vect-stmts.c: Likewise.
8288 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8290 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8291 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8292 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8293 (aarch64_vectorize_vec_perm_const): ...this new function.
8294 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8295 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8296 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8297 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8298 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8299 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8300 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8302 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8303 check for NEON modes.
8304 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8305 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8306 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8307 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8309 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8310 the old VEC_PERM_CONST conditions.
8311 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8312 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8313 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8314 (ia64_vectorize_vec_perm_const_ok): Merge into...
8315 (ia64_vectorize_vec_perm_const): ...this new function.
8316 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8317 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8318 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8319 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8320 * config/mips/mips.c (mips_expand_vec_perm_const)
8321 (mips_vectorize_vec_perm_const_ok): Merge into...
8322 (mips_vectorize_vec_perm_const): ...this new function.
8323 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8324 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8325 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8326 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8327 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8328 (rs6000_expand_vec_perm_const): Delete.
8329 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8331 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8332 (altivec_expand_vec_perm_const_le): Take each operand individually.
8333 Operate on constant selectors rather than rtxes.
8334 (altivec_expand_vec_perm_const): Likewise. Update call to
8335 altivec_expand_vec_perm_const_le.
8336 (rs6000_expand_vec_perm_const): Delete.
8337 (rs6000_vectorize_vec_perm_const_ok): Delete.
8338 (rs6000_vectorize_vec_perm_const): New function.
8339 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8340 an element count and rtx array.
8341 (rs6000_expand_extract_even): Update call accordingly.
8342 (rs6000_expand_interleave): Likewise.
8343 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8344 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8345 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8346 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8347 (rs6000_expand_vec_perm_const): Delete.
8348 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8349 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8350 (altivec_expand_vec_perm_const_le): Take each operand individually.
8351 Operate on constant selectors rather than rtxes.
8352 (altivec_expand_vec_perm_const): Likewise. Update call to
8353 altivec_expand_vec_perm_const_le.
8354 (rs6000_expand_vec_perm_const): Delete.
8355 (rs6000_vectorize_vec_perm_const_ok): Delete.
8356 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8357 reference to the SPE evmerge intructions.
8358 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8359 an element count and rtx array.
8360 (rs6000_expand_extract_even): Update call accordingly.
8361 (rs6000_expand_interleave): Likewise.
8362 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8363 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8365 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8367 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8369 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8370 vector mode and that that mode matches the mode of the data
8372 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8373 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8374 directly using expand_vec_perm_1 when forcing selectors into
8376 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8378 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8380 * optabs-query.h (can_vec_perm_p): Delete.
8381 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8382 * optabs-query.c (can_vec_perm_p): Split into...
8383 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8384 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8385 particular selector is valid.
8386 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8387 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8388 (vect_grouped_load_supported): Likewise.
8389 (vect_shift_permute_load_chain): Likewise.
8390 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8391 (vect_transform_slp_perm_load): Likewise.
8392 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8393 (vectorizable_bswap): Likewise.
8394 (vect_gen_perm_mask_checked): Likewise.
8395 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8396 implementations of variable permutation vectors into account
8397 when deciding which selector to use.
8398 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8399 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8400 with a false third argument.
8401 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8402 to test whether the constant selector is valid and can_vec_perm_var_p
8403 to test whether a variable selector is valid.
8405 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8407 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8408 * optabs-query.c (can_vec_perm_p): Likewise.
8409 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8410 instead of vec_perm_indices.
8411 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8412 (vect_gen_perm_mask_checked): Likewise,
8413 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8414 (vect_gen_perm_mask_checked): Likewise,
8416 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8418 * optabs-query.h (qimode_for_vec_perm): Declare.
8419 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8420 (qimode_for_vec_perm): ...this new function.
8421 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8423 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8425 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8426 does not have a conditional at the top.
8428 2018-01-02 Richard Biener <rguenther@suse.de>
8430 * ipa-inline.c (big_speedup_p): Fix expression.
8432 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8435 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8438 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8442 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8443 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8444 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8445 cond_taken_branch_cost 3->4.
8447 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8449 PR tree-optimization/83581
8450 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8451 TODO_cleanup_cfg if any changes have been made.
8454 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8455 convert_modes if target mode has the right side, but different mode
8459 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8460 last argument when extracting from CONCAT. If either from_real or
8461 from_imag is NULL, use expansion through memory. If result is not
8462 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8463 the parts directly to inner mode, if even that fails, use expansion
8467 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8468 check for bswap in mode rather than HImode and use that in expand_unop
8471 Copyright (C) 2018 Free Software Foundation, Inc.
8473 Copying and distribution of this file, with or without modification,
8474 are permitted in any medium without royalty provided the copyright
8475 notice and this notice are preserved.