1 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5 Add strict argument and use it.
6 (rs6000_split_multireg_move): Update for new strict argument.
7 (mem_operand_gpr): Disallow all non-offsettable addresses.
8 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10 2018-03-09 Jakub Jelinek <jakub@redhat.com>
13 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
14 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
15 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
18 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
19 decl, use remap_type if we want to use the type.
21 2018-03-09 Martin Sebor <msebor@redhat.com>
23 PR tree-optimization/84526
24 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
26 (builtin_access::generic_overlap): Be prepared to handle non-array
29 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
31 PR rtl-optimization/84682
32 * lra-constraints.c (process_address_1): Check is_address flag
33 for address constraints.
34 (process_alt_operands): Likewise.
35 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
36 preprocess_constraints.
37 * recog.h (preprocess_constraints): Add oploc parameter.
39 * recog.c (preprocess_constraints): Test address_operand for
40 CT_ADDRESS constraints.
42 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
45 * lra-assigns.c (assign_by_spills): Return a flag of reload
46 assignment failure. Do not process the reload assignment
47 failures. Do not spill other reload pseudos if they has the same
49 (lra_assign): Add a return arg. Set up from the result of
50 assign_by_spills call.
51 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
52 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
53 usage_insns if it is not NULL.
54 (spill_hard_reg_in_range): New function.
55 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
56 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
58 (lra_assign): Change prototype.
59 * lra.c (lra): Add code to deal with fails by splitting hard reg
62 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
65 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
66 Accept complain bool parameter. Only emit errors if it is true.
67 (arm_parse_cpu_option_name): Likewise.
68 (arm_target_thumb_only): Adjust callers of the above.
69 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
70 prototype to take a default true bool parameter.
71 (arm_parse_arch_option_name): Likewise.
73 2018-03-09 David Malcolm <dmalcolm@redhat.com>
74 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
78 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
79 * configure: Regenerate.
80 * configure.ac ("linker --version-script option"): New.
81 ("linker soname option"): New.
83 2018-03-09 Richard Biener <rguenther@suse.de>
85 PR tree-optimization/84775
86 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
87 immediate uses of predicate stmts and mark them modified.
90 PR tree-optimization/84178
91 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
93 (version_loop_for_if_conversion): Delay update_ssa call.
94 (tree_if_conversion): Delay update_ssa until after predicate
97 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
100 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
101 when the function accesses prior frames.
103 2018-03-08 Jakub Jelinek <jakub@redhat.com>
106 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
107 gen_llsym, otherwise call maybe_gen_llsym.
110 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
111 has ',' character inside of it.
113 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
116 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
117 as clobbering CC_REGNUM.
119 2018-03-08 Richard Biener <rguenther@suse.de>
122 * tree-scalar-evolution.c: Include tree-into-ssa.h.
123 (follow_copies_to_constant): Do not follow SSA names registered
126 2018-03-08 Richard Biener <rguenther@suse.de>
128 PR tree-optimization/84178
129 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
131 (version_loop_for_if_conversion): Delay update_ssa call.
132 (tree_if_conversion): Delay update_ssa until after predicate
135 2018-03-08 David Malcolm <dmalcolm@redhat.com>
137 PR tree-optimization/84178
138 * tree-if-conv.c (release_bb_predicate): Remove the
139 the assertion that the stmts have NULL use_ops.
140 Discard the statements, asserting that they haven't
141 yet been added to a BB.
143 2018-03-08 Richard Biener <rguenther@suse.de>
145 PR tree-optimization/84746
146 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
147 (phi_translate): Pass in destination ANTIC_OUT set.
148 (phi_translate_1): Likewise. For a simplified result lookup
149 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
150 (phi_translate_set): Adjust.
151 (do_pre_regular_insertion): Likewise.
152 (do_pre_partial_partial_insertion): Likewise.
154 2018-03-08 Martin Liska <mliska@suse.cz>
156 PR gcov-profile/84735
157 * doc/gcov.texi: Document usage of profile files.
158 * gcov-io.h: Document changes in the format.
160 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
164 * dwarf2out.c (struct dw_line_info_table): Update comments for
166 (FORCE_RESET_NEXT_VIEW): New.
167 (FORCE_RESETTING_VIEW_P): New.
168 (RESETTING_VIEW_P): Check for -1 too.
169 (ZERO_VIEW_P): Likewise.
170 (new_line_info_table): Force-reset next view.
171 (dwarf2out_begin_function): Likewise.
172 (dwarf2out_source_line): Simplify zero_view_p initialization.
173 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
174 view directly. Omit view when omitting .loc at line 0.
176 2018-03-08 Jakub Jelinek <jakub@redhat.com>
178 PR tree-optimization/84740
179 * tree-switch-conversion.c (process_switch): Call build_constructors
180 only if info.phi_count is non-zero.
182 PR tree-optimization/84739
183 * tree-tailcall.c (find_tail_calls): Check call arguments against
184 DECL_ARGUMENTS (current_function_decl) rather than
185 DECL_ARGUMENTS (func) when checking for tail recursion.
187 2018-03-07 Jakub Jelinek <jakub@redhat.com>
189 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
190 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
191 Volker Reichelt's entry and add entries for people that perform
192 GCC fuzzy testing and report numerous bugs.
194 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
197 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
198 readonly data in sdata, if that is disabled.
199 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
200 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
201 -mreadonly-in-sdata option.
203 2018-03-07 Martin Sebor <msebor@redhat.com>
205 PR tree-optimization/84468
206 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
207 basic block when looking for nul assignment.
209 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
212 * except.h (output_function_exception_table): Adjust prototype.
213 * except.c (output_function_exception_table): Remove FNNAME parameter
214 and add SECTION parameter. Ouput one part of the table at a time.
215 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
216 the first part of the exception table and emit unwind directives.
217 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
218 (i386_pe_seh_cold_init): Likewise.
219 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
220 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
221 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
222 (ix86_output_call_insn): Emit a nop in one more case for SEH.
223 * config/i386/winnt.c: Include except.h.
224 (struct seh_frame_state): Add reg_offset, after_prologue and
225 in_cold_section fields.
226 (i386_pe_seh_end_prologue): Set seh->after_prologue.
227 (i386_pe_seh_cold_init): New function.
228 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
229 to seh->in_cold_section.
230 (seh_emit_push): Record the offset of the push.
231 (seh_emit_save): Record the offet of the save.
232 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
233 Test seh->after_prologue to disregard the epilogue.
234 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
235 (i386_pe_end_cold_function): New function.
237 2018-03-07 Jakub Jelinek <jakub@redhat.com>
240 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
241 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
244 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
246 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
247 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
250 * multiple_target.c: Include tree-inline.h and intl.h.
251 (expand_target_clones): Diagnose and fail if node->definition and
252 !tree_versionable_function_p (node->decl).
254 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
256 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
258 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
259 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
260 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
262 2018-03-06 Jakub Jelinek <jakub@redhat.com>
265 * combine.c (try_combine): Use reg_or_subregno instead of handling
266 just paradoxical SUBREGs and REGs.
268 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
270 * config/arc/arc.c (arc_finalize_pic): Remove function.
271 (arc_must_save_register): We use single base PIC register, remove
272 checks to save/restore the PIC register.
273 (arc_expand_prologue): Likewise.
274 * config/arc/arc-protos.h (arc_set_default_type_attributes):
276 (arc_verify_short): Likewise.
277 (arc_attr_type): Likewise.
278 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
279 (walk_stores): Likewise.
280 (arc_address_cost): Make it static.
281 (arc_verify_short): Likewise.
282 (branch_dest): Likewise.
283 (arc_attr_type): Likewise.
284 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
285 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
286 (arc_final_prescan_insn): Remove inserting the nops due to
287 hardware hazards. It is done in reorg step.
288 (insn_length_variant_t): Remove.
289 (insn_length_parameters_t): Likewise.
290 (arc_insn_length_parameters): Likewise.
291 (arc_get_insn_variants): Likewise.
292 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
294 2018-03-06 Jakub Jelinek <jakub@redhat.com>
297 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
300 PR tree-optimization/84687
301 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
303 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
305 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
307 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
308 Rename to ppc_speculation_barrier.
309 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
310 __builtin_ppc_speculation_barrier.
312 2018-03-05 Jakub Jelinek <jakub@redhat.com>
315 * combine.c (combine_simplify_rtx): Don't try to simplify if
316 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
319 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
321 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
322 to 32 bytes when compiling for POWER9.
324 2018-03-05 Jakub Jelinek <jakub@redhat.com>
327 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
328 regparm >= 3 with no arg reg available also for calls with
329 flag_force_indirect_call. Pass decl to ix86_function_regparm.
332 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
334 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
336 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
339 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
341 2018-03-05 Richard Biener <rguenther@suse.de>
343 PR tree-optimization/84486
344 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
345 When inserting a __builtin_assume_aligned call set the LHS
346 SSA name alignment info accordingly.
348 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
350 PR tree-optimization/84114
351 * config/aarch64/aarch64.c (aarch64_reassociation_width)
352 Avoid reassociation of FLOAT_MODE addition.
354 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
356 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
357 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
358 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
359 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
360 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
361 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
362 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
364 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
365 __builtin_ia32_wbinvd): New builtins.
366 (SPECIAL_ARGS2): New.
367 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
368 (SPECIAL_ARGS2): New.
369 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
370 (ix86_valid_target_attribute_inner_p): Ditto.
371 (ix86_init_mmx_sse_builtins): Add special_args2.
372 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
373 TARGET_WBNOINVD_P): New.
374 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
375 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
376 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
377 * config/i386/immintrin.h (_wbinvd): New intrinsic.
378 * config/i386/pconfigintrin.h: New file.
379 * config/i386/wbnoinvdintrin.h: Ditto.
380 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
381 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
383 2018-03-05 Richard Biener <rguenther@suse.de>
385 PR tree-optimization/84670
386 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
388 (BB_VISITED_WITH_VISITED_SUCCS): New define.
389 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
390 (compute_antic_aux): Only assert the number of values in ANTIC_IN
391 doesn't grow if all successors (recursively) were visited at least
394 2018-03-05 Richard Biener <rguenther@suse.de>
396 PR tree-optimization/84650
397 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
398 if executed in the loop pipeline.
400 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
402 * doc/configfiles.texi (Configuration Files): Move info about
403 conditionalizing $target-protos.h to...
404 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
405 differs from $target-protos.h.
407 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
408 Chung-Ju Wu <jasonwucj@gmail.com>
410 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
411 * config/nds32/nds32-multiple.md (setmemsi): Define.
412 * config/nds32/nds32-memory-manipulation.c
413 (nds32_gen_dup_4_byte_to_word_value): New.
414 (emit_setmem_word_loop): New.
415 (emit_setmem_byte_loop): New.
416 (nds32_expand_setmem_loop): New.
417 (nds32_expand_setmem_loop_v3m): New.
418 (nds32_expand_setmem_unroll): New.
419 (nds32_expand_setmem): New.
421 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
422 Chung-Ju Wu <jasonwucj@gmail.com>
424 * config/nds32/nds32-memory-manipulation.c
425 (nds32_emit_load_store): New.
426 (nds32_emit_post_inc_load_store): New.
427 (nds32_emit_mem_move): New.
428 (nds32_emit_mem_move_block): New.
429 (nds32_expand_movmemsi_loop_unknown_size): New.
430 (nds32_expand_movmemsi_loop_known_size): New.
431 (nds32_expand_movmemsi_loop): New.
432 (nds32_expand_movmemsi_unroll): New.
433 (nds32_expand_movmemqi): Rename ...
434 (nds32_expand_movmemsi): ... to this.
435 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
436 (movmemsi): ... to this.
437 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
438 (nds32_expand_movmemsi): ... to this.
440 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
441 Monk Chiang <sh.chiang04@gmail.com>
442 Chung-Ju Wu <jasonwucj@gmail.com>
444 * config/nds32/nds32-protos.h
445 (nds32_expand_load_multiple): New arguments.
446 (nds32_expand_store_multiple): Ditto.
447 (nds32_valid_multiple_load_store): Rename ...
448 (nds32_valid_multiple_load_store_p): ... to this.
449 * config/nds32/nds32-memory-manipulation.c
450 (nds32_expand_load_multiple): Refine implementation.
451 (nds32_expand_store_multiple): Ditto.
452 * config/nds32/nds32-multiple.md
453 (load_multiple): Update nds32_expand_load_multiple interface.
454 (store_multiple): Update nds32_expand_store_multiple interface.
455 * config/nds32/nds32-predicates.c
456 (nds32_valid_multiple_load_store): Rename ...
457 (nds32_valid_multiple_load_store_p): ... to this and refine
459 * config/nds32/predicates.md
460 (nds32_load_multiple_and_update_address_operation): New predicate.
461 (nds32_store_multiple_and_update_address_operation): New predicate.
463 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
464 Chung-Ju Wu <jasonwucj@gmail.com>
466 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
467 (combo): New attribute.
468 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
470 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
472 * config/nds32/nds32.opt: Change -mcmodel= default value.
474 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
475 Monk Chiang <sh.chiang04@gmail.com>
476 Chung-Ju Wu <jasonwucj@gmail.com>
478 * config/nds32/constants.md (unspec_element): New enum.
479 * config/nds32/constraints.md (Umw): New constraint.
480 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
481 * config/nds32/nds32-intrinsic.md: Likewise.
482 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
483 (nds32_valid_smw_lwm_base_p): New.
484 (nds32_output_smw_single_word): New.
485 (nds32_output_lmw_single_word): New.
486 (nds32_expand_unaligned_load): New.
487 (nds32_expand_unaligned_store): New.
488 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
489 (nds32_output_smw_single_word): Declare.
490 (nds32_output_lmw_single_word): Declare.
491 (nds32_expand_unaligned_load): Declare.
492 (nds32_expand_unaligned_store): Declare.
493 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
494 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
495 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
496 NDS32_BUILTIN_UASTORE_DW.
497 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
500 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
501 Kito Cheng <kito.cheng@gmail.com>
502 Chung-Ju Wu <jasonwucj@gmail.com>
504 * config/nds32/nds32-intrinsic.c
505 (nds32_expand_builtin_null_ftype_reg): Delete.
506 (nds32_expand_builtin_reg_ftype_imm): Ditto.
507 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
508 (nds32_read_argument): New.
509 (nds32_legitimize_target): Ditto.
510 (nds32_legitimize_argument): Ditto.
511 (nds32_check_constant_argument): Ditto.
512 (nds32_expand_unop_builtin): Ditto.
513 (nds32_expand_unopimm_builtin): Ditto.
514 (nds32_expand_binop_builtin): Ditto.
515 (nds32_builtin_decl_impl): Ditto.
516 (builtin_description): Ditto.
517 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
518 (nds32_init_builtins_impl): Ditto.
519 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
520 (nds32_builtin_decl): New.
521 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
522 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
524 2018-03-02 Jeff Law <law@redhat.com>
526 * reorg.c (stop_search_p): Handle DEBUG_INSN.
527 (redundant_insn, fill_simple_delay_slots): Likewise.
528 (fill_slots_from_thread): Likewise.
529 * resource.c (mark_referenced_resources): Likewise.
530 (mark_set_resources, find_dead_or_set_registers): Likewise.
532 2018-03-02 Jakub Jelinek <jakub@redhat.com>
534 * substring-locations.h (format_warning_va): Formatting fix for
536 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
538 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
539 * substring-locations.c: Include intl.h.
540 (format_warning_va): Turned into small wrapper around
541 format_warning_n_va, renamed to ...
542 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
543 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
545 (format_warning_at_substring_n): New function.
546 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
547 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
548 format_warning_at_substring with just a shorter name instead of
549 const function pointer.
550 (fmtwarn_n): New function.
551 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
552 appropriate, get rid of all the fmtstr temporaries, move conditionals
553 with G_() wrapped string literals directly into fmtwarn arguments,
554 cast dir.len to (int), formatting fixes.
556 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
558 * doc/invoke.texi: Remove "Cilk Plus" references.
560 2018-03-02 Jakub Jelinek <jakub@redhat.com>
561 Richard Biener <rguenther@suse.de>
564 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
565 for error or warning attributes if CALL_FROM_THUNK_P is set.
568 2018-03-02 Jakub Jelinek <jakub@redhat.com>
571 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
572 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
575 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
576 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
578 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
579 instead of -1U in last predictors element's probability member.
581 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
584 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
585 arguments if they are comparable.
587 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
589 PR tree-optimization/84634
590 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
591 masks and masked_loop_p with a single loop_masks, making sure it's
592 null for bb vectorization.
594 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
596 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
597 (vect_analyze_data_ref_access): Use loop->safe_len rather than
598 loop->force_vectorize to check whether there is no alias.
600 2018-03-02 Jakub Jelinek <jakub@redhat.com>
603 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
605 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
607 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
608 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
609 instead of a loop around prev_real_insn.
610 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
614 * config/i386/i386.c (ix86_print_operand): Use conditional
615 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
618 2018-03-02 Richard Biener <rguenther@suse.de>
620 PR tree-optimization/84427
621 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
622 (bitmap_set_subtract_values): Rewrite to handle multiple
625 (prune_clobbered_mems): Likewise.
626 (phi_translate): Take edge instead of pred/phiblock.
627 (phi_translate_1): Likewise.
628 (phi_translate_set): Likewise. Insert all translated
629 exprs for a value into the set, keeping possibly multiple
630 expressions per value.
631 (compute_antic_aux): Adjust for phi_translate changes.
632 When intersecting union the expressions and prune those
633 not in the final value set, keeping possibly multiple
634 expressions per value. Do not use value-insertion
635 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
636 all expressions. Add verification that the value-sets
637 only shrink during iteration.
638 (compute_partial_antic_aux): Adjust for the phi_translate changes.
639 (do_pre_regular_insertion): Likewise.
640 (do_pre_partial_partial_insertion): Likewise.
642 2018-03-02 Richard Biener <rguenther@suse.de>
645 * config/darwin.c (saved_debug_info_level): New static global.
646 (darwin_asm_lto_start): Disable debug info generation for LTO out.
647 (darwin_asm_lto_end): Restore debug info generation settings.
649 2018-03-01 Martin Liska <mliska@suse.cz>
652 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
655 2018-03-01 Richard Biener <rguenther@suse.de>
658 * dwarf2out.c (gen_variable_die): Properly handle late VLA
659 type annotation with LTO when debug was disabled at compile-time.
661 2018-03-01 Matthew Fortune <mfortune@gmail.com>
663 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
665 (mips_final_postscan_insn): Likewise.
667 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
669 PR rtl-optimization/84528
670 * alias.c (init_alias_target): Add commentary.
671 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
672 a unique base value if the frame pointer is not eliminated
673 to the stack pointer.
675 2018-03-01 Tom de Vries <tom@codesourcery.com>
677 PR rtl-optimization/83327
678 * lra-int.h (hard_regs_spilled_into): Declare.
679 * lra.c (hard_regs_spilled_into): Define.
680 (init_reg_info): Init hard_regs_spilled_into.
681 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
682 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
683 (process_bb_lives): Handle hard_regs_spilled_into.
684 (lra_create_live_ranges_1): Before doing liveness propagation, clear
685 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
687 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
689 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
690 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
691 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
692 * config/rs6000/aix72.h: New file.
694 2018-02-28 Jakub Jelinek <jakub@redhat.com>
696 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
697 instead of warning_at with conditional singular and plural messages
701 * stor-layout.c (update_alignment_for_field): For
702 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
703 && !DECL_PACKED (field), do the alignment update, just use
704 only desired_align instead of MAX (type_align, desired_align)
706 (place_field): Don't do known_align < desired_align handling
707 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
708 is non-NULL, instead do it after rli->prev_field handling and
709 only if not within a bitfield word. For DECL_PACKED (field)
710 use type_align of BITS_PER_UNIT.
712 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
714 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
715 superfluous parentheses and trailing spaces.
717 2018-02-28 Richard Biener <rguenther@suse.de>
719 PR tree-optimization/84584
720 * graphite-scop-detection.c (scop_detection::add_scop): Discard
721 SCoPs with fake exit edge.
723 2018-02-28 Martin Liska <mliska@suse.cz>
726 * timevar.c (timer::print): Fix format to properly print 100%
729 2018-02-28 Richard Biener <rguenther@suse.de>
732 * genmatch.c (capture_info::walk_match): Do not mark
733 captured expressions without operands as expr_p given
734 they act more like predicates and should be subject to
735 "lost tail" side-effect preserving.
737 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
739 PR rtl-optimization/81611
740 * auto-inc-dec.c (attempt_change): Move dead note from
741 mem_insn if it's the next use of regno
742 (find_address): Take address use of reg holding
743 non-incremented value. Add parm to limit search to the named
745 (merge_in_block): Attempt to use a mem insn that is the next
746 use of the original regno.
748 2018-02-27 Martin Sebor <msebor@redhat.com>
751 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
752 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
754 2018-02-27 Martin Sebor <msebor@redhat.com>
757 * diagnostic-core.h (warning_n, error_n, inform_n): Change
758 n argument to unsigned HOST_WIDE_INT.
759 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
760 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
761 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
762 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
764 2018-02-27 Richard Biener <rguenther@suse.de>
766 PR tree-optimization/84512
767 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
768 Do not use the estimate returned from record_stmt_cost for
769 the scalar iteration cost but sum properly using add_stmt_cost.
771 2018-02-27 Richard Biener <rguenther@suse.de>
773 PR tree-optimization/84466
774 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
775 Adjust last change to less strictly validate use operands.
777 2018-02-27 Martin Liska <mliska@suse.cz>
779 PR gcov-profile/84548
780 * gcov.c (process_file): Allow partial overlap and consider it
781 also as group functions.
782 (output_lines): Properly calculate range of lines for a group.
784 2018-02-27 Martin Liska <mliska@suse.cz>
786 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
787 'ggc' suffixes. Change first column width.
788 (timer::print): Fix formatting of the column.
790 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
792 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
793 preserve inline entry blocks for the sake of debug inline
794 entry point markers alone.
795 (remove_unused_locals): Suggest in comments a better place to
796 force the preservation of inline entry blocks that are
797 otherwise unused, but do not preserve them.
799 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
801 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
803 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
806 * config/i386/constraints.md (Bs): Replace
807 ix86_indirect_branch_register with
808 TARGET_INDIRECT_BRANCH_REGISTER.
810 * config/i386/i386.md (indirect_jump): Likewise.
811 (tablejump): Likewise.
812 (*sibcall_memory): Likewise.
813 (*sibcall_value_memory): Likewise.
814 Peepholes of indirect call and jump via memory: Likewise.
815 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
816 (*sibcall_value_GOT_32): Likewise.
817 * config/i386/predicates.md (indirect_branch_operand): Likewise.
818 (GOT_memory_operand): Likewise.
819 (call_insn_operand): Likewise.
820 (sibcall_insn_operand): Likewise.
821 (GOT32_symbol_operand): Likewise.
822 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
824 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
826 PR rtl-optimization/83496
827 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
828 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
829 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
830 redundant insn, if any.
831 (relax_delay_slots): Likewise.
832 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
834 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
836 PR tree-optimization/83965
837 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
838 that grouped statements are part of a reduction chain. Return
839 true if the statement is not marked as a reduction itself but
841 (vect_recog_dot_prod_pattern): Don't check whether the statement
842 is part of a group here.
843 (vect_recog_sad_pattern): Likewise.
844 (vect_recog_widen_sum_pattern): Likewise.
846 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
849 * final.c (rest_of_clean_state): Also look for calls inside sequences.
851 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
854 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
856 (ix86_output_indirect_function_return): New prototype.
857 (ix86_split_simple_return_pop_internal): Likewise.
858 * config/i386/i386.c (indirect_return_via_cx): New.
859 (indirect_return_via_cx_bnd): Likewise.
860 (indirect_thunk_name): Handle return va CX_REG.
861 (output_indirect_thunk_function): Create alias for
862 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
863 (ix86_output_indirect_jmp): Remove the bool argument.
864 (ix86_output_indirect_function_return): New function.
865 (ix86_split_simple_return_pop_internal): Likewise.
866 * config/i386/i386.md (*indirect_jump): Don't pass false
867 to ix86_output_indirect_jmp.
868 (*tablejump_1): Likewise.
869 (simple_return_pop_internal): Change it to define_insn_and_split.
870 Call ix86_split_simple_return_pop_internal to split it for
872 (simple_return_indirect_internal): Call
873 ix86_output_indirect_function_return instead of
874 ix86_output_indirect_jmp.
876 2018-02-26 Jakub Jelinek <jakub@redhat.com>
879 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
880 memset and value initialization afterwards.
882 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
884 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
886 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
889 * common/config/aarch64/aarch64-common.c
890 (aarch_option_optimization_table[]): Switch
891 off fomit-frame-pointer
893 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
894 Chung-Ju Wu <jasonwucj@gmail.com>
896 * config/nds32/nds32-multiple.md (load_multiple): Disallow
898 (store_multiple): Ditto.
900 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
902 * config.gcc: Add --with-cpu support for nds32 target.
903 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
904 * config/nds32/nds32.opt: Add -mcpu= option.
906 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
908 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
909 isel=yes): Warn for these deprecated options.
911 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
913 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
914 ISA_2_5_MASKS_EMBEDDED.
916 2018-02-23 Jakub Jelinek <jakub@redhat.com>
918 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
919 p->max as pointers rather than using iterative_hash_expr.
921 2018-02-23 Carl Love <cel@us.ibm.com>
923 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
924 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
926 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
927 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
928 P8V_BUILTIN_VEC_VUNSIGNED2.
930 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
933 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
934 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
936 * lra-constraints.c (curr_insn_transform): Set up
937 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
939 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
940 * lra-eliminations.c (spill_pseudos): Ditto.
941 (process_insn_for_elimination): Ditto.
942 * lra-lives.c (reg_early_clobber_p): Use the new macros.
943 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
944 LRA_NON_CLOBBERED_ALT.
946 2018-02-22 Martin Sebor <msebor@redhat.com>
948 PR tree-optimization/84480
949 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
950 to maybe_diag_stxncpy_trunc. Call it.
951 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
952 from gimple_fold_builtin_strcpy. Print inlining stack.
953 (handle_builtin_stxncpy): Print inlining stack.
954 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
956 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
959 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
960 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
961 and -fcheck-pointer-bounds are used together.
962 (indirect_thunk_prefix): New enum.
963 (indirect_thunk_need_prefix): New function.
964 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
965 "_nt" instead of "_bnd" for NOTRACK prefix.
966 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
967 (output_indirect_thunk_function): Likewise.
969 (ix86_code_end): Update output_indirect_thunk_function calls.
970 (ix86_output_indirect_branch_via_reg): Replace
971 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
972 (ix86_output_indirect_branch_via_push): Likewise.
973 (ix86_output_function_return): Likewise.
974 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
975 incompatible with -fcf-protection=branch and
976 -fcheck-pointer-bounds.
978 2018-02-22 Steve Ellcey <sellcey@cavium.com>
981 * config/aarch64/aarch64.c (aarch64_print_address_internal):
982 Change gcc_assert call to output_operand_lossage.
984 2018-02-22 Steve Ellcey <sellcey@cavium.com>
986 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
988 2018-02-22 DJ Delorie <dj@redhat.com>
989 Sebastian Perta <sebastian.perta@renesas.com>
990 Oleg Endo <olegendo@gcc.gnu.org>
992 * config/rx/rx.c (rx_rtx_costs): New function.
993 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
995 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
997 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
999 2018-02-22 Martin Liska <mliska@suse.cz>
1002 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1003 Add "native" as a possible value.
1005 2018-02-22 Martin Liska <mliska@suse.cz>
1008 * config/i386/i386.c (ix86_option_override_internal):
1009 Add "native" as a possible value for -march and -mtune.
1011 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1014 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1015 to all type variants.
1017 PR tree-optimization/84503
1018 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1019 width as info->bitpos + info->bitsize - start.
1020 (merged_store_group::merge_overlapping): Simplify width computation.
1021 (check_no_overlap): New function.
1022 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1023 start + width and last_order of the group, fail if check_no_overlap
1025 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1026 to group if check_no_overlap fails.
1028 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1030 * config/rs6000/altivec.md: Delete contraint arguments to
1031 define_expand, define_split, and define_peephole2, and in
1032 define_insn_and_split if always unused.
1033 * config/rs6000/darwin.md: Ditto.
1034 * config/rs6000/dfp.md: Ditto.
1035 * config/rs6000/rs6000.md: Ditto.
1036 * config/rs6000/sync.md: Ditto.
1037 * config/rs6000/vector.md: Ditto.
1038 * config/rs6000/vsx.md: Ditto.
1040 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1042 * config/rs6000/altivec.md: Write output control strings as braced
1043 blocks instead of double-quoted strings.
1044 * config/rs6000/darwin.md: Ditto.
1045 * config/rs6000/rs6000.md: Ditto.
1046 * config/rs6000/vector.md: Ditto.
1047 * config/rs6000/vsx.md: Ditto.
1049 2018-02-21 Jason Merrill <jason@redhat.com>
1051 PR c++/84314 - ICE with templates and fastcall attribute.
1052 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1054 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1056 * ipa-cp.c (determine_versionability): Fix comment typos.
1058 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1061 * ipa-cp.c (determine_versionability): Do not version functions caling
1064 2018-02-21 Martin Liska <mliska@suse.cz>
1067 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1068 Add "native" as a possible value.
1069 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1070 the macro when native cpu detection is available.
1072 2018-02-21 Martin Liska <mliska@suse.cz>
1075 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1076 Add "native" as a possible value.
1077 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1078 when native cpu detection is available.
1080 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1081 Martin Sebor <msebor@redhat.com>
1083 PR tree-optimization/84478
1084 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1086 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1087 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1088 support which is conservatively correct, for 2 only stay conservative
1089 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1090 argument to the 2 argument get_range_strlen, adjust 6 arg
1091 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1093 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1094 (gimple_fold_builtin_strlen): Pass true as last argument to
1097 2018-02-20 Martin Sebor <msebor@redhat.com>
1100 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1101 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1102 (builtin_memref::builtin_memref): Factor out parts into
1103 set_base_and_offset and call it.
1105 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1108 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1109 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1110 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1111 search at the associated MODE_INT.
1113 2018-02-20 Jeff Law <law@redhat.com>
1116 PR tree-optimization/81592
1118 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1119 for range data rather than using global data.
1120 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1121 range data rather than using global data.
1122 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1123 pass it to children as needed.
1124 (struct directive::fmtresult): Similarly.
1125 (struct directive::set_width): Similarly.
1126 (struct directive::set_precision): Similarly.
1127 (format_integer, format_directive, parse_directive): Similarly.
1128 (format_none): Accept unnamed vr_values parameter.
1129 (format_percent, format_floating, format_character): Similarly.
1130 (format_string, format_plain): Similarly.
1131 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1132 the EVRP range analyzer for range data rather than using global data.
1133 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1134 gimple-ssa-evrp-analyze.h
1135 (class sprintf_dom_walker): Add after_dom_children member function.
1136 Add evrp_range_analyzer member.
1137 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1138 range analyzer as needed.
1139 (sprintf_dom_walker::after_dom_children): New member function.
1140 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1142 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1143 (evrp_range_analyzer::pop_to_marker): Likewise.
1145 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1147 PR tree-optimization/84419
1148 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1149 with the required type if its current type is compatible but
1152 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1155 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1156 after vectorization.
1158 2018-02-20 Martin Liska <mliska@suse.cz>
1161 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1162 possible values if we don't have a hint.
1164 2018-02-20 Martin Liska <mliska@suse.cz>
1168 * final.c (shorten_branches): Build align_tab array with one
1170 * opts.c (finish_options): Add alignment option limit check.
1171 (MAX_CODE_ALIGN): Likewise.
1172 (MAX_CODE_ALIGN_VALUE): Likewise.
1173 * doc/invoke.texi: Document maximum allowed option value for
1174 all -falign-* options.
1176 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1179 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1180 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1181 * var-tracking.c (emit_note_insn_var_location): Remove all references
1182 to NOTE_INSN_CALL_ARG_LOCATION.
1183 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1184 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1185 Use copy_rtx_if_shared.
1186 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1187 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1188 (dwarf2out_var_location): Remove handling of
1189 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1191 * final.c (final_scan_insn): Remove all references to
1192 NOTE_INSN_CALL_ARG_LOCATION.
1193 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1194 before dumping final insns.
1195 * except.c (emit_note_eh_region_end): Remove all references to
1196 NOTE_INSN_CALL_ARG_LOCATION.
1197 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1198 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1199 * config/arc/arc.c (hwloop_optimize): Likewise.
1200 * config/arm/arm.c (create_fix_barrier): Likewise.
1201 * config/s390/s390.c (s390_chunkify_start): Likewise.
1202 * config/sh/sh.c (find_barrier): Likewise.
1203 * config/i386/i386.c (rest_of_insert_endbranch,
1204 ix86_seh_fixup_eh_fallthru): Likewise.
1205 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1206 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1207 * config/frv/frv.c (frv_function_prologue): Likewise.
1208 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1210 (note_outside_basic_block_p): Remove all references to
1211 NOTE_INSN_CALL_ARG_LOCATION.
1212 * gengtype.c (adjust_field_rtx_def): Likewise.
1213 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1215 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1216 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1219 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1222 PR tree-optimization/84452
1223 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1224 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1227 2018-02-19 Martin Liska <mliska@suse.cz>
1230 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1232 2018-02-19 Martin Liska <mliska@suse.cz>
1233 Richard Sandiford <richard.sandiford@linaro.org>
1235 PR tree-optimization/82491
1236 * gimple-fold.c (get_base_constructor): Make earlier bail out
1239 2018-02-19 Carl Love <cel@us.ibm.com>
1241 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1242 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1244 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1245 P8V_BUILTIN_VEC_NEG.
1247 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1249 * config/rl78/rl78.md (movdf): New define expand.
1251 2018-02-19 Martin Liska <mliska@suse.cz>
1254 * doc/invoke.texi: Fix typo.
1255 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1257 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1259 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1260 handle rs6000_single_float and rs6000_double_float specially for
1263 2018-02-16 Jeff Law <law@redhat.com>
1265 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1268 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1271 * ipa-inline.c (inline_small_functions): Fix a typo.
1273 2018-02-16 Nathan Sidwell <nathan@acm.org>
1275 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1277 2018-02-16 Carl Love <cel@us.ibm.com>
1279 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1280 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1281 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1282 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1283 expansion to P8V_BUILTIN_VEC_FLOAT2.
1285 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1287 PR rtl-optimization/70023
1288 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1289 src_regno into account.
1291 2018-02-16 Carl Love <cel@us.ibm.com>
1293 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1294 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1295 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1296 * config/rs6000/rs6000.c: Remove case statements for
1297 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1298 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1299 and P9V_BUILTIN_VEC_VINSERT4B.
1300 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1301 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1302 * config/rs6000/vsx.md:
1303 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1306 2018-02-16 Carl Love <cel@us.ibm.com>
1308 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1310 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1312 * config/rs6000/rs6000-c.c: Add the definitions for
1313 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1314 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1315 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1316 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1317 definition for insert4b and define insn *insert3b_internal.
1318 * doc/extend.texi: Add documentation for vec_extract4b.
1320 2018-02-16 Nathan Sidwell <nathan@acm.org>
1322 * doc/extend.texi (Backwards Compatibility): Mention friend
1323 injection. Note for-scope is deprecated.
1324 * doc/invoke.texi (-ffriend-injection): Deprecate.
1326 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1328 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1329 that moved to I2, also allow destinations that are a paradoxical
1330 subreg (instead of a normal reg).
1332 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1335 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1338 2018-02-16 Richard Biener <rguenther@suse.de>
1340 PR tree-optimization/84037
1341 PR tree-optimization/84016
1343 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1344 Adjust vec_construct for the fact we need additional higher latency
1345 128bit inserts for AVX256 and AVX512 vector builds.
1346 (ix86_add_stmt_cost): Scale vector construction cost for
1349 2018-02-16 Richard Biener <rguenther@suse.de>
1351 PR tree-optimization/84417
1352 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1353 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1354 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1356 2018-02-16 Martin Liska <mliska@suse.cz>
1359 * internal-fn.def (ASAN_CHECK): Set proper flags.
1360 (ASAN_MARK): Likewise.
1362 2018-02-16 Julia Koval <julia.koval@intel.com>
1364 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1365 from PTA_CANNONLAKE.
1367 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1370 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1371 Use ++iter rather than iter++ for std::list iterators.
1372 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1373 defer deleting them until all nodes in the forest are processed. Do
1374 free even leaf nodes. Change to_process into auto_vec.
1377 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1378 * vec.h (vec_default_construct): Use memset instead of placement new
1379 if BROKEN_VALUE_INITIALIZATION is defined.
1380 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1381 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1384 PR rtl-optimization/83723
1385 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1386 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1387 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1389 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1391 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1393 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1395 PR rtl-optimization/81443
1396 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1397 from inner REGs to paradoxical SUBREGs.
1399 2018-02-16 Richard Biener <rguenther@suse.de>
1401 PR tree-optimization/84399
1402 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1403 For operands we can analyze at their definition make sure we can
1404 analyze them at each use as well.
1406 2018-02-16 Richard Biener <rguenther@suse.de>
1408 PR tree-optimization/84190
1409 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1410 volatile accesses if the decl isn't volatile.
1412 2018-02-15 Jason Merrill <jason@redhat.com>
1414 PR c++/84314 - ICE with templates and fastcall attribute.
1415 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1416 TYPE_CANONICAL on an existing type.
1418 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1420 PR tree-optimization/84383
1421 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1422 dstoff nor call operand_equal_p if dstbase is NULL.
1424 PR tree-optimization/84334
1425 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1426 also a CONSTANT_CLASS_P, punt.
1428 2018-02-14 Jim Wilson <jimw@sifive.com>
1430 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1431 first SMALL_OPERAND check. New local min_second_step. Move assert
1432 to where locals are set. Add TARGET_RVC support.
1433 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1435 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1437 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1439 2018-02-14 Martin Sebor <msebor@redhat.com>
1441 PR tree-optimization/83698
1442 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1443 arrays constrain the offset range to their bounds.
1444 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1445 (builtin_access::overlap): Avoid setting the size of overlap if it's
1447 (maybe_diag_overlap): Also consider arrays when deciding what values
1448 of offsets to include in diagnostics.
1450 2018-02-14 Martin Sebor <msebor@redhat.com>
1453 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1454 that correspond to the kind of a declaration.
1456 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1459 * config/pa/pa.md: Load address of PIC label using the linkage table
1460 if the label is nonlocal.
1462 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1464 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1465 warning message if user requests -maltivec=be.
1466 * doc/invoke.texi: Document deprecation of -maltivec=be.
1468 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1471 * config/rs6000/rs6000-c.c: Update definitions for
1472 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1473 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1475 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1478 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1479 add _get_ssp intrinsics. Remove argument from
1480 __builtin_ia32_rdssp[d|q].
1481 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1482 * config/i386/i386-builtin.def: Remove argument from
1483 __builtin_ia32_rdssp[d|q].
1484 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1485 ix86_expand_special_args_builtin for _rdssp[d|q].
1486 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1487 Clear register before usage.
1488 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1489 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1491 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1493 PR tree-optimization/84357
1494 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1495 operand 1 of an ARRAY_REF too.
1497 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1500 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1501 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1503 (set_of_reg): New struct.
1504 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1505 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1506 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1508 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1509 Split into bitclr, bitset, bitinvert patterns if appropriate.
1510 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1511 use rx_fuse_in_memory_bitop.
1512 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1513 to named insn, correct maximum insn length.
1515 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1518 * machmode.def: Define a complex mode for PARTIAL_INT.
1519 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1521 * doc/rtl.texi: Document CSPImode.
1522 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1524 (msp430_hard_regno_nregs_with_padding): Likewise.
1526 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1529 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1531 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1533 PR rtl-optimization/84169
1534 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1535 we generated a parallel as new i3 and we split that to new i2 and i3
1536 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1537 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1538 those to i2, not i1. Partially rewrite this scan code.
1540 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1543 * stor-layout.c (place_field): For variable length fields, adjust
1544 offset_align afterwards not just based on the field's alignment,
1545 but also on the size.
1548 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1549 of exps and logs in the use_exp2 case.
1551 2018-02-13 Jeff Law <law@redhat.com>
1553 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1556 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1559 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1563 * common.opt (gas-loc-support, gas-locview-support): New.
1564 (ginline-points, ginternal-reset-location-views): New.
1565 * doc/invoke.texi: Document them. Use @itemx where intended.
1566 (gvariable-location-views): Adjust.
1567 * target.def (reset_location_view): New.
1568 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1569 (TARGET_RESET_LOCATION_VIEW): New.
1570 * doc/tm.texi: Rebuilt.
1571 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1572 (dwarf2out_default_as_locview_support): New.
1573 (output_asm_line_debug_info): Use option variables.
1574 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1575 (output_loc_list): Likewise.
1576 (add_high_low_attributes): Check option variables.
1577 Don't output entry view attribute in strict mode.
1578 (gen_inlined_subroutine_die): Check option variables.
1579 (dwarf2out_inline_entry): Likewise.
1580 (init_sections_and_labels): Likewise.
1581 (dwarf2out_early_finish): Likewise.
1582 (maybe_reset_location_view): New, from...
1583 (dwarf2out_var_location): ... here. Call it.
1584 * debug.h (dwarf2out_default_as_loc_support): Declare.
1585 (dwarf2out_default_as_locview_support): Declare.
1586 * hooks.c (hook_int_rtx_insn_0): New.
1587 * hooks.h (hook_int_rtx_insn_0): Declare.
1588 * toplev.c (process_options): Take -gas-loc-support and
1589 -gas-locview-support from dwarf2out. Enable
1590 -gvariable-location-views by default only with locview
1591 assembler support. Enable -ginternal-reset-location-views by
1592 default only if the target defines the corresponding hook.
1593 Enable -ginline-points by default if location views are
1594 enabled; force it disabled if statement frontiers are
1596 * tree-inline.c (expand_call_inline): Check option variables.
1597 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1599 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1601 PR tree-optimization/84321
1602 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1603 handling. Also check whether the anti-range contains any values
1604 that satisfy the mask; switch to a VR_RANGE if not.
1606 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1609 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1611 2018-02-13 Martin Jambor <mjambor@suse.cz>
1614 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1615 of call statements, also set location of a load to a temporary.
1617 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1619 * config/rl78/rl78.c (add_vector_labels): New function.
1620 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1621 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1622 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1623 which checks that no arguments are passed.
1624 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1625 * doc/extend.texi: Documentation for the new attribute.
1627 2018-02-13 Andreas Schwab <schwab@suse.de>
1629 * config/riscv/linux.h (CPP_SPEC): Define.
1631 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1634 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1635 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1636 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1637 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1638 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1639 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1640 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1641 temporarily for AES and PCLMUL builtins.
1643 PR tree-optimization/84339
1644 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1645 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1649 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1650 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1651 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1653 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1655 * omp-simd-clone.h: New file.
1656 * omp-simd-clone.c: Include omp-simd-clone.h.
1657 (expand_simd_clones): No longer static.
1658 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1659 cgraph.h and omp-simd-clone.h.
1660 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1661 (vect_recog_widen_shift_pattern): Formatting fix.
1662 (vect_pattern_recog_1): Don't check optab for calls.
1665 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1666 operands[2] into a REG before using gen_lowpart on it.
1668 2018-02-12 Jeff Law <law@redhat.com>
1671 * config/sh/sh.c (find_barrier): Consider a sibling call
1674 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1675 successfully back substituting a reg.
1677 2018-02-12 Richard Biener <rguenther@suse.de>
1679 PR tree-optimization/84037
1680 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1681 parameter, move visited init to caller.
1682 (vect_slp_analyze_operations): Separate cost from validity
1683 check, initialize visited once for all instances.
1684 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1686 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1688 (vect_model_promotion_demotion_cost): Likewise.
1689 (vectorizable_bswap): Guard cost modeling with !slp_node
1690 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1692 (vectorizable_call): Likewise.
1693 (vectorizable_conversion): Likewise.
1694 (vectorizable_assignment): Likewise.
1695 (vectorizable_shift): Likewise.
1696 (vectorizable_operation): Likewise.
1697 (vectorizable_store): Likewise.
1698 (vectorizable_load): Likewise.
1699 (vectorizable_condition): Likewise.
1700 (vectorizable_comparison): Likewise.
1702 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1705 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1706 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1707 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1709 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1712 * params.def (inline-min-speedup): Increase from 8 to 15.
1713 (max-inline-insns-auto): Decrease from 40 to 30.
1714 * ipa-split.c (consider_split): Add some buffer for function to
1715 be considered inlining candidate.
1716 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1719 2018-02-12 Richard Biener <rguenther@suse.de>
1721 PR tree-optimization/84037
1722 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1723 matched stmts if we cannot swap the non-matched ones.
1725 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1727 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1728 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1729 _mm_maskz_scalef_round_ss): New intrinsics.
1730 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1731 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1732 __builtin_ia32_scalefss_round): Remove.
1733 (__builtin_ia32_scalefsd_mask_round,
1734 __builtin_ia32_scalefss_mask_round): New intrinsics.
1735 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1736 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1737 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1738 "<round_constraint>")): Changed to ...
1739 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1740 "<round_scalar_constraint>")): ... this.
1741 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1742 %0, %1, %2<round_op3>}"): Changed to ...
1743 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1744 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1745 %2<round_scalar_mask_op3>}"): ... this.
1746 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1748 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1750 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1751 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1752 (_mm_maskz_sqrt_round_ss): New intrinsics.
1753 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1754 (__builtin_ia32_sqrtsd_mask_round)
1755 (__builtin_ia32_sqrtss_mask_round): New builtins.
1756 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1757 (__builtin_ia32_sqrtss_round): Remove.
1758 (__builtin_ia32_sqrtsd_mask_round)
1759 (__builtin_ia32_sqrtss_mask_round): New builtins.
1760 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1761 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1762 ((match_operand:VF_128 1 "vector_operand"
1763 "xBm,<round_constraint>")): Changed to ...
1764 ((match_operand:VF_128 1 "vector_operand"
1765 "xBm,<round_scalar_constraint>")): ... this.
1766 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1767 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1768 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1769 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1770 %<iptr>1<round_scalar_mask_op3>}): ... this.
1771 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1772 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1774 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1777 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1778 Cast vec_cmpeq result to correct type.
1779 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1780 Cast vec_cmpgt result to correct type.
1782 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1784 * final.c (final_scan_insn_1): Renamed from...
1785 (final_scan_insn): ... this. New wrapper, to recover
1786 seen from the outermost call in recursive ones.
1787 * config/sparc/sparc.c (output_return): Drop seen from call.
1788 (output_sibcall): Likewise.
1789 * config/visium/visium.c (output_branch): Likewise.
1791 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1793 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1796 2018-02-10 Alan Modra <amodra@gmail.com>
1799 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1800 Specify LR as an input.
1802 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1805 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1806 remove_member_access_dummy_vars): New functions.
1807 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1808 lower_omp_1, execute_lower_omp): Use them.
1810 PR rtl-optimization/84308
1811 * shrink-wrap.c (spread_components): Release todo vector.
1813 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1815 PR rtl-optimization/57193
1816 * ira-color.c (struct allocno_color_data): Add member
1817 conflict_allocno_hard_prefs.
1818 (update_conflict_allocno_hard_prefs): New.
1819 (bucket_allocno_compare_func): Add a preference based on
1820 conflict_allocno_hard_prefs.
1821 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1822 (color_allocnos): Remove a dead code. Initiate
1823 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1825 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1828 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1829 constraint from =wa to wa. Avoid a subreg on the output operand,
1830 instead use a pseudo and subreg it in a move.
1831 (p9_xxbrd_<mode>): Changed to ...
1832 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1833 (p9_xxbrd_v2df): New expander.
1834 (p9_xxbrw_<mode>): Changed to ...
1835 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1836 (p9_xxbrw_v4sf): New expander.
1838 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1840 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1842 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1845 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1846 multiply in 32-bit mode.
1847 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1848 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1851 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1853 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1854 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1855 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1856 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1858 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1861 * dwarf2out.c (is_trivial_indirect_ref): New function.
1862 (dwarf2out_late_global_decl): Do not generate a location
1863 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1864 and that are not defined in the current unit.
1866 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1868 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1869 instead of a libcall for UNORDERED.
1871 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1874 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1875 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1877 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1880 * config/s390/s390.c (s390_set_current_function): Invoke
1881 s390_indirect_branch_settings also if fndecl didn't change.
1883 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1885 * config/rs6000/rs6000.md (blockage): Set length to zero.
1887 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1889 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1891 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1894 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1895 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1899 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1900 PARALLEL incoming that failed vt_get_decl_and_offset check.
1903 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1904 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1906 (get_variable_section): For decls in named .bss* sections pass true as
1907 second argument to bss_initializer_p.
1909 2018-02-09 Marek Polacek <polacek@redhat.com>
1910 Jakub Jelinek <jakub@redhat.com>
1913 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1914 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1915 Sync some changes from cxx_fold_indirect_ref.
1917 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1919 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1921 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1922 (BLOCK_INLINE_ENTRY_LABEL): New.
1923 (dwarf2out_var_location): Disregard inline entry markers.
1924 (inline_entry_data): New struct.
1925 (inline_entry_data_hasher): New hashtable type.
1926 (inline_entry_data_hasher::hash): New.
1927 (inline_entry_data_hasher::equal): New.
1928 (inline_entry_data_table): New variable.
1929 (add_high_low_attributes): Add DW_AT_entry_pc and
1930 DW_AT_GNU_entry_view attributes if a pending entry is found
1931 in inline_entry_data_table. Add old entry_pc attribute only
1932 if debug nonbinding markers are disabled.
1933 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1934 markers are enabled.
1935 (block_within_block_p, dwarf2out_inline_entry): New.
1936 (dwarf2out_finish): Check that no entries remained in
1937 inline_entry_data_table.
1938 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1939 (final_scan_insn, notice_source_line): Likewise.
1940 (rest_of_clean_state): Skip inline entry markers.
1941 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1943 * gimple.c (gimple_build_debug_inline_entry): New.
1944 * gimple.h (enum gimple_debug_subcode): Add
1945 GIMPLE_DEBUG_INLINE_ENTRY.
1946 (gimple_build_debug_inline_entry): Declare.
1947 (gimple_debug_inline_entry_p): New.
1948 (gimple_debug_nonbind_marker_p): Adjust.
1949 * insn-notes.def (INLINE_ENTRY): New.
1950 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1951 inline entry marker notes.
1952 (print_insn): Likewise.
1953 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1954 (INSN_DEBUG_MARKER_KIND): Likewise.
1955 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1956 * tree-inline.c (expand_call_inline): Build and insert
1957 debug_inline_entry stmt.
1958 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1959 inline entry blocks early, if nonbind markers are enabled.
1960 (dump_scope_block): Dump fragment info.
1961 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1962 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1963 (gimple_build_debug_inline_entry): New.
1964 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1965 Enable/disable inline entry points too.
1966 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1967 (DEBUG_INSN): Describe inline entry markers.
1969 * common.opt (gvariable-location-views): New.
1970 (gvariable-location-views=incompat5): New.
1971 * config.in: Rebuilt.
1972 * configure: Rebuilt.
1973 * configure.ac: Test assembler for view support.
1974 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1975 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1976 * dwarf2out.c (var_loc_view): New typedef.
1977 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1978 (dwarf2out_locviews_in_attribute): New.
1979 (dwarf2out_locviews_in_loclist): New.
1980 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1981 (enum dw_line_info_opcode): Add LI_adv_address.
1982 (struct dw_line_info_table): Add view.
1983 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1984 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1985 (zero_view_p): New variable.
1986 (ZERO_VIEW_P): New macro.
1987 (output_asm_line_debug_info): New.
1988 (struct var_loc_node): Add view.
1989 (add_AT_view_list, AT_loc_list): New.
1990 (add_var_loc_to_decl): Add view param. Test it against last.
1991 (new_loc_list): Add view params. Record them.
1992 (AT_loc_list_ptr): Handle loc and view lists.
1993 (view_list_to_loc_list_val_node): New.
1994 (print_dw_val): Handle dw_val_class_view_list.
1995 (size_of_die): Likewise.
1996 (value_format): Likewise.
1997 (loc_list_has_views): New.
1998 (gen_llsym): Set vl_symbol too.
1999 (maybe_gen_llsym, skip_loc_list_entry): New.
2000 (dwarf2out_maybe_output_loclist_view_pair): New.
2001 (output_loc_list): Output view list or entries too.
2002 (output_view_list_offset): New.
2003 (output_die): Handle dw_val_class_view_list.
2004 (output_dwarf_version): New.
2005 (output_compilation_unit_header): Use it.
2006 (output_skeleton_debug_sections): Likewise.
2007 (output_rnglists, output_line_info): Likewise.
2008 (output_pubnames, output_aranges): Update version comments.
2009 (output_one_line_info_table): Output view numbers in asm comments.
2010 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2011 Call maybe_gen_llsym.
2012 (loc_list_from_tree_1): Adjust.
2013 (add_AT_location_description): Create view list attribute if
2014 needed, check it's absent otherwise.
2015 (convert_cfa_to_fb_loc_list): Adjust.
2016 (maybe_emit_file): Call output_asm_line_debug_info for test.
2017 (dwarf2out_var_location): Reset views as needed. Precompute
2018 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2019 attribute. Set view.
2020 (new_line_info_table): Reset next view.
2021 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2022 (dwarf2out_source_line): Likewise. Output view resets and labels to
2023 the assembler, or select appropriate line info opcodes.
2024 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2025 (optimize_string_length): Catch it. Adjust.
2026 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2027 dw_val_class_view_list, and remove it if no longer needed.
2028 (hash_loc_list): Hash view numbers.
2029 (loc_list_hasher::equal): Compare them.
2030 (optimize_location_lists): Check whether a view list symbol is
2031 needed, and whether the locview attribute is present, and
2032 whether they match. Remove the locview attribute if no longer
2034 (index_location_lists): Call skip_loc_list_entry for test.
2035 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2036 Use output_dwarf_version.
2037 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2038 (struct dw_val_node): Add val_view_list.
2039 * final.c (SEEN_NEXT_VIEW): New.
2040 (set_next_view_needed): New.
2041 (clear_next_view_needed): New.
2042 (maybe_output_next_view): New.
2043 (final_start_function): Rename to...
2044 (final_start_function_1): ... this. Take pointer to FIRST,
2045 add SEEN parameter. Emit param bindings in the initial view.
2046 (final_start_function): Reintroduce SEEN-less interface.
2047 (final): Rename to...
2048 (final_1): ... this. Take SEEN parameter. Output final pending
2049 next view at the end.
2050 (final): Reintroduce seen-less interface.
2051 (final_scan_insn): Output pending next view before switching
2052 sections or ending a block. Mark the next view as needed when
2053 outputting variable locations. Notify debug backend of section
2054 changes, and of location view changes.
2055 (rest_of_handle_final): Adjust.
2056 * toplev.c (process_options): Autodetect value for debug variable
2057 location views option. Warn on incompat5 without -gdwarf-5.
2058 * doc/invoke.texi (gvariable-location-views): New.
2059 (gvariable-location-views=incompat5): New.
2060 (gno-variable-location-views): New.
2062 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2064 PR tree-optimization/84136
2065 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2066 that the result of find_edge is non-NULL.
2068 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2071 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2072 storing integer register in SImode. Fix cost of 256 and 512
2073 byte aligned SSE register store.
2075 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2077 * config/i386/i386.c (ix86_multiplication_cost): Fix
2078 multiplication cost for TARGET_AVX512DQ.
2080 2018-02-08 Marek Polacek <polacek@redhat.com>
2082 PR tree-optimization/84238
2083 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2086 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2088 PR tree-optimization/84265
2089 * tree-vect-stmts.c (vectorizable_store): Don't treat
2090 VMAT_CONTIGUOUS accesses as grouped.
2091 (vectorizable_load): Likewise.
2093 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2095 PR tree-optimization/81635
2096 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2097 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2098 (test_round_for_mask): New functions.
2099 (wide_int_cc_tests): Call test_round_for_mask.
2100 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2101 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2102 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2103 range returned by get_range_info.
2105 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2108 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2109 * symtab.c: Include builtins.h
2110 (symtab_node::output_to_lto_symbol_table_p): Move here
2111 from lto-streamer-out.c:output_symbol_p.
2112 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2113 (output_symbol_p): Move all logic to symtab.c
2114 (produce_symtab): Update.
2116 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2118 * config/s390/s390-opts.h (enum indirect_branch): Define.
2119 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2120 (s390_indirect_branch_via_thunk)
2121 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2122 (enum s390_indirect_branch_type): Define.
2123 * config/s390/s390.c (struct s390_frame_layout, struct
2124 machine_function): Remove.
2125 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2126 (indirect_branch_table_label_no, indirect_branch_table_name):
2128 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2129 (enum s390_indirect_branch_option): Define.
2130 (s390_return_addr_from_memory): New function.
2131 (s390_handle_string_attribute): New function.
2132 (s390_attribute_table): Add new attribute handler.
2133 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2134 (s390_indirect_branch_via_thunk): New function.
2135 (s390_indirect_branch_via_inline_thunk): New function.
2136 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2137 sibling call optimization for non z10 compiles.
2138 (s390_emit_call): Force indirect branch target to be a single
2139 register. Add r1 clobber for non-z10 compiles.
2140 (s390_emit_epilogue): Emit return jump via return_use expander.
2141 (s390_reorg): Handle JUMP_INSNs as execute targets.
2142 (s390_option_override_internal): Perform validity checks for the
2143 new command line options.
2144 (s390_indirect_branch_attrvalue): New function.
2145 (s390_indirect_branch_settings): New function.
2146 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2147 (s390_output_indirect_thunk_function): New function.
2148 (s390_code_end): Implement target hook.
2149 (s390_case_values_threshold): Implement target hook.
2150 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2152 * config/s390/s390.h (struct s390_frame_layout)
2153 (struct machine_function): Move here from s390.c.
2154 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2155 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2156 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2157 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2158 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2159 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2160 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2161 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2162 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2163 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2164 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2165 (mnemonic attribute): Add values which aren't recognized
2167 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2168 pattern for branch conversion. Fix mnemonic attribute.
2169 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2170 indirect branch via thunk if requested.
2171 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2172 ("*indirect_jump"): Disable for branch conversion using out of
2174 ("indirect_jump_via_thunk<mode>_z10")
2175 ("indirect_jump_via_thunk<mode>")
2176 ("indirect_jump_via_inlinethunk<mode>_z10")
2177 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2178 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2179 ("casesi_jump_via_inlinethunk<mode>_z10")
2180 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2181 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2182 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2183 ("*indirect2_jump"): Disable for branch conversion.
2184 ("casesi_jump"): Turn into expander and expand patterns for branch
2186 ("return_use"): New expander.
2187 ("*return"): Emit return via thunk and rename it to ...
2188 ("*return<mode>"): ... this one.
2189 * config/s390/s390.opt: Add new options and and enum for the
2192 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2194 * lra-constraints.c (match_reload): Unconditionally use
2195 gen_lowpart_SUBREG, rather than selecting between that
2196 and equivalent gen_rtx_SUBREG code.
2198 2018-02-08 Richard Biener <rguenther@suse.de>
2200 PR tree-optimization/84233
2201 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2202 changed flag instead of boguously re-using phi_inserted.
2204 2018-02-08 Martin Jambor <mjambor@suse.cz>
2206 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2207 static local variables.
2209 2018-02-08 Richard Biener <rguenther@suse.de>
2211 PR tree-optimization/84278
2212 * tree-vect-stmts.c (vectorizable_store): When looking for
2213 smaller vector types to perform grouped strided loads/stores
2214 make sure the mode is supported by the target.
2215 (vectorizable_load): Likewise.
2217 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2219 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2220 Increase LDP/STP opportunities by adding adjacent callee-saves.
2222 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2224 PR rtl-optimization/84068
2225 PR rtl-optimization/83459
2226 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2228 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2230 PR tree-optimization/84224
2231 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2232 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2235 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2238 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2239 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2240 count, remove one USE.
2242 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2244 * doc/install.texi (Configuration): Document the
2245 --with-long-double-format={ibm,ieee} PowerPC configuration
2249 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2250 Convert from define_expand to be define_insn_and_split. Rework
2251 float/double/_Float128 conversions to QI/HI/SImode to work with
2252 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2253 conversions to QI/HImode types did a store and then a load to
2254 truncate the value. For conversions to VSX registers, don't split
2255 the insn, instead emit the code directly. Use the code iterator
2256 any_fix to combine signed and unsigned conversions.
2257 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2258 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2259 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2260 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2261 (fix_<mode>di2_hw): Likewise.
2262 (fixuns_<mode>di2_hw): Likewise.
2263 (fix_<mode>si2_hw): Likewise.
2264 (fixuns_<mode>si2_hw): Likewise.
2265 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2266 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2267 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2268 fix<uns>_trunc<SFDF:mode>si2_p8.
2269 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2271 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2272 (fix<uns>_<mode>_mem): Likewise.
2273 (fctiw<u>z_<mode>_mem): Likewise.
2274 (fix<uns>_<mode>_mem): Likewise.
2275 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2276 the register allocator from doing a direct move to the GPRs to do
2277 a store, and instead use the ISA 3.0 store byte/half-word from
2278 vector register instruction. For IEEE 128-bit floating point,
2279 also optimize stores of 32-bit ints.
2280 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2282 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2284 * genextract.c (push_pathstr_operand): New function to support
2286 (walk_rtx): Call push_pathstr_operand.
2287 (print_path): Support [a-zA-Z].
2289 2018-02-07 Richard Biener <rguenther@suse.de>
2291 PR tree-optimization/84037
2292 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2293 (cse_and_gimplify_to_preheader): Declare.
2294 (vect_get_place_in_interleaving_chain): Likewise.
2295 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2297 (_loop_vec_info::~_loop_vec_info): Delete it.
2298 (cse_and_gimplify_to_preheader): New function.
2299 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2300 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2301 (vectorizable_load): Likewise. For grouped stores always base
2302 the IV on the first element.
2303 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2304 condition before gimplifying.
2306 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2308 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2309 *DIV_EXPR and *MOD_EXPR.
2311 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2314 * config/i386/i386.c (ix86_option_override_internal): Mask out
2315 the CF_SET bit when checking -fcf-protection.
2317 2018-02-07 Tom de Vries <tom@codesourcery.com>
2320 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2323 2018-02-07 Richard Biener <rguenther@suse.de>
2325 PR tree-optimization/84204
2326 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2329 PR tree-optimization/84205
2330 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2331 special-case isl_ast_op_zdiv_r.
2333 PR tree-optimization/84223
2334 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2335 Only add conditions from within the region.
2336 (gather_bbs::after_dom_children): Adjust.
2338 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2341 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2342 * config/avr/avr.md: Only post-reload split REG-REG moves if
2343 either register is GENERAL_REG_P.
2345 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2347 PR tree-optimization/84235
2348 * tree-ssa-scopedtables.c
2349 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2350 if the subtraction is performed in floating point type where NaNs are
2351 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2352 build 1. Formatting fix.
2354 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2357 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2358 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2359 and skip it regardless of bb boundaries. Use CALL_P macro,
2360 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2361 unnecessarily, formatting fix.
2363 2018-02-06 Michael Collison <michael.collison@arm.com>
2365 * config/arm/thumb2.md:
2366 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2367 (*thumb_mov_notscc): Ditto.
2369 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2372 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2373 unsigned_fix, not "s".
2375 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2377 * configure.ac (gcc_fn_eh_frame_ro): New function.
2378 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2379 correct .eh_frame permissions.
2380 * configure: Regenerate.
2382 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2384 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2387 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2389 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2390 Display warning message for -mno-speculate-indirect-jumps.
2392 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2394 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2396 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2398 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2400 PR tree-optimization/84225
2401 * tree-eh.c (find_trapping_overflow): Only call
2402 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2404 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2407 * config/i386/i386.c: Reimplement the check of possible options
2408 -mibt/-mshstk conbination. Change error messages.
2409 * doc/invoke.texi: Fix a typo: remove extra '='.
2411 2018-02-06 Marek Polacek <polacek@redhat.com>
2413 PR tree-optimization/84228
2414 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2416 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2419 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2420 emitted arch directives.
2421 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2422 __ARM_FEATURE_COPROC before changing architectures.
2424 2018-02-06 Richard Biener <rguenther@suse.de>
2426 * config/i386/i386.c (print_reg): Fix typo.
2427 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2429 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2431 * configure: Regenerate.
2433 2018-02-05 Martin Sebor <msebor@redhat.com>
2435 PR tree-optimization/83369
2436 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2439 2018-02-05 Martin Liska <mliska@suse.cz>
2441 * doc/invoke.texi: Cherry-pick upstream r323995.
2443 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2445 * ira.c (ira_init_register_move_cost): Adjust comment.
2447 2018-02-05 Martin Liska <mliska@suse.cz>
2449 PR gcov-profile/84137
2450 * doc/gcov.texi: Fix typo in documentation.
2452 2018-02-05 Martin Liska <mliska@suse.cz>
2454 PR gcov-profile/83879
2455 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2456 using dlopen functionality.
2458 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2460 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2461 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2462 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2463 _mm_maskz_range_round_ss): New intrinsics.
2464 (__builtin_ia32_rangesd128_round)
2465 (__builtin_ia32_rangess128_round): Remove.
2466 (__builtin_ia32_rangesd128_mask_round,
2467 __builtin_ia32_rangess128_mask_round): New builtins.
2468 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2469 __builtin_ia32_rangess128_round): Remove.
2470 (__builtin_ia32_rangesd128_mask_round,
2471 __builtin_ia32_rangess128_mask_round): New builtins.
2472 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2473 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2474 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2475 "<round_saeonly_constraint>")): Changed to ...
2476 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2477 "<round_saeonly_scalar_constraint>")): ... this.
2478 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2479 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2480 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2481 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2482 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2484 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2486 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2488 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2489 Remove all values except native, 8540 and 8548.
2491 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2493 * config/i386/i386.c (ix86_output_function_return): Pass
2494 INVALID_REGNUM, instead of -1, as invalid register number to
2495 indirect_thunk_name and output_indirect_thunk.
2497 2018-02-02 Julia Koval <julia.koval@intel.com>
2499 * config.gcc: Add -march=icelake.
2500 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2501 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2502 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2503 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2504 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2505 (processor_target_table): Add icelake.
2506 (ix86_option_override_internal): Handle new PTAs.
2507 (get_builtin_code_for_version): Handle icelake.
2508 (M_INTEL_COREI7_ICELAKE): New.
2509 (fold_builtin_cpu): Handle icelake.
2510 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2511 * doc/invoke.texi: Add -march=icelake.
2513 2018-02-02 Julia Koval <julia.koval@intel.com>
2515 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2516 to wide_int_bitmask.
2517 * wide-int-bitmask.h: New.
2519 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2522 * config/i386/i386.md: Replace Pmode with word_mode in
2523 builtin_setjmp_setup and builtin_longjmp to support x32.
2525 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2529 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2531 (rs6000_supported_cpu_names): New static variable.
2532 (linux_cpu_translation_table): Likewise.
2533 (elf_platform) <cpu>: Define new static variable and use it.
2534 Translate kernel AT_PLATFORM name to canonical name if needed.
2535 Error if platform name is unknown.
2537 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2540 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2542 2018-02-01 Jeff Law <law@redhat.com>
2545 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2546 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2547 the scratch if RELEASE_VIA_POP is false.
2548 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2549 If we have to save a temporary register, decrement SIZE appropriately.
2550 Pass new arguments to release_scratch_register_on_entry.
2551 (ix86_adjust_stack_and_probe): Likewise.
2552 (ix86_emit_probe_stack_range): Pass new arguments to
2553 release_scratch_register_on_entry.
2555 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2557 PR rtl-optimization/84157
2558 * combine.c (change_zero_ext): Use REG_P predicate in
2559 front of HARD_REGISTER_P predicate.
2561 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2563 * config/avr/avr.c (avr_option_override): Move disabling of
2564 -fdelete-null-pointer-checks to...
2565 * common/config/avr/avr-common.c (avr_option_optimization_table):
2568 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2570 PR tree-optimization/81635
2571 * tree-data-ref.c (split_constant_offset_1): For types that
2572 wrap on overflow, try to use range info to prove that wrapping
2575 2018-02-01 Renlin Li <renlin.li@arm.com>
2578 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2580 (aarch64_register_move_cost): Likewise.
2581 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2583 (REG_CLASS_NAMES): Likewise.
2584 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2585 TAILCALL_ADDR_REGS. Remove IP registers.
2586 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2588 2018-02-01 Richard Biener <rguenther@suse.de>
2590 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2591 for specifying RPO order and allow NULL for that.
2592 * domwalk.c (dom_walker::dom_walker): Likewise.
2593 (dom_walker::walk): Handle NULL RPO order.
2594 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2596 (rewrite_update_dom_walker): Likewise.
2597 (mark_def_dom_walker): Likewise.
2599 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2601 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2602 (aarch64_maybe_expand_sve_subreg_move): Declare.
2603 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2604 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2606 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2607 that are semantically a reverse operation.
2608 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2609 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2610 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2612 (aarch64_can_change_mode_class): For big-endian, forbid changes
2613 between two SVE modes if they have different element sizes.
2615 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2617 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2618 the TImode handling for big-endian targets.
2620 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2622 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2623 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2625 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2626 Remove BSWAP handing for big-endian targets and use the form of
2627 LD1RQ appropariate for the mode.
2629 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2631 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2632 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2635 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2638 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2639 check for operands that need to go through aarch64_sve_reload_be.
2641 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2643 PR tree-optimization/81661
2644 PR tree-optimization/84117
2645 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2646 * tree-eh.c: Include gimplify.h.
2647 (find_trapping_overflow, replace_trapping_overflow,
2648 rewrite_to_non_trapping_overflow): New functions.
2649 * tree-vect-loop.c: Include tree-eh.h.
2650 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2651 * tree-data-ref.c: Include tree-eh.h.
2652 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2654 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2656 PR rtl-optimization/84123
2657 * combine.c (change_zero_ext): Check if hard register satisfies
2658 can_change_dest_mode before calling gen_lowpart_SUBREG.
2660 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2663 * ira.c (ira_init_register_move_cost): Remove assert.
2665 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2667 PR rtl-optimization/84071
2668 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2669 * doc/tm.texi: Regenerate.
2671 2018-01-31 Richard Biener <rguenther@suse.de>
2673 PR tree-optimization/84132
2674 * tree-data-ref.c (analyze_miv_subscript): Properly
2675 check whether evolution_function_is_affine_multivariate_p
2676 before calling gcd_of_steps_may_divide_p.
2678 2018-01-31 Julia Koval <julia.koval@intel.com>
2681 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2682 * config/i386/i386.md (rdpid_rex64) New.
2683 (rdpid): Make 32bit only.
2685 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2688 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2689 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2691 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2694 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2696 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2698 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2700 PR rtl-optimization/84071
2701 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2702 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2704 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2706 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2707 (arc_attribute_table): Add 'aux' attribute.
2708 (arc_in_small_data_p): Consider aux like variables.
2709 (arc_is_aux_reg_p): New function.
2710 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2711 (arc_get_aux_arg): New function.
2712 (prepare_move_operands): Handle aux-register access.
2713 (arc_handle_aux_attribute): New function.
2714 * doc/extend.texi (ARC Variable attributes): Add subsection.
2716 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2718 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2719 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2720 (arc_attribute_table): Add 'uncached' attribute.
2721 (arc_print_operand): Print '.di' flag for uncached memory
2723 (arc_in_small_data_p): Do not consider for small data the uncached
2725 (arc_is_uncached_mem_p): New function.
2726 * config/arc/predicates.md (compact_store_memory_operand): Check
2727 for uncached memory accesses.
2728 (nonvol_nonimm_operand): Likewise.
2729 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2731 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2734 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2735 falign-loops=): Add Optimization flag.
2737 2018-01-30 Jeff Law <law@redhat.com>
2740 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2741 INT_REGISTERS_SAVED. Check it prior to calling
2742 get_scratch_register_on_entry.
2743 (ix86_adjust_stack_and_probe): Similarly.
2744 (ix86_emit_probe_stack_range): Similarly.
2745 (ix86_expand_prologue): Corresponding changes.
2747 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2750 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2751 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2753 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2756 * lra-constraints.c (curr_insn_transform): Process AND in the
2759 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2761 PR rtl-optimization/83986
2762 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2763 dependence against last_pending_memory_flush in addition to
2766 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2768 PR tree-optimization/81611
2769 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2772 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2775 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2778 2018-01-30 Richard Biener <rguenther@suse.de>
2779 Jakub Jelinek <jakub@redhat.com>
2781 PR tree-optimization/84111
2782 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2783 inner loops added during recursion, as they don't have up-to-date
2786 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2789 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2790 (can_inline_edge_by_limits_p): ... here.
2791 (can_early_inline_edge_p, check_callers,
2792 update_caller_keys, update_callee_keys, recursive_inlining,
2793 add_new_edges_to_heap, speculation_useful_p,
2794 inline_small_functions,
2795 inline_small_functions, flatten_function,
2796 inline_to_all_callers_1): Update.
2798 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2800 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2803 2018-01-30 Richard Biener <rguenther@suse.de>
2805 PR tree-optimization/83008
2806 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2807 invariant and constant vector uses in stmts when they need
2810 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2813 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2814 * configure: Regenerate.
2816 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2818 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2820 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2821 Use gen_rtx_REG rather than gen_lowpart.
2823 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2825 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2826 rather than 0 when creating partial subregs.
2828 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2830 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2833 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2836 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2837 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2838 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2839 flags. This restores the settings used before the 2017-07-24.
2840 Turning off pre increment/decrement/modify allows IVOPTS to
2841 optimize DF/SF loops where the index is an int.
2843 2018-01-29 Richard Biener <rguenther@suse.de>
2844 Kelvin Nilsen <kelvin@gcc.gnu.org>
2847 * tree-vect-stmts.c (vectorizable_call): Don't call
2848 targetm.vectorize_builtin_md_vectorized_function if callee is
2851 2018-01-22 Carl Love <cel@us.ibm.com>
2853 * doc/extend.tex: Fix typo in second arg in
2854 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2856 2018-01-29 Richard Biener <rguenther@suse.de>
2858 PR tree-optimization/84086
2859 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2860 (flush_ssaname_freelist): When SSA names were released reset
2861 the SCEV hash table.
2863 2018-01-29 Richard Biener <rguenther@suse.de>
2865 PR tree-optimization/84057
2866 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2867 removed paths when removing edges.
2869 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2871 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2872 -mfunction-return=@var{choice}.
2874 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2877 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2879 (layout::print_source_line): Likewise.
2880 (test_get_line_width_without_trailing_whitespace): Add test cases.
2882 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2885 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2888 2018-01-26 Jim Wilson <jimw@sifive.com>
2890 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2892 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2895 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2897 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2898 and CMP + SUB-immediate -> SUBS.
2900 2018-01-26 Martin Sebor <msebor@redhat.com>
2902 PR tree-optimization/83896
2903 * tree-ssa-strlen.c (get_string_len): Rename...
2904 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2905 Avoid assuming length is constant.
2906 (handle_char_store): Use HOST_WIDE_INT for string length.
2908 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2911 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2912 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2914 2018-01-26 Richard Biener <rguenther@suse.de>
2916 PR rtl-optimization/84003
2917 * dse.c (record_store): Only record redundant stores when
2918 the earlier store aliases at least all accesses the later one does.
2920 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2922 PR rtl-optimization/83985
2923 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2924 REG_CFA_RESTORE insns.
2925 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2926 REG_CFA_RESTORE insns here.
2929 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2930 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2932 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2934 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2935 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2936 (arc_init): Likewise.
2937 (arc_override_options): Likewise.
2938 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2940 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2942 * config/arc/arc.h (TARGET_DBNZ): Define.
2943 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2944 properly set the tune attribute.
2945 (dbnz): Use TARGET_DBNZ guard.
2946 * config/arc/arc.opt (mtune): Add core3 option.
2948 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2950 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2951 recognize new pic like addresses.
2952 (arc_delegitimize_address): Clean up.
2954 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2956 * config/arc/arc-arches.def: Option mrf16 valid for all
2958 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2959 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2960 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2961 * config/arc/arc-tables.opt: Regenerate.
2962 * config/arc/arc.c (arc_conditional_register_usage): Handle
2963 reduced register file case.
2964 (arc_file_start): Set must have build attributes.
2965 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2967 * config/arc/arc.opt (mrf16): Add new option.
2968 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2969 * config/arc/genmultilib.awk: Handle new mrf16 option.
2970 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2971 * config/arc/t-multilib: Regenerate.
2972 * doc/invoke.texi (ARC Options): Document mrf16 option.
2974 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2976 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2977 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2978 (arc_attribute_table): Add 'secure_call' attribute.
2979 (arc_print_operand): Print secure call operand.
2980 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2982 (arc_is_secure_call_p): New function. * config/arc/arc.md
2983 (call_i): Add support for sjli instruction.
2984 (call_value_i): Likewise.
2985 * config/arc/constraints.md (Csc): New constraint.
2987 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2988 John Eric Martin <John.Martin@emmicro-us.com>
2990 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2991 * config/arc/arc.c (_arc_jli_section): New struct.
2992 (arc_jli_section): New type.
2993 (rc_jli_sections): New static variable.
2994 (arc_handle_jli_attribute): New function.
2995 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2996 (arc_file_end): New function.
2997 (TARGET_ASM_FILE_END): Define.
2998 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2999 (arc_add_jli_section): New function.
3000 (jli_call_scan): Likewise.
3001 (arc_reorg): Call jli_call_scan.
3002 (arc_output_addsi): Remove 'S' from printing asm operand.
3003 (arc_is_jli_call_p): New function.
3004 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3006 (movhi_insn): Likewise.
3007 (movsi_insn): Likewise.
3008 (movsi_set_cc_insn): Likewise.
3009 (loadqi_update): Likewise.
3010 (load_zeroextendqisi_update): Likewise.
3011 (load_signextendqisi_update): Likewise.
3012 (loadhi_update): Likewise.
3013 (load_zeroextendhisi_update): Likewise.
3014 (load_signextendhisi_update): Likewise.
3015 (loadsi_update): Likewise.
3016 (loadsf_update): Likewise.
3017 (movsicc_insn): Likewise.
3018 (bset_insn): Likewise.
3019 (bxor_insn): Likewise.
3020 (bclr_insn): Likewise.
3021 (bmsk_insn): Likewise.
3022 (bicsi3_insn): Likewise.
3023 (cmpsi_cc_c_insn): Likewise.
3024 (movsi_ne): Likewise.
3025 (movsi_cond_exec): Likewise.
3026 (clrsbsi2): Likewise.
3037 (call_i): Remove 'S' asm letter, add jli instruction.
3038 (call_value_i): Likewise.
3039 * config/arc/arc.op (mjli-always): New option.
3040 * config/arc/constraints.md (Cji): New constraint.
3041 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3043 (subsf3_fpx): Likewise.
3044 (mulsf3_fpx): Likewise.
3045 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3047 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3049 * doc/invoke.texi (ARC): Document mjli-always option.
3051 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3053 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3054 avoid addition with 0 and use incw and decw where possible.
3056 2018-01-26 Richard Biener <rguenther@suse.de>
3058 PR tree-optimization/81082
3059 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3060 association if it requires casting to unsigned.
3061 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3062 from fold_plusminus_mult_expr to catch important cases late when
3063 range info is available.
3065 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3067 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3068 * configure.ac (hidden_linkonce): New test.
3069 * configure: Regenerate.
3070 * config.in: Regenerate.
3072 2018-01-26 Julia Koval <julia.koval@intel.com>
3074 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3075 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3076 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3077 _mm_mask_bitshuffle_epi64_mask): Fix type.
3078 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3079 USI_FTYPE_V4DI_V4DI_USI): Remove.
3080 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3081 __builtin_ia32_vpshufbitqmb256_mask,
3082 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3083 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3084 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3086 2018-01-26 Alan Modra <amodra@gmail.com>
3089 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3090 UNSPEC_VBPERMQ. Sort other unspecs.
3092 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3094 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3096 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3099 * predict.c (drop_profile): Do not push/pop cfun; update also
3101 (handle_missing_profiles): Fix logic looking for zero profiles.
3103 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3106 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3107 on functions with #pragma omp declare simd or functions with simd
3109 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3110 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3111 Remove trailing \n from warning_at calls.
3113 2018-01-25 Tom de Vries <tom@codesourcery.com>
3116 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3117 for neutered workers.
3119 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3122 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3123 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3125 2018-01-24 Jeff Law <law@redhat.com>
3128 * i386.c (get_probe_interval): Move to earlier point.
3129 (ix86_compute_frame_layout): If -fstack-clash-protection and
3130 the frame is larger than the probe interval, then use pushes
3131 to save registers rather than reg->mem moves.
3132 (ix86_expand_prologue): Remove conditional for int_registers_saved
3135 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3138 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3139 min/max for never referenced object.
3141 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3144 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3146 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3147 attributes from DECL_ATTRIBUTES (decl) without affecting
3148 DECL_ATTRIBUTES (current_function_decl).
3149 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3150 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3152 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3154 PR tree-optimization/83979
3155 * fold-const.c (fold_comparison): Use constant_boolean_node
3156 instead of boolean_{true,false}_node.
3158 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3160 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3163 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3165 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3166 Simplify the clause that sets the length attribute.
3167 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3168 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3169 clause that sets the length attribute.
3170 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3172 2018-01-24 Tom de Vries <tom@codesourcery.com>
3175 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3176 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3177 Add strict parameter.
3178 (prevent_branch_around_nothing): Insert dummy insn between branch to
3179 label and label with no ptx insn inbetween.
3180 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3182 2018-01-24 Tom de Vries <tom@codesourcery.com>
3185 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3186 for neutered threads in warp.
3187 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3189 2018-01-24 Richard Biener <rguenther@suse.de>
3191 PR tree-optimization/83176
3192 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3195 2018-01-24 Richard Biener <rguenther@suse.de>
3197 PR tree-optimization/82819
3198 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3199 code generating pluses that are no-ops in the target precision.
3201 2018-01-24 Richard Biener <rguenther@suse.de>
3204 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3206 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3208 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3209 to merge probabilities.
3210 * predict.c (probably_never_executed): Also mark as cold functions
3211 with global 0 profile and guessed local profile.
3212 * profile-count.c (profile_probability::combine_with_count): New
3214 * profile-count.h (profile_probability::operator*,
3215 profile_probability::operator*=, profile_probability::operator/,
3216 profile_probability::operator/=): Reduce precision to adjusted
3217 and set value to guessed on contradictory divisions.
3218 (profile_probability::combine_with_freq): Remove.
3219 (profile_probability::combine_wiht_count): Declare.
3220 (profile_count::force_nonzero):: Set to adjusted.
3221 (profile_count::probability_in):: Set quality to adjusted.
3222 * tree-ssa-tail-merge.c (replace_block_by): Use
3225 2018-01-23 Andrew Waterman <andrew@sifive.com>
3226 Jim Wilson <jimw@sifive.com>
3228 * config/riscv/riscv.c (riscv_stack_boundary): New.
3229 (riscv_option_override): Set riscv_stack_boundary. Handle
3230 riscv_preferred_stack_boundary_arg.
3231 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3232 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3233 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3234 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3235 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3236 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3238 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3241 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3242 of struct ix86_frame.
3243 (ix86_expand_epilogue): Likewise. Add a local variable for
3244 the reg_save_offset field in struct ix86_frame.
3246 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3248 PR tree-optimization/82604
3249 * tree-loop-distribution.c (enum partition_kind): New enum item
3250 PKIND_PARTIAL_MEMSET.
3251 (partition_builtin_p): Support above new enum item.
3252 (generate_code_for_partition): Ditto.
3253 (compute_access_range): Differentiate cases that equality can be
3254 proven at all loops, the innermost loops or no loops.
3255 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3256 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3257 (finalize_partitions, distribute_loop): Don't fuse partition of
3258 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3259 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3262 2018-01-23 Martin Liska <mliska@suse.cz>
3264 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3265 order to ignore the predictor.
3266 (PRED_POLYMORPHIC_CALL): Likewise.
3267 (PRED_RECURSIVE_CALL): Likewise.
3269 2018-01-23 Martin Liska <mliska@suse.cz>
3271 * tree-profile.c (tree_profiling): Print function header to
3272 aware reader which function we are working on.
3273 * value-prof.c (gimple_find_values_to_profile): Do not print
3274 not interesting value histograms.
3276 2018-01-23 Martin Liska <mliska@suse.cz>
3278 * profile-count.h (enum profile_quality): Add
3279 profile_uninitialized as the first value. Do not number values
3280 as they are zero based.
3281 (profile_count::verify): Update sanity check.
3282 (profile_probability::verify): Likewise.
3284 2018-01-23 Nathan Sidwell <nathan@acm.org>
3286 * doc/invoke.texi (ffor-scope): Deprecate.
3288 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3290 PR tree-optimization/83510
3291 * domwalk.c (set_all_edges_as_executable): New function.
3292 (dom_walker::dom_walker): Convert bool param
3293 "skip_unreachable_blocks" to enum reachability. Move setup of
3294 edge flags to set_all_edges_as_executable and only do it when
3295 reachability is REACHABLE_BLOCKS.
3296 * domwalk.h (enum dom_walker::reachability): New enum.
3297 (dom_walker::dom_walker): Convert bool param
3298 "skip_unreachable_blocks" to enum reachability.
3299 (set_all_edges_as_executable): New decl.
3300 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3301 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3303 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3304 but converting true to REACHABLE_BLOCKS.
3305 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3307 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3308 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3309 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3311 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3312 if check_all_array_refs will be called.
3314 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3316 * tree.c (selftest::test_location_wrappers): Add more test
3319 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3321 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3322 (selftest::test_bit_in_range): Likewise.
3324 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3327 * doc/sourcebuild.texi (vect_float): Say that the selector
3328 only describes the situation when -funsafe-math-optimizations is on.
3329 (vect_float_strict): Document.
3331 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3333 PR tree-optimization/83965
3334 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3335 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3336 instead of checking only for a reduction.
3337 (vect_recog_widen_sum_pattern): Likewise.
3339 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3341 * predict.c (probably_never_executed): Only use precise profile info.
3342 (compute_function_frequency): Skip after inlining hack since we now
3343 have quality checking.
3345 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3347 * profile-count.h (profile_probability::very_unlikely,
3348 profile_probability::unlikely, profile_probability::even): Set
3349 precision to guessed.
3351 2018-01-23 Richard Biener <rguenther@suse.de>
3353 PR tree-optimization/83963
3354 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3355 Properly terminate dominator walk when crossing the exit edge not
3356 when visiting its source block.
3358 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3361 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3362 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3364 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3366 PR tree-optimization/83957
3367 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3368 semicolon after for body surrounded by braces.
3370 PR tree-optimization/83081
3371 * profile-count.h (profile_probability::split): New method.
3372 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3373 Use profile_probability::split.
3374 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3375 when splitting a single conditional jump into 2.
3377 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3379 PR tree-optimization/69452
3380 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3383 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3385 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3386 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3387 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3389 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3391 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3392 * config/rl78/rl78.md (movdi): New define_expand.
3393 * config/rl78/rl78.c (rl78_split_movdi): New function.
3395 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3398 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3400 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3401 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3402 128-bit to produce an UNSPEC move to get the double word with the
3403 signbit and then a shift directly to do signbit.
3404 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3405 implementation with a new version that just does either a direct
3406 move or a regular move. Move memory interface to separate insns.
3407 Move insns so they are next to the expander.
3408 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3409 with signbit move. Split big and little endian case.
3410 (signbit<mode>2_dm_mem_le): Likewise.
3411 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3412 (signbit<mode>2_dm2): Likewise.
3414 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3416 * config/rl78/rl78.md (anddi3): New define_expand.
3418 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3420 * config/rl78/rl78.md (umindi3): New define_expand.
3422 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3424 * config/rl78/rl78.md (smindi3): New define_expand.
3426 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3428 * config/rl78/rl78.md (smaxdi3): New define_expand.
3430 2018-01-22 Carl Love <cel@us.ibm.com>
3432 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3433 LVX_V1TI): Add macro expansion.
3434 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3435 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3436 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3437 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3438 Change check to determine if the instruction is a byte reversing
3439 entry. Fix typo in comment.
3440 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3441 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3442 Add def_builtin calls for new builtins.
3443 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3444 Add define_insn expansion.
3446 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3448 * config/rl78/rl78.md (umaxdi3): New define_expand.
3450 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3452 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3453 for non-QImode registers.
3455 2018-01-22 Richard Biener <rguenther@suse.de>
3457 PR tree-optimization/83963
3458 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3459 including the loop exit block.
3460 (scop_detection::merge_sese): Likewise.
3461 (scop_detection::add_scop): Do it here instead.
3463 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3465 * doc/sourcebuild.texi (arm_softfloat): Document.
3467 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3470 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3471 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3472 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3474 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3475 David Edelsohn <dje.gcc@gmail.com>
3478 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3479 Change "crset eq" to "crset 2".
3480 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3481 (*call_indirect_aix<mode>_nospec): Likewise.
3482 (*call_value_indirect_aix<mode>_nospec): Likewise.
3483 (*call_indirect_elfv2<mode>_nospec): Likewise.
3484 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3485 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3486 change assembly output from . to $.
3487 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3488 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3489 (*tablejump<mode>_internal1_nospec): Likewise.
3491 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3494 * config/sh/sh_optimize_sett_clrt.cc:
3495 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3497 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3499 PR tree-optimization/83940
3500 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3501 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3502 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3503 use it to pass back the definition type.
3504 (vect_check_store_rhs): Likewise rhs_dt_out.
3505 (vect_build_gather_load_calls): Add a mask_dt argument and use
3506 it instead of a call to vect_is_simple_use.
3507 (vectorizable_store): Update calls to vect_check_load_store_mask
3508 and vect_check_store_rhs. Use the dt returned by the latter instead
3509 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3510 instead of calls to vect_is_simple_use. Pass the scalar rather
3511 than the vector operand to vect_is_simple_use when handling
3512 second and subsequent copies of an rhs value.
3513 (vectorizable_load): Update calls to vect_check_load_store_mask
3514 and vect_build_gather_load_calls. Use the cached mask_dt and
3515 gs_info.offset_dt instead of calls to vect_is_simple_use.
3517 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3520 * tree-emutls.c: Include gimplify.h.
3521 (lower_emutls_2): New function.
3522 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3523 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3524 it before further processing.
3527 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3528 UINTVAL (trueop1) instead of INTVAL (op1).
3530 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3534 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3535 INCOMING_FRAME_SP_OFFSET if not defined.
3536 (scan_trace): Add ENTRY argument. If true and
3537 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3538 emit a note to adjust the CFA offset.
3539 (create_cfi_notes): Adjust scan_trace callers.
3540 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3541 INCOMING_FRAME_SP_OFFSET in the CIE.
3542 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3543 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3545 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3546 * doc/tm.texi: Regenerated.
3548 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3550 PR rtl-optimization/83147
3551 * lra-constraints.c (remove_inheritance_pseudos): Use
3552 lra_substitute_pseudo_within_insn.
3554 2018-01-19 Tom de Vries <tom@codesourcery.com>
3555 Cesar Philippidis <cesar@codesourcery.com>
3558 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3560 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3563 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3564 spaces for function labels.
3566 2018-01-19 Martin Liska <mliska@suse.cz>
3568 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3569 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3570 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3571 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3572 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3573 (PRED_CONST_RETURN): Change from 69 to 65.
3574 (PRED_NULL_RETURN): Change from 91 to 71.
3575 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3576 (PRED_LOOP_GUARD): Change from 66 to 73.
3578 2018-01-19 Martin Liska <mliska@suse.cz>
3580 * predict.c (predict_insn_def): Add new assert.
3581 (struct branch_predictor): Change type to signed integer.
3582 (test_prediction_value_range): Amend test to cover
3584 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3585 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3586 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3587 (PRED_LOOP_IV_COMPARE): Likewise.
3588 * predict.h (PROB_UNINITIALIZED): Define new constant.
3590 2018-01-19 Martin Liska <mliska@suse.cz>
3592 * predict.c (dump_prediction): Add new format for
3593 analyze_brprob.py script which is enabled with -details
3595 * profile-count.h (precise_p): New function.
3597 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3599 PR tree-optimization/83922
3600 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3601 there are no statements that need masking.
3602 (vect_active_double_reduction_p): New function.
3603 (vect_analyze_loop_operations): Use it when handling phis that
3604 are not in the loop header.
3606 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3608 PR tree-optimization/83914
3609 * tree-vect-loop.c (vectorizable_induction): Don't convert
3610 init_expr or apply the peeling adjustment for inductions
3611 that are nested within the vectorized loop.
3613 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3615 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3618 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3622 * function.h (gimplify_parameters): Add gimple_seq * argument.
3623 * function.c: Include gimple.h and options.h.
3624 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3625 for the added local temporaries if needed.
3626 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3627 if there are any parameter cleanups, wrap whole body into a
3628 try/finally with the cleanups.
3630 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3633 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3634 Use GET_MODE_CLASS for scalar floating point.
3636 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3640 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3641 Fix call of call_cgraph_insertion_hooks.
3643 2018-01-18 Martin Sebor <msebor@redhat.com>
3645 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3647 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3650 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3653 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3656 * common.opt: (-ffile-prefix-map): New option.
3657 * opts.c (common_handle_option): Defer it.
3658 * opts-global.c (handle_common_deferred_options): Handle it.
3659 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3660 * file-prefix-map.h: New file.
3661 (remap_debug_filename, add_debug_prefix_map): ...here.
3662 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3663 * final.c (debug_prefix_map, add_debug_prefix_map
3664 remap_debug_filename): Move to...
3665 * file-prefix-map.c: New file.
3666 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3667 generalize, get rid of alloca(), use strrchr() instead of strchr().
3668 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3669 Implement in terms of add_prefix_map().
3670 (remap_macro_filename, remap_debug_filename): Implement in term of
3672 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3673 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3674 * dbxout.c: Include file-prefix-map.h.
3675 * varasm.c: Likewise.
3676 * vmsdbgout.c: Likewise.
3677 * xcoffout.c: Likewise.
3678 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3679 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3680 * doc/invoke.texi (-ffile-prefix-map): Document.
3681 (-fdebug-prefix-map): Update description.
3683 2018-01-18 Martin Liska <mliska@suse.cz>
3685 * config/i386/i386.c (indirect_thunk_name): Document that also
3687 (output_indirect_thunk): Document why both instructions
3688 (pause and lfence) are generated.
3690 2018-01-18 Richard Biener <rguenther@suse.de>
3692 PR tree-optimization/83887
3693 * graphite-scop-detection.c
3694 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3695 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3696 (scop_detection::merge_sese): Re-implement with a flood-fill
3697 algorithm that properly finds a SESE region if it exists.
3699 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3702 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3703 pointer_diff optimizations use view_convert instead of convert.
3705 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3707 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3708 Generate different code for -mno-speculate-indirect-jumps.
3709 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3710 (*call_indirect_aix<mode>): Disable for
3711 -mno-speculate-indirect-jumps.
3712 (*call_indirect_aix<mode>_nospec): New define_insn.
3713 (*call_value_indirect_aix<mode>): Disable for
3714 -mno-speculate-indirect-jumps.
3715 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3716 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3717 -mno-speculate-indirect-jumps.
3718 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3720 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3722 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3723 long double type, set the flags for noting the default long double
3724 type, even if we don't pass or return a long double type.
3726 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3729 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3732 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3734 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3735 support for merge[hl].
3736 (fold_mergehl_helper): New helper function.
3737 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3738 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3739 (altivec_vmrglw_direct): Add xxmrglw insn.
3741 2018-01-17 Andrew Waterman <andrew@sifive.com>
3743 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3744 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3746 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3749 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3750 call the lto_location_cache before reading the
3751 DECL_SOURCE_LOCATION of the types.
3753 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3754 Richard Sandiford <richard.sandiford@linaro.org>
3756 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3757 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3758 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3759 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3760 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3762 * config/aarch64/constraints.md (aarch64_movti_operand):
3764 * config/aarch64/predicates.md (Uti): Add new constraint.
3766 2018-01-17 Carl Love <cel@us.ibm.com>
3768 * config/rs6000/vsx.md (define_expand xl_len_r,
3769 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3770 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3772 (define_expand, define_insn): Move the shift left from the
3773 define_insn to the define_expand for lxvl and stxvl instructions.
3774 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3775 and XL_LEN_R definitions to PURE.
3777 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3779 * config/i386/i386.c (indirect_thunk_name): Declare regno
3780 as unsigned int. Compare regno with INVALID_REGNUM.
3781 (output_indirect_thunk): Ditto.
3782 (output_indirect_thunk_function): Ditto.
3783 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3784 in the call to output_indirect_thunk_function.
3786 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3789 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3790 rather than the size of inner_type to determine the stack slot size
3791 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3793 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3796 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3799 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3801 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3802 endian Linux systems to optionally enable multilibs for selecting
3803 the long double type if the user configured an explicit type.
3804 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3805 have no long double multilibs if not defined.
3806 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3807 warn if the user used -mabi={ieee,ibm}longdouble and we built
3808 multilibs for long double.
3809 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3810 appropriate multilib option.
3811 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3813 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3814 for building long double multilibs.
3815 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3817 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3819 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3822 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3824 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3827 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3830 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3833 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3835 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3836 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3838 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3840 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3841 different rtl trees depending on TARGET_64BIT.
3842 (rs6000_gen_lvx): Likewise.
3844 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3846 * config/visium/visium.md (nop): Tweak comment.
3847 (hazard_nop): Likewise.
3849 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3851 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3852 -mspeculate-indirect-jumps.
3853 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3854 for -mno-speculate-indirect-jumps.
3855 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3856 (*call_value_indirect_elfv2<mode>): Disable for
3857 -mno-speculate-indirect-jumps.
3858 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3859 (indirect_jump): Emit different RTL for
3860 -mno-speculate-indirect-jumps.
3861 (*indirect_jump<mode>): Disable for
3862 -mno-speculate-indirect-jumps.
3863 (*indirect_jump<mode>_nospec): New define_insn.
3864 (tablejump): Emit different RTL for
3865 -mno-speculate-indirect-jumps.
3866 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3867 (tablejumpsi_nospec): New define_expand.
3868 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3869 (tablejumpdi_nospec): New define_expand.
3870 (*tablejump<mode>_internal1): Disable for
3871 -mno-speculate-indirect-jumps.
3872 (*tablejump<mode>_internal1_nospec): New define_insn.
3873 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3876 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3878 * caller-save.c (insert_save): Drop unnecessary parameter. All
3881 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3882 Richard Biener <rguenth@suse.de>
3885 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3886 return early, inline manually is_gimple_sizepos. Make sure if we
3887 call gimplify_expr we don't end up with a gimple constant.
3888 * tree.c (variably_modified_type_p): Don't return true for
3889 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3890 * gimplify.h (is_gimple_sizepos): Remove.
3892 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3894 PR tree-optimization/83857
3895 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3896 vectorizable_live_operation for pure SLP statements.
3897 (vectorizable_live_operation): Handle PHIs.
3899 2018-01-16 Richard Biener <rguenther@suse.de>
3901 PR tree-optimization/83867
3902 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3903 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3905 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3908 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3909 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3910 If off is not INTEGER_CST, issue a may not be aligned warning
3911 rather than isn't aligned. Use isn%'t rather than isn't.
3912 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3914 <case MULT_EXPR>: Improve the case when bottom and one of the
3915 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3916 operand, in that case check if the other operand is multiple of
3917 bottom divided by the INTEGER_CST operand.
3919 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3922 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3923 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3924 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3925 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3926 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3927 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3928 (pa_function_arg_size): New function.
3930 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3932 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3933 in a separate statement.
3935 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3937 PR tree-optimization/83847
3938 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3939 group gathers and scatters.
3941 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3943 PR rtl-optimization/86620
3944 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3946 PR rtl-optimization/83213
3947 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3948 to last if both are JUMP_INSNs.
3950 PR tree-optimization/83843
3951 * gimple-ssa-store-merging.c
3952 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3953 store_immediate_info for bswap/nop orig_stores.
3955 2018-01-15 Andrew Waterman <andrew@sifive.com>
3957 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3959 <UDIV>: Increase cost if !TARGET_DIV.
3961 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3963 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3964 (define_attr "cr_logical_3op"): New.
3965 (cceq_ior_compare): Adjust.
3966 (cceq_ior_compare_complement): Adjust.
3967 (*cceq_rev_compare): Adjust.
3968 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3969 (is_cracked_insn): Adjust.
3970 (insn_must_be_first_in_group): Adjust.
3971 * config/rs6000/40x.md: Adjust.
3972 * config/rs6000/440.md: Adjust.
3973 * config/rs6000/476.md: Adjust.
3974 * config/rs6000/601.md: Adjust.
3975 * config/rs6000/603.md: Adjust.
3976 * config/rs6000/6xx.md: Adjust.
3977 * config/rs6000/7450.md: Adjust.
3978 * config/rs6000/7xx.md: Adjust.
3979 * config/rs6000/8540.md: Adjust.
3980 * config/rs6000/cell.md: Adjust.
3981 * config/rs6000/e300c2c3.md: Adjust.
3982 * config/rs6000/e500mc.md: Adjust.
3983 * config/rs6000/e500mc64.md: Adjust.
3984 * config/rs6000/e5500.md: Adjust.
3985 * config/rs6000/e6500.md: Adjust.
3986 * config/rs6000/mpc.md: Adjust.
3987 * config/rs6000/power4.md: Adjust.
3988 * config/rs6000/power5.md: Adjust.
3989 * config/rs6000/power6.md: Adjust.
3990 * config/rs6000/power7.md: Adjust.
3991 * config/rs6000/power8.md: Adjust.
3992 * config/rs6000/power9.md: Adjust.
3993 * config/rs6000/rs64.md: Adjust.
3994 * config/rs6000/titan.md: Adjust.
3996 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3998 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3999 ix86_indirect_branch_register logic.
4001 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4003 * config/i386/constraints.md (Bs): Update
4004 ix86_indirect_branch_register check. Don't check
4005 ix86_indirect_branch_register with GOT_memory_operand.
4007 * config/i386/predicates.md (GOT_memory_operand): Don't check
4008 ix86_indirect_branch_register here.
4009 (GOT32_symbol_operand): Likewise.
4011 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4013 * config/i386/predicates.md (constant_call_address_operand):
4014 Rewrite ix86_indirect_branch_register logic.
4015 (sibcall_insn_operand): Likewise.
4017 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4019 * config/i386/constraints.md (Bs): Replace
4020 ix86_indirect_branch_thunk_register with
4021 ix86_indirect_branch_register.
4023 * config/i386/i386.md (indirect_jump): Likewise.
4024 (tablejump): Likewise.
4025 (*sibcall_memory): Likewise.
4026 (*sibcall_value_memory): Likewise.
4027 Peepholes of indirect call and jump via memory: Likewise.
4028 * config/i386/i386.opt: Likewise.
4029 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4030 (GOT_memory_operand): Likewise.
4031 (call_insn_operand): Likewise.
4032 (sibcall_insn_operand): Likewise.
4033 (GOT32_symbol_operand): Likewise.
4035 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4038 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4039 type rather than type addr's type points to.
4040 (expand_omp_atomic_mutex): Likewise.
4041 (expand_omp_atomic): Likewise.
4043 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4046 * config/i386/i386.c (output_indirect_thunk_function): Use
4047 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4048 for __x86_return_thunk.
4050 2018-01-15 Richard Biener <rguenther@suse.de>
4053 * expmed.c (extract_bit_field_1): Fix typo.
4055 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4058 * config/arm/iterators.md (VF): New mode iterator.
4059 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4060 Remove integer-related logic from pattern.
4061 (neon_vabd<mode>_3): Likewise.
4063 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4066 * common.opt (fstrict-overflow): No longer an alias.
4067 (fwrapv-pointer): New option.
4068 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4069 also for pointer types based on flag_wrapv_pointer.
4070 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4071 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4072 opts->x_flag_wrapv got set.
4073 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4074 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4075 POINTER_TYPE_OVERFLOW_UNDEFINED.
4076 * match.pd: Likewise in address comparison pattern.
4077 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4079 2018-01-15 Richard Biener <rguenther@suse.de>
4082 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4083 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4084 Reset type names to their identifier if their TYPE_DECL doesn't
4085 have linkage (and thus is used for ODR and devirt).
4086 (save_debug_info_for_decl): Remove.
4087 (save_debug_info_for_type): Likewise.
4088 (add_tree_to_fld_list): Adjust.
4089 * tree-pretty-print.c (dump_generic_node): Make dumping of
4090 type names more robust.
4092 2018-01-15 Richard Biener <rguenther@suse.de>
4094 * BASE-VER: Bump to 8.0.1.
4096 2018-01-14 Martin Sebor <msebor@redhat.com>
4099 * builtins.c (check_access): Avoid warning when the no-warning bit
4102 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4104 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4105 * ira-color (allocno_hard_regs_compare): Likewise.
4107 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4110 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4111 Use .pushsection/.popsection.
4113 2018-01-14 Martin Sebor <msebor@redhat.com>
4116 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4118 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4120 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4121 entry from extra_headers.
4122 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4123 extra_headers, make the list bitwise identical to the i?86-*-* one.
4125 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4127 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4128 -mcmodel=large with -mindirect-branch=thunk,
4129 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4130 -mfunction-return=thunk-extern.
4131 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4132 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4133 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4135 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4137 * config/i386/i386.c (print_reg): Print the name of the full
4138 integer register without '%'.
4139 (ix86_print_operand): Handle 'V'.
4140 * doc/extend.texi: Document 'V' modifier.
4142 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4144 * config/i386/constraints.md (Bs): Disallow memory operand for
4145 -mindirect-branch-register.
4147 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4148 (GOT_memory_operand): Likewise.
4149 (call_insn_operand): Likewise.
4150 (sibcall_insn_operand): Likewise.
4151 (GOT32_symbol_operand): Likewise.
4152 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4153 for -mindirect-branch-register.
4154 (tablejump): Likewise.
4155 (*sibcall_memory): Likewise.
4156 (*sibcall_value_memory): Likewise.
4157 Disallow peepholes of indirect call and jump via memory for
4158 -mindirect-branch-register.
4159 (*call_pop): Replace m with Bw.
4160 (*call_value_pop): Likewise.
4161 (*sibcall_pop_memory): Replace m with Bs.
4162 * config/i386/i386.opt (mindirect-branch-register): New option.
4163 * doc/invoke.texi: Document -mindirect-branch-register option.
4165 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4167 * config/i386/i386-protos.h (ix86_output_function_return): New.
4168 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4169 set function_return_type.
4170 (indirect_thunk_name): Add ret_p to indicate thunk for function
4172 (output_indirect_thunk_function): Pass false to
4173 indirect_thunk_name.
4174 (ix86_output_indirect_branch_via_reg): Likewise.
4175 (ix86_output_indirect_branch_via_push): Likewise.
4176 (output_indirect_thunk_function): Create alias for function
4177 return thunk if regno < 0.
4178 (ix86_output_function_return): New function.
4179 (ix86_handle_fndecl_attribute): Handle function_return.
4180 (ix86_attribute_table): Add function_return.
4181 * config/i386/i386.h (machine_function): Add
4182 function_return_type.
4183 * config/i386/i386.md (simple_return_internal): Use
4184 ix86_output_function_return.
4185 (simple_return_internal_long): Likewise.
4186 * config/i386/i386.opt (mfunction-return=): New option.
4187 (indirect_branch): Mention -mfunction-return=.
4188 * doc/extend.texi: Document function_return function attribute.
4189 * doc/invoke.texi: Document -mfunction-return= option.
4191 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4193 * config/i386/i386-opts.h (indirect_branch): New.
4194 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4195 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4196 with local indirect jump when converting indirect call and jump.
4197 (ix86_set_indirect_branch_type): New.
4198 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4199 (indirectlabelno): New.
4200 (indirect_thunk_needed): Likewise.
4201 (indirect_thunk_bnd_needed): Likewise.
4202 (indirect_thunks_used): Likewise.
4203 (indirect_thunks_bnd_used): Likewise.
4204 (INDIRECT_LABEL): Likewise.
4205 (indirect_thunk_name): Likewise.
4206 (output_indirect_thunk): Likewise.
4207 (output_indirect_thunk_function): Likewise.
4208 (ix86_output_indirect_branch_via_reg): Likewise.
4209 (ix86_output_indirect_branch_via_push): Likewise.
4210 (ix86_output_indirect_branch): Likewise.
4211 (ix86_output_indirect_jmp): Likewise.
4212 (ix86_code_end): Call output_indirect_thunk_function if needed.
4213 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4215 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4216 (ix86_attribute_table): Add indirect_branch.
4217 * config/i386/i386.h (machine_function): Add indirect_branch_type
4218 and has_local_indirect_jump.
4219 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4221 (tablejump): Likewise.
4222 (*indirect_jump): Use ix86_output_indirect_jmp.
4223 (*tablejump_1): Likewise.
4224 (simple_return_indirect_internal): Likewise.
4225 * config/i386/i386.opt (mindirect-branch=): New option.
4226 (indirect_branch): New.
4229 (thunk-inline): Likewise.
4230 (thunk-extern): Likewise.
4231 * doc/extend.texi: Document indirect_branch function attribute.
4232 * doc/invoke.texi: Document -mindirect-branch= option.
4234 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4237 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4239 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4241 * ipa-inline.c (want_inline_small_function_p): Return false if
4242 inlining has already failed with CIF_FINAL_ERROR.
4243 (update_caller_keys): Call want_inline_small_function_p before
4245 (update_callee_keys): Likewise.
4247 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4249 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4251 (rs6000_quadword_masked_address_p): Likewise.
4252 (quad_aligned_load_p): Likewise.
4253 (quad_aligned_store_p): Likewise.
4254 (const_load_sequence_p): Add comment to describe the outer-most loop.
4255 (mimic_memory_attributes_and_flags): New function.
4256 (rs6000_gen_stvx): Likewise.
4257 (replace_swapped_aligned_store): Likewise.
4258 (rs6000_gen_lvx): Likewise.
4259 (replace_swapped_aligned_load): Likewise.
4260 (replace_swapped_load_constant): Capitalize argument name in
4261 comment describing this function.
4262 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4263 and stores that access quad-word aligned addresses and replace
4264 with stvx or lvx instructions when appropriate.
4265 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4266 New function prototype.
4267 (rs6000_quadword_masked_address_p): Likewise.
4268 (rs6000_gen_lvx): Likewise.
4269 (rs6000_gen_stvx): Likewise.
4270 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4271 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4272 when memory address is aligned.
4273 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4274 this split to select lvx instruction when memory address is aligned.
4275 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4276 instruction when memory address is aligned.
4277 (*vsx_le_perm_load_v16qi): Likewise.
4278 (four unnamed splitters): Modify to select the stvx instruction
4279 when memory is aligned.
4281 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4283 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4284 which appears in the queue multiple times.
4286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4287 Alan Hayward <alan.hayward@arm.com>
4288 David Sherwood <david.sherwood@arm.com>
4290 * tree-vectorizer.h (vec_lower_bound): New structure.
4291 (_loop_vec_info): Add check_nonzero and lower_bounds.
4292 (LOOP_VINFO_CHECK_NONZERO): New macro.
4293 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4294 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4295 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4296 fields. Make seg_len the distance travelled, not including the
4298 (dr_direction_indicator): Declare.
4299 (dr_zero_step_indicator): Likewise.
4300 (dr_known_forward_stride_p): Likewise.
4301 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4303 (runtime_alias_check_p): Allow runtime alias checks with
4305 (operator ==): Compare access_size and align.
4306 (prune_runtime_alias_test_list): Rework for new distinction between
4307 the access_size and seg_len.
4308 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4310 (get_segment_min_max): New function.
4311 (create_intersect_range_checks): Use it.
4312 (dr_step_indicator): New function.
4313 (dr_direction_indicator): Likewise.
4314 (dr_zero_step_indicator): Likewise.
4315 (dr_known_forward_stride_p): Likewise.
4316 * tree-loop-distribution.c (data_ref_segment_size): Return
4317 DR_STEP * (niters - 1).
4318 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4320 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4321 (vect_preserves_scalar_order_p): New function, split out from...
4322 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4323 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4324 (vect_vfa_access_size): New function.
4325 (vect_vfa_align): Likewise.
4326 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4327 (dump_lower_bound): New function.
4328 (vect_check_lower_bound): Likewise.
4329 (vect_small_gap_p): Likewise.
4330 (vectorizable_with_step_bound_p): Likewise.
4331 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4332 depencies if the vectorization factor is 1. Convert the checks
4333 for nonzero steps into checks on the bounds of DR_STEP. Try using
4334 a bunds check for variable steps if the minimum required step is
4335 relatively small. Update calls to the dr_with_seg_len
4336 constructor and to vect_compile_time_alias.
4337 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4339 (vect_loop_versioning): Call it.
4340 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4342 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4345 Alan Hayward <alan.hayward@arm.com>
4346 David Sherwood <david.sherwood@arm.com>
4348 * doc/sourcebuild.texi (vect_scatter_store): Document.
4349 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4351 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4353 * genopinit.c (main): Add supports_vec_scatter_store and
4354 supports_vec_scatter_store_cached to target_optabs.
4355 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4356 IFN_MASK_SCATTER_STORE.
4357 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4359 * internal-fn.h (internal_store_fn_p): Declare.
4360 (internal_fn_stored_value_index): Likewise.
4361 * internal-fn.c (scatter_store_direct): New macro.
4362 (expand_scatter_store_optab_fn): New function.
4363 (direct_scatter_store_optab_supported_p): New macro.
4364 (internal_store_fn_p): New function.
4365 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4366 IFN_MASK_SCATTER_STORE.
4367 (internal_fn_mask_index): Likewise.
4368 (internal_fn_stored_value_index): New function.
4369 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4371 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4372 * optabs-query.c (supports_vec_scatter_store_p): New function.
4373 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4374 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4375 true for scatter stores.
4376 (vect_gather_scatter_fn_p): Handle scatter stores too.
4377 (vect_check_gather_scatter): Consider using scatter stores if
4378 supports_vec_scatter_store_p.
4379 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4381 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4382 internal_fn_stored_value_index.
4383 (check_load_store_masking): Handle scatter stores too.
4384 (vect_get_store_rhs): Make public.
4385 (vectorizable_call): Use internal_store_fn_p.
4386 (vectorizable_store): Handle scatter store internal functions.
4387 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4388 when deciding whether the end of the group has been reached.
4389 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4390 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4391 (mask_scatter_store<mode>): New insns.
4393 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4394 Alan Hayward <alan.hayward@arm.com>
4395 David Sherwood <david.sherwood@arm.com>
4397 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4398 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4399 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4401 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4402 Use vect_truncate_gather_scatter_offset if we can't treat the
4403 operation as a normal gather load or scatter store.
4404 (get_group_load_store_type): Take the gather_scatter_info
4405 as argument. Try using a gather load or scatter store for
4406 single-element groups.
4407 (get_load_store_type): Update calls to get_group_load_store_type
4408 and vect_use_strided_gather_scatters_p.
4410 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4411 Alan Hayward <alan.hayward@arm.com>
4412 David Sherwood <david.sherwood@arm.com>
4414 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4415 optional tree argument.
4416 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4418 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4419 but continue to use the current value as a fallback.
4420 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4421 to compare the updates.
4422 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4423 (get_load_store_type): Use it when handling a strided access.
4424 (vect_get_strided_load_store_ops): New function.
4425 (vect_get_data_ptr_increment): Likewise.
4426 (vectorizable_load): Handle strided gather loads. Always pass
4427 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4430 Alan Hayward <alan.hayward@arm.com>
4431 David Sherwood <david.sherwood@arm.com>
4433 * doc/md.texi (gather_load@var{m}): Document.
4434 (mask_gather_load@var{m}): Likewise.
4435 * genopinit.c (main): Add supports_vec_gather_load and
4436 supports_vec_gather_load_cached to target_optabs.
4437 * optabs-tree.c (init_tree_optimization_optabs): Use
4438 ggc_cleared_alloc to allocate target_optabs.
4439 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4440 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4442 * internal-fn.h (internal_load_fn_p): Declare.
4443 (internal_gather_scatter_fn_p): Likewise.
4444 (internal_fn_mask_index): Likewise.
4445 (internal_gather_scatter_fn_supported_p): Likewise.
4446 * internal-fn.c (gather_load_direct): New macro.
4447 (expand_gather_load_optab_fn): New function.
4448 (direct_gather_load_optab_supported_p): New macro.
4449 (direct_internal_fn_optab): New function.
4450 (internal_load_fn_p): Likewise.
4451 (internal_gather_scatter_fn_p): Likewise.
4452 (internal_fn_mask_index): Likewise.
4453 (internal_gather_scatter_fn_supported_p): Likewise.
4454 * optabs-query.c (supports_at_least_one_mode_p): New function.
4455 (supports_vec_gather_load_p): Likewise.
4456 * optabs-query.h (supports_vec_gather_load_p): Declare.
4457 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4458 and memory_type field.
4459 (NUM_PATTERNS): Bump to 15.
4460 * tree-vect-data-refs.c: Include internal-fn.h.
4461 (vect_gather_scatter_fn_p): New function.
4462 (vect_describe_gather_scatter_call): Likewise.
4463 (vect_check_gather_scatter): Try using internal functions for
4464 gather loads. Recognize existing calls to a gather load function.
4465 (vect_analyze_data_refs): Consider using gather loads if
4466 supports_vec_gather_load_p.
4467 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4468 (vect_get_gather_scatter_offset_type): Likewise.
4469 (vect_convert_mask_for_vectype): Likewise.
4470 (vect_add_conversion_to_patterm): Likewise.
4471 (vect_try_gather_scatter_pattern): Likewise.
4472 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4473 (vect_vect_recog_func_ptrs): Add it.
4474 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4475 internal_fn_mask_index and internal_gather_scatter_fn_p.
4476 (check_load_store_masking): Take the gather_scatter_info as an
4477 argument and handle gather loads.
4478 (vect_get_gather_scatter_ops): New function.
4479 (vectorizable_call): Check internal_load_fn_p.
4480 (vectorizable_load): Likewise. Handle gather load internal
4482 (vectorizable_store): Update call to check_load_store_masking.
4483 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4484 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4485 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4486 (aarch64_gather_scale_operand_d): New predicates.
4487 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4488 (mask_gather_load<mode>): New insns.
4490 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4491 Alan Hayward <alan.hayward@arm.com>
4492 David Sherwood <david.sherwood@arm.com>
4494 * optabs.def (fold_left_plus_optab): New optab.
4495 * doc/md.texi (fold_left_plus_@var{m}): Document.
4496 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4497 * internal-fn.c (fold_left_direct): Define.
4498 (expand_fold_left_optab_fn): Likewise.
4499 (direct_fold_left_optab_supported_p): Likewise.
4500 * fold-const-call.c (fold_const_fold_left): New function.
4501 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4502 * tree-parloops.c (valid_reduction_p): New function.
4503 (gather_scalar_reductions): Use it.
4504 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4505 (vect_finish_replace_stmt): Declare.
4506 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4507 (needs_fold_left_reduction_p): New function, split out from...
4508 (vect_is_simple_reduction): ...here. Accept reductions that
4509 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4510 (vect_force_simple_reduction): Also store the reduction type in
4511 the assignment's STMT_VINFO_REDUC_TYPE.
4512 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4513 (merge_with_identity): New function.
4514 (vect_expand_fold_left): Likewise.
4515 (vectorize_fold_left_reduction): Likewise.
4516 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4517 scalar phi in place for it. Check for target support and reject
4518 cases that would reassociate the operation. Defer the transform
4519 phase to vectorize_fold_left_reduction.
4520 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4521 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4522 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4526 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4527 call to ifc_temp_var.
4529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4530 Alan Hayward <alan.hayward@arm.com>
4531 David Sherwood <david.sherwood@arm.com>
4533 * target.def (legitimize_address_displacement): Take the original
4534 offset as a poly_int.
4535 * targhooks.h (default_legitimize_address_displacement): Update
4537 * targhooks.c (default_legitimize_address_displacement): Likewise.
4538 * doc/tm.texi: Regenerate.
4539 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4540 as an argument, moving assert of ad->disp == ad->disp_term to...
4541 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4542 Try calling targetm.legitimize_address_displacement before expanding
4543 the address rather than afterwards, and adjust for the new interface.
4544 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4545 Match the new hook interface. Handle SVE addresses.
4546 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4551 * Makefile.in (OBJS): Add early-remat.o.
4552 * target.def (select_early_remat_modes): New hook.
4553 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4554 * doc/tm.texi: Regenerate.
4555 * targhooks.h (default_select_early_remat_modes): Declare.
4556 * targhooks.c (default_select_early_remat_modes): New function.
4557 * timevar.def (TV_EARLY_REMAT): New timevar.
4558 * passes.def (pass_early_remat): New pass.
4559 * tree-pass.h (make_pass_early_remat): Declare.
4560 * early-remat.c: New file.
4561 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4563 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4566 Alan Hayward <alan.hayward@arm.com>
4567 David Sherwood <david.sherwood@arm.com>
4569 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4570 vfm1 with a bound_epilog parameter.
4571 (vect_do_peeling): Update calls accordingly, and move the prologue
4572 call earlier in the function. Treat the base bound_epilog as 0 for
4573 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4574 this base when peeling for gaps.
4575 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4576 with fully-masked loops.
4577 (vect_estimate_min_profitable_iters): Handle the single peeled
4578 iteration in that case.
4580 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4581 Alan Hayward <alan.hayward@arm.com>
4582 David Sherwood <david.sherwood@arm.com>
4584 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4585 single-element interleaving even if the size is not a power of 2.
4586 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4587 accesses for single-element interleaving if the group size is
4590 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4591 Alan Hayward <alan.hayward@arm.com>
4592 David Sherwood <david.sherwood@arm.com>
4594 * doc/md.texi (fold_extract_last_@var{m}): Document.
4595 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4596 * optabs.def (fold_extract_last_optab): New optab.
4597 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4598 * internal-fn.c (fold_extract_direct): New macro.
4599 (expand_fold_extract_optab_fn): Likewise.
4600 (direct_fold_extract_optab_supported_p): Likewise.
4601 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4602 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4603 EXTRACT_LAST_REDUCTION.
4604 (get_initial_def_for_reduction): Do not create an initial vector
4605 for EXTRACT_LAST_REDUCTION reductions.
4606 (vectorizable_reduction): Leave the scalar phi in place for
4607 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4608 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4609 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4610 transform phase to vectorizable_condition.
4611 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4613 (vect_finish_stmt_generation): ...here.
4614 (vect_finish_replace_stmt): New function.
4615 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4616 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4618 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4621 Alan Hayward <alan.hayward@arm.com>
4622 David Sherwood <david.sherwood@arm.com>
4624 * doc/md.texi (extract_last_@var{m}): Document.
4625 * optabs.def (extract_last_optab): New optab.
4626 * internal-fn.def (EXTRACT_LAST): New internal function.
4627 * internal-fn.c (cond_unary_direct): New macro.
4628 (expand_cond_unary_optab_fn): Likewise.
4629 (direct_cond_unary_optab_supported_p): Likewise.
4630 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4631 loops using EXTRACT_LAST.
4632 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4633 (extract_last_<mode>): ...this optab.
4634 (vec_extract<mode><Vel>): Update accordingly.
4636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4637 Alan Hayward <alan.hayward@arm.com>
4638 David Sherwood <david.sherwood@arm.com>
4640 * target.def (empty_mask_is_expensive): New hook.
4641 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4642 * doc/tm.texi: Regenerate.
4643 * targhooks.h (default_empty_mask_is_expensive): Declare.
4644 * targhooks.c (default_empty_mask_is_expensive): New function.
4645 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4646 if the target says that empty masks are expensive.
4647 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4649 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4651 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4652 Alan Hayward <alan.hayward@arm.com>
4653 David Sherwood <david.sherwood@arm.com>
4655 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4656 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4657 (vect_use_loop_mask_for_alignment_p): New function.
4658 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4659 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4660 niters_skip argument. Make sure that the first niters_skip elements
4661 of the first iteration are inactive.
4662 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4663 Update call to vect_set_loop_masks_directly.
4664 (get_misalign_in_elems): New function, split out from...
4665 (vect_gen_prolog_loop_niters): ...here.
4666 (vect_update_init_of_dr): Take a code argument that specifies whether
4667 the adjustment should be added or subtracted.
4668 (vect_update_init_of_drs): Likewise.
4669 (vect_prepare_for_masked_peels): New function.
4670 (vect_do_peeling): Skip prologue peeling if we're using a mask
4671 instead. Update call to vect_update_inits_of_drs.
4672 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4674 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4675 alignment. Do not include the number of peeled iterations in
4676 the minimum threshold in that case.
4677 (vectorizable_induction): Adjust the start value down by
4678 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4679 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4680 Take the number of skipped iterations into account when calculating
4682 * tree-vect-stmts.c (vect_gen_while_not): New function.
4684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4685 Alan Hayward <alan.hayward@arm.com>
4686 David Sherwood <david.sherwood@arm.com>
4688 * doc/sourcebuild.texi (vect_fully_masked): Document.
4689 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4691 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4693 (vect_analyze_loop_2): ...here. Don't check the vectorization
4694 factor against the number of loop iterations if the loop is
4697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4698 Alan Hayward <alan.hayward@arm.com>
4699 David Sherwood <david.sherwood@arm.com>
4701 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4702 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4703 (dump_groups): Update accordingly.
4704 (iv_use::mem_type): New member variable.
4705 (address_p): New function.
4706 (record_use): Add a mem_type argument and initialize the new
4708 (record_group_use): Add a mem_type argument. Use address_p.
4709 Remove obsolete null checks of base_object. Update call to record_use.
4710 (find_interesting_uses_op): Update call to record_group_use.
4711 (find_interesting_uses_cond): Likewise.
4712 (find_interesting_uses_address): Likewise.
4713 (get_mem_type_for_internal_fn): New function.
4714 (find_address_like_use): Likewise.
4715 (find_interesting_uses_stmt): Try find_address_like_use before
4716 calling find_interesting_uses_op.
4717 (addr_offset_valid_p): Use the iv mem_type field as the type
4718 of the addressed memory.
4719 (add_autoinc_candidates): Likewise.
4720 (get_address_cost): Likewise.
4721 (split_small_address_groups_p): Use address_p.
4722 (split_address_groups): Likewise.
4723 (add_iv_candidate_for_use): Likewise.
4724 (autoinc_possible_for_pair): Likewise.
4725 (rewrite_groups): Likewise.
4726 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4727 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4728 (get_alias_ptr_type_for_ptr_address): New function.
4729 (rewrite_use_address): Rewrite address uses in calls that were
4730 identified by find_address_like_use.
4732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4733 Alan Hayward <alan.hayward@arm.com>
4734 David Sherwood <david.sherwood@arm.com>
4736 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4738 * gimple-expr.h (is_gimple_addressable: Likewise.
4739 * gimple-expr.c (is_gimple_address): Likewise.
4740 * internal-fn.c (expand_call_mem_ref): New function.
4741 (expand_mask_load_optab_fn): Use it.
4742 (expand_mask_store_optab_fn): Likewise.
4744 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4745 Alan Hayward <alan.hayward@arm.com>
4746 David Sherwood <david.sherwood@arm.com>
4748 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4749 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4750 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4751 (cond_umax@var{mode}): Document.
4752 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4753 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4754 (cond_umin_optab, cond_umax_optab): New optabs.
4755 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4756 (COND_IOR, COND_XOR): New internal functions.
4757 * internal-fn.h (get_conditional_internal_fn): Declare.
4758 * internal-fn.c (cond_binary_direct): New macro.
4759 (expand_cond_binary_optab_fn): Likewise.
4760 (direct_cond_binary_optab_supported_p): Likewise.
4761 (get_conditional_internal_fn): New function.
4762 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4763 Cope with reduction statements that are vectorized as calls rather
4765 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4766 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4767 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4768 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4769 (UNSPEC_COND_EOR): New unspecs.
4770 (optab): Add mappings for them.
4771 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4772 (sve_int_op, sve_fp_op): New int attributes.
4774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4775 Alan Hayward <alan.hayward@arm.com>
4776 David Sherwood <david.sherwood@arm.com>
4778 * optabs.def (while_ult_optab): New optab.
4779 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4780 * internal-fn.def (WHILE_ULT): New internal function.
4781 * internal-fn.h (direct_internal_fn_supported_p): New override
4782 that takes two types as argument.
4783 * internal-fn.c (while_direct): New macro.
4784 (expand_while_optab_fn): New function.
4785 (convert_optab_supported_p): Likewise.
4786 (direct_while_optab_supported_p): New macro.
4787 * wide-int.h (wi::udiv_ceil): New function.
4788 * tree-vectorizer.h (rgroup_masks): New structure.
4789 (vec_loop_masks): New typedef.
4790 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4792 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4793 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4794 (vect_max_vf): New function.
4795 (slpeel_make_loop_iterate_ntimes): Delete.
4796 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4797 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4798 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4799 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4800 internal-fn.h, stor-layout.h and optabs-query.h.
4801 (vect_set_loop_mask): New function.
4802 (add_preheader_seq): Likewise.
4803 (add_header_seq): Likewise.
4804 (interleave_supported_p): Likewise.
4805 (vect_maybe_permute_loop_masks): Likewise.
4806 (vect_set_loop_masks_directly): Likewise.
4807 (vect_set_loop_condition_masked): Likewise.
4808 (vect_set_loop_condition_unmasked): New function, split out from
4809 slpeel_make_loop_iterate_ntimes.
4810 (slpeel_make_loop_iterate_ntimes): Rename to..
4811 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4812 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4813 (vect_do_peeling): Update call accordingly.
4814 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4816 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4817 mask_compare_type, can_fully_mask_p and fully_masked_p.
4818 (release_vec_loop_masks): New function.
4819 (_loop_vec_info): Use it to free the loop masks.
4820 (can_produce_all_loop_masks_p): New function.
4821 (vect_get_max_nscalars_per_iter): Likewise.
4822 (vect_verify_full_masking): Likewise.
4823 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4824 retries, and free the mask rgroups before retrying. Check loop-wide
4825 reasons for disallowing fully-masked loops. Make the final decision
4826 about whether use a fully-masked loop or not.
4827 (vect_estimate_min_profitable_iters): Do not assume that peeling
4828 for the number of iterations will be needed for fully-masked loops.
4829 (vectorizable_reduction): Disable fully-masked loops.
4830 (vectorizable_live_operation): Likewise.
4831 (vect_halve_mask_nunits): New function.
4832 (vect_double_mask_nunits): Likewise.
4833 (vect_record_loop_mask): Likewise.
4834 (vect_get_loop_mask): Likewise.
4835 (vect_transform_loop): Handle the case in which the final loop
4836 iteration might handle a partial vector. Call vect_set_loop_condition
4837 instead of slpeel_make_loop_iterate_ntimes.
4838 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4839 (check_load_store_masking): New function.
4840 (prepare_load_store_mask): Likewise.
4841 (vectorizable_store): Handle fully-masked loops.
4842 (vectorizable_load): Likewise.
4843 (supportable_widening_operation): Use vect_halve_mask_nunits for
4845 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4846 (vect_gen_while): New function.
4847 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4848 (aarch64_uqdec<mode>): New insn.
4850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4851 Alan Hayward <alan.hayward@arm.com>
4852 David Sherwood <david.sherwood@arm.com>
4854 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4855 (reduc_xor_scal_optab): New optabs.
4856 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4857 (reduc_xor_scal_@var{m}): Document.
4858 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4859 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4861 * fold-const-call.c (fold_const_call): Handle them.
4862 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4863 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4864 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4865 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4866 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4867 (UNSPEC_XORV): New unspecs.
4868 (optab): Add entries for them.
4869 (BITWISEV): New int iterator.
4870 (bit_reduc_op): New int attributes.
4872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4873 Alan Hayward <alan.hayward@arm.com>
4874 David Sherwood <david.sherwood@arm.com>
4876 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4877 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4878 * optabs.def (vec_shl_insert_optab): New optab.
4879 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4880 (duplicate_and_interleave): Likewise.
4881 * tree-vect-loop.c: Include internal-fn.h.
4882 (neutral_op_for_slp_reduction): New function, split out from
4883 get_initial_defs_for_reduction.
4884 (get_initial_def_for_reduction): Handle option 2 for variable-length
4885 vectors by loading the neutral value into a vector and then shifting
4886 the initial value into element 0.
4887 (get_initial_defs_for_reduction): Replace the code argument with
4888 the neutral value calculated by neutral_op_for_slp_reduction.
4889 Use gimple_build_vector for constant-length vectors.
4890 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4891 but the first group_size elements have a neutral value.
4892 Use duplicate_and_interleave otherwise.
4893 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4894 Update call to get_initial_defs_for_reduction. Handle SLP
4895 reductions for variable-length vectors by creating one vector
4896 result for each scalar result, with the elements associated
4897 with other scalar results stubbed out with the neutral value.
4898 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4899 Require IFN_VEC_SHL_INSERT for double reductions on
4900 variable-length vectors, or SLP reductions that have
4901 a neutral value. Require can_duplicate_and_interleave_p
4902 support for variable-length unchained SLP reductions if there
4903 is no neutral value, such as for MIN/MAX reductions. Also require
4904 the number of vector elements to be a multiple of the number of
4905 SLP statements when doing variable-length unchained SLP reductions.
4906 Update call to vect_create_epilog_for_reduction.
4907 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4908 and remove initial values.
4909 (duplicate_and_interleave): Make public.
4910 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4911 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4913 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4914 Alan Hayward <alan.hayward@arm.com>
4915 David Sherwood <david.sherwood@arm.com>
4917 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4918 (can_duplicate_and_interleave_p): New function.
4919 (vect_get_and_check_slp_defs): Take the vector of statements
4920 rather than just the current one. Remove excess parentheses.
4921 Restriction rejectinon of vect_constant_def and vect_external_def
4922 for variable-length vectors to boolean types, or types for which
4923 can_duplicate_and_interleave_p is false.
4924 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4925 (duplicate_and_interleave): New function.
4926 (vect_get_constant_vectors): Use gimple_build_vector for
4927 constant-length vectors and suitable variable-length constant
4928 vectors. Use duplicate_and_interleave for other variable-length
4929 vectors. Don't defer the update when inserting new statements.
4931 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4932 Alan Hayward <alan.hayward@arm.com>
4933 David Sherwood <david.sherwood@arm.com>
4935 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4936 min_profitable_iters doesn't go negative.
4938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4939 Alan Hayward <alan.hayward@arm.com>
4940 David Sherwood <david.sherwood@arm.com>
4942 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4943 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4944 * optabs.def (vec_mask_load_lanes_optab): New optab.
4945 (vec_mask_store_lanes_optab): Likewise.
4946 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4947 (MASK_STORE_LANES): Likewise.
4948 * internal-fn.c (mask_load_lanes_direct): New macro.
4949 (mask_store_lanes_direct): Likewise.
4950 (expand_mask_load_optab_fn): Handle masked operations.
4951 (expand_mask_load_lanes_optab_fn): New macro.
4952 (expand_mask_store_optab_fn): Handle masked operations.
4953 (expand_mask_store_lanes_optab_fn): New macro.
4954 (direct_mask_load_lanes_optab_supported_p): Likewise.
4955 (direct_mask_store_lanes_optab_supported_p): Likewise.
4956 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4958 (vect_load_lanes_supported): Likewise.
4959 * tree-vect-data-refs.c (strip_conversion): New function.
4960 (can_group_stmts_p): Likewise.
4961 (vect_analyze_data_ref_accesses): Use it instead of checking
4962 for a pair of assignments.
4963 (vect_store_lanes_supported): Take a masked_p parameter.
4964 (vect_load_lanes_supported): Likewise.
4965 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4966 vect_store_lanes_supported and vect_load_lanes_supported.
4967 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4968 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4969 parameter. Don't allow gaps for masked accesses.
4970 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4971 and vect_load_lanes_supported.
4972 (get_load_store_type): Take a masked_p parameter and update
4973 call to get_group_load_store_type.
4974 (vectorizable_store): Update call to get_load_store_type.
4975 Handle IFN_MASK_STORE_LANES.
4976 (vectorizable_load): Update call to get_load_store_type.
4977 Handle IFN_MASK_LOAD_LANES.
4979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4980 Alan Hayward <alan.hayward@arm.com>
4981 David Sherwood <david.sherwood@arm.com>
4983 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4985 * config/aarch64/aarch64-protos.h
4986 (aarch64_sve_struct_memory_operand_p): Declare.
4987 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4988 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4989 (VPRED, vpred): Handle SVE structure modes.
4990 * config/aarch64/constraints.md (Utx): New constraint.
4991 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4992 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4993 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4994 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4995 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4996 structure modes. Split into pieces after RA.
4997 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4998 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5000 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5001 SVE structure modes.
5002 (aarch64_classify_address): Likewise.
5003 (sizetochar): Move earlier in file.
5004 (aarch64_print_operand): Handle SVE register lists.
5005 (aarch64_array_mode): New function.
5006 (aarch64_sve_struct_memory_operand_p): Likewise.
5007 (TARGET_ARRAY_MODE): Redefine.
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 * target.def (array_mode): New target hook.
5014 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5015 * doc/tm.texi: Regenerate.
5016 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5017 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5018 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5020 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5024 Alan Hayward <alan.hayward@arm.com>
5025 David Sherwood <david.sherwood@arm.com>
5027 * fold-const.c (fold_binary_loc): Check the argument types
5028 rather than the result type when testing for a vector operation.
5030 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5032 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5033 * doc/tm.texi: Regenerate.
5035 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5036 Alan Hayward <alan.hayward@arm.com>
5037 David Sherwood <david.sherwood@arm.com>
5039 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5040 (sve): Document new AArch64 extension.
5041 * doc/md.texi (w): Extend the description of the AArch64
5042 constraint to include SVE vectors.
5043 (Upl, Upa): Document new AArch64 predicate constraints.
5044 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5046 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5047 (msve-vector-bits=): New option.
5048 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5049 SVE when these are disabled.
5050 (sve): New extension.
5051 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5052 modes. Adjust their number of units based on aarch64_sve_vg.
5053 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5054 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5055 aarch64_addr_query_type.
5056 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5057 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5058 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5059 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5060 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5061 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5062 (aarch64_simd_imm_zero_p): Delete.
5063 (aarch64_check_zero_based_sve_index_immediate): Declare.
5064 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5065 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5066 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5067 (aarch64_sve_float_mul_immediate_p): Likewise.
5068 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5070 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5071 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5072 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5073 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5074 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5075 (aarch64_regmode_natural_size): Likewise.
5076 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5077 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5079 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5080 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5081 for VG and the SVE predicate registers.
5082 (V_ALIASES): Add a "z"-prefixed alias.
5083 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5084 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5085 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5086 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5087 (REG_CLASS_NAMES): Add entries for them.
5088 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5089 and the predicate registers.
5090 (aarch64_sve_vg): Declare.
5091 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5092 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5093 (REGMODE_NATURAL_SIZE): Define.
5094 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5096 * config/aarch64/aarch64.c: Include cfgrtl.h.
5097 (simd_immediate_info): Add a constructor for series vectors,
5098 and an associated step field.
5099 (aarch64_sve_vg): New variable.
5100 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5101 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5102 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5103 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5104 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5105 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5106 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5107 (aarch64_get_mask_mode): New functions.
5108 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5109 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5110 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5111 predicate modes and predicate registers. Explicitly restrict
5112 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5113 to store a vector mode if it is recognized by
5114 aarch64_classify_vector_mode.
5115 (aarch64_regmode_natural_size): New function.
5116 (aarch64_hard_regno_caller_save_mode): Return the original mode
5118 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5119 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5120 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5121 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5123 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5124 does not overlap dest if the function is frame-related. Handle
5126 (aarch64_split_add_offset): New function.
5127 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5128 them aarch64_add_offset.
5129 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5130 and update call to aarch64_sub_sp.
5131 (aarch64_add_cfa_expression): New function.
5132 (aarch64_expand_prologue): Pass extra temporary registers to the
5133 functions above. Handle the case in which we need to emit new
5134 DW_CFA_expressions for registers that were originally saved
5135 relative to the stack pointer, but now have to be expressed
5136 relative to the frame pointer.
5137 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5139 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5140 IP0 and IP1 values for SVE frames.
5141 (aarch64_expand_vec_series): New function.
5142 (aarch64_expand_sve_widened_duplicate): Likewise.
5143 (aarch64_expand_sve_const_vector): Likewise.
5144 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5145 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5146 into the register, rather than emitting a SET directly.
5147 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5148 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5149 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5150 (offset_9bit_signed_scaled_p): New functions.
5151 (aarch64_replicate_bitmask_imm): New function.
5152 (aarch64_bitmask_imm): Use it.
5153 (aarch64_cannot_force_const_mem): Reject expressions involving
5154 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5155 (aarch64_classify_index): Handle SVE indices, by requiring
5156 a plain register index with a scale that matches the element size.
5157 (aarch64_classify_address): Handle SVE addresses. Assert that
5158 the mode of the address is VOIDmode or an integer mode.
5159 Update call to aarch64_classify_symbol.
5160 (aarch64_classify_symbolic_expression): Update call to
5161 aarch64_classify_symbol.
5162 (aarch64_const_vec_all_in_range_p): New function.
5163 (aarch64_print_vector_float_operand): Likewise.
5164 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5165 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5166 and the FP immediates 1.0 and 0.5.
5167 (aarch64_print_address_internal): Handle SVE addresses.
5168 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5169 (aarch64_regno_regclass): Handle predicate registers.
5170 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5172 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5173 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5174 (aarch64_convert_sve_vector_bits): New function.
5175 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5176 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5178 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5179 Handle SVE vector and predicate modes. Accept VL-based constants
5180 that need only one temporary register, and VL offsets that require
5181 no temporary registers.
5182 (aarch64_conditional_register_usage): Mark the predicate registers
5183 as fixed if SVE isn't available.
5184 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5185 Return true for SVE vector and predicate modes.
5186 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5187 rather than an unsigned int. Handle SVE modes.
5188 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5190 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5192 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5193 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5194 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5195 (aarch64_sve_float_mul_immediate_p): New functions.
5196 (aarch64_sve_valid_immediate): New function.
5197 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5198 Explicitly reject structure modes. Check for INDEX constants.
5199 Handle PTRUE and PFALSE constants.
5200 (aarch64_check_zero_based_sve_index_immediate): New function.
5201 (aarch64_simd_imm_zero_p): Delete.
5202 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5203 vector modes. Accept constants in the range of CNT[BHWD].
5204 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5205 ask for an Advanced SIMD mode.
5206 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5207 (aarch64_simd_vector_alignment): Handle SVE predicates.
5208 (aarch64_vectorize_preferred_vector_alignment): New function.
5209 (aarch64_simd_vector_alignment_reachable): Use it instead of
5211 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5212 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5214 (MAX_VECT_LEN): Delete.
5215 (expand_vec_perm_d): Add a vec_flags field.
5216 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5217 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5218 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5220 (aarch64_evpc_rev): Rename to...
5221 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5222 (aarch64_evpc_rev_global): New function.
5223 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5224 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5226 (aarch64_evpc_sve_tbl): New function.
5227 (aarch64_expand_vec_perm_const_1): Update after rename of
5228 aarch64_evpc_rev. Handle SVE permutes too, trying
5229 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5230 than aarch64_evpc_tbl.
5231 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5232 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5233 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5234 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5235 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5236 (aarch64_expand_sve_vcond): New functions.
5237 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5238 of aarch64_vector_mode_p.
5239 (aarch64_dwarf_poly_indeterminate_value): New function.
5240 (aarch64_compute_pressure_classes): Likewise.
5241 (aarch64_can_change_mode_class): Likewise.
5242 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5243 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5244 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5245 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5246 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5247 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5248 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5249 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5251 (Dn, Dl, Dr): Accept const as well as const_vector.
5252 (Dz): Likewise. Compare against CONST0_RTX.
5253 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5254 of "vector" where appropriate.
5255 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5256 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5257 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5258 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5259 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5260 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5261 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5262 (v_int_equiv): Extend to SVE modes.
5263 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5265 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5266 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5267 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5268 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5269 (SVE_COND_FP_CMP): New int iterators.
5270 (perm_hilo): Handle the new unpack unspecs.
5271 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5273 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5274 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5275 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5276 (aarch64_equality_operator, aarch64_constant_vector_operand)
5277 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5278 (aarch64_sve_nonimmediate_operand): Likewise.
5279 (aarch64_sve_general_operand): Likewise.
5280 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5281 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5282 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5283 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5284 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5285 (aarch64_sve_float_arith_immediate): Likewise.
5286 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5287 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5288 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5289 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5290 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5291 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5292 (aarch64_sve_float_arith_operand): Likewise.
5293 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5294 (aarch64_sve_float_mul_operand): Likewise.
5295 (aarch64_sve_vec_perm_operand): Likewise.
5296 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5297 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5298 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5299 as well as const_vector.
5300 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5301 in file. Use CONST0_RTX and CONSTM1_RTX.
5302 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5303 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5304 Use aarch64_simd_imm_zero.
5305 * config/aarch64/aarch64-sve.md: New file.
5306 * config/aarch64/aarch64.md: Include it.
5307 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5308 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5309 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5310 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5311 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5312 (sve): New attribute.
5313 (enabled): Disable instructions with the sve attribute unless
5315 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5316 aarch64_expand_mov_immediate.
5317 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5318 CNT[BHSD] immediates.
5319 (movti): Split CONST_POLY_INT moves into two halves.
5320 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5321 Split additions that need a temporary here if the destination
5322 is the stack pointer.
5323 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5324 (*add<mode>3_poly_1): New instruction.
5325 (set_clobber_cc): New expander.
5327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5329 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5330 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5331 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5332 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5333 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5334 Change innermode from fixed_mode_size to machine_mode.
5335 (simplify_subreg): Update call accordingly. Handle a constant-sized
5336 subreg of a variable-length CONST_VECTOR.
5338 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5339 Alan Hayward <alan.hayward@arm.com>
5340 David Sherwood <david.sherwood@arm.com>
5342 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5343 (add_offset_to_base): New function, split out from...
5344 (create_mem_ref): ...here. When handling a scale other than 1,
5345 check first whether the address is valid without the offset.
5346 Add it into the base if so, leaving the index and scale as-is.
5348 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5351 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5352 fold_for_warn before checking if arg2 is INTEGER_CST.
5354 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5356 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5357 (store_multiple_operation): Delete.
5358 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5359 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5360 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5361 guarded by TARGET_STRING.
5362 (rs6000_output_load_multiple): Delete.
5363 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5364 OPTION_MASK_STRING / TARGET_STRING handling.
5365 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5366 (const rs6000_opt_masks) <"string">: Change mask to 0.
5367 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5368 (MASK_STRING): Delete.
5369 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5371 (load_multiple): Delete.
5378 (store_multiple): Delete.
5385 (movmemsi_8reg): Delete.
5386 (corresponding unnamed define_insn): Delete.
5387 (movmemsi_6reg): Delete.
5388 (corresponding unnamed define_insn): Delete.
5389 (movmemsi_4reg): Delete.
5390 (corresponding unnamed define_insn): Delete.
5391 (movmemsi_2reg): Delete.
5392 (corresponding unnamed define_insn): Delete.
5393 (movmemsi_1reg): Delete.
5394 (corresponding unnamed define_insn): Delete.
5395 * config/rs6000/rs6000.opt (mno-string): New.
5396 (mstring): Replace by deprecation warning stub.
5397 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5399 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5401 * regrename.c (regrename_do_replace): If replacing the same
5402 reg multiple times, try to reuse last created gen_raw_REG.
5405 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5406 main to workaround a bug in GDB.
5408 2018-01-12 Tom de Vries <tom@codesourcery.com>
5411 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5413 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5415 PR rtl-optimization/80481
5416 * ira-color.c (get_cap_member): New function.
5417 (allocnos_conflict_by_live_ranges_p): Use it.
5418 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5419 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5421 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5424 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5425 (*saddl_se_1): Ditto.
5427 (*ssubl_se_1): Ditto.
5429 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5431 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5432 rather than wi::to_widest for DR_INITs.
5433 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5434 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5435 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5437 (vect_analyze_group_access_1): Note that here.
5439 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5441 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5442 polynomial type sizes.
5444 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5446 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5447 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5448 (gimple_add_tmp_var): Likewise.
5450 2018-01-12 Martin Liska <mliska@suse.cz>
5452 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5453 (gimple_alloc_sizes): Likewise.
5454 (dump_gimple_statistics): Use PRIu64 in printf format.
5455 * gimple.h: Change uint64_t to int.
5457 2018-01-12 Martin Liska <mliska@suse.cz>
5459 * tree-core.h: Use uint64_t instead of int.
5460 * tree.c (tree_node_counts): Likewise.
5461 (tree_node_sizes): Likewise.
5462 (dump_tree_statistics): Use PRIu64 in printf format.
5464 2018-01-12 Martin Liska <mliska@suse.cz>
5466 * Makefile.in: As qsort_chk is implemented in vec.c, add
5467 vec.o to linkage of gencfn-macros.
5468 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5469 passing the info to record_node_allocation_statistics.
5470 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5472 * ggc-common.c (struct ggc_usage): Add operator== and use
5473 it in operator< and compare function.
5474 * mem-stats.h (struct mem_usage): Likewise.
5475 * vec.c (struct vec_usage): Remove operator< and compare
5476 function. Can be simply inherited.
5478 2018-01-12 Martin Jambor <mjambor@suse.cz>
5481 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5482 * tree-ssa-math-opts.c: Include domwalk.h.
5483 (convert_mult_to_fma_1): New function.
5484 (fma_transformation_info): New type.
5485 (fma_deferring_state): Likewise.
5486 (cancel_fma_deferring): New function.
5487 (result_of_phi): Likewise.
5488 (last_fma_candidate_feeds_initial_phi): Likewise.
5489 (convert_mult_to_fma): Added deferring logic, split actual
5490 transformation to convert_mult_to_fma_1.
5491 (math_opts_dom_walker): New type.
5492 (math_opts_dom_walker::after_dom_children): New method, body moved
5493 here from pass_optimize_widening_mul::execute, added deferring logic
5495 (pass_optimize_widening_mul::execute): Moved most of code to
5496 math_opts_dom_walker::after_dom_children.
5497 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5498 * config/i386/i386.c (ix86_option_override_internal): Added
5499 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5501 2018-01-12 Richard Biener <rguenther@suse.de>
5504 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5505 inline instance vars.
5507 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5510 * config/rx/rx.c (rx_is_restricted_memory_address):
5513 2018-01-12 Richard Biener <rguenther@suse.de>
5515 PR tree-optimization/80846
5516 * target.def (split_reduction): New target hook.
5517 * targhooks.c (default_split_reduction): New function.
5518 * targhooks.h (default_split_reduction): Declare.
5519 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5520 target requests first reduce vectors by combining low and high
5522 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5523 (get_vectype_for_scalar_type_and_size): Export.
5524 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5525 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5526 * doc/tm.texi: Regenerate.
5527 * config/i386/i386.c (ix86_split_reduction): Implement
5528 TARGET_VECTORIZE_SPLIT_REDUCTION.
5530 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5533 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5534 in PIC mode except for TARGET_VXWORKS_RTP.
5535 * config/sparc/sparc.c: Include cfgrtl.h.
5536 (TARGET_INIT_PIC_REG): Define.
5537 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5538 (sparc_pic_register_p): New predicate.
5539 (sparc_legitimate_address_p): Use it.
5540 (sparc_legitimize_pic_address): Likewise.
5541 (sparc_delegitimize_address): Likewise.
5542 (sparc_mode_dependent_address_p): Likewise.
5543 (gen_load_pcrel_sym): Remove 4th parameter.
5544 (load_got_register): Adjust call to above. Remove obsolete stuff.
5545 (sparc_expand_prologue): Do not call load_got_register here.
5546 (sparc_flat_expand_prologue): Likewise.
5547 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5548 (sparc_use_pseudo_pic_reg): New function.
5549 (sparc_init_pic_reg): Likewise.
5550 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5551 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5553 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5555 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5556 Add item for branch_cost.
5558 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5560 PR rtl-optimization/83565
5561 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5562 not extend the result to a larger mode for rotate operations.
5563 (num_sign_bit_copies1): Likewise.
5565 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5568 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5570 Use values-Xc.o for -pedantic.
5571 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5573 2018-01-12 Martin Liska <mliska@suse.cz>
5576 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5578 (possible_polymorphic_call_targets): Use it.
5579 (ipa_devirt): Likewise.
5581 2018-01-12 Martin Liska <mliska@suse.cz>
5583 * profile-count.h (enum profile_quality): Use 0 as invalid
5584 enum value of profile_quality.
5586 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5588 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5589 -mext-string options.
5591 2018-01-12 Richard Biener <rguenther@suse.de>
5593 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5594 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5595 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5597 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5599 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5601 * configure.ac (--with-long-double-format): Add support for the
5602 configuration option to change the default long double format on
5604 * config.gcc (powerpc*-linux*-*): Likewise.
5605 * configure: Regenerate.
5606 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5607 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5608 used without modification.
5610 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5612 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5613 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5614 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5615 MISC_BUILTIN_SPEC_BARRIER.
5616 (rs6000_init_builtins): Likewise.
5617 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5619 (speculation_barrier): New define_insn.
5620 * doc/extend.texi: Document __builtin_speculation_barrier.
5622 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5625 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5626 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5627 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5629 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5630 integral modes instead of "ss" and "sd".
5631 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5632 vectors with 32-bit and 64-bit elements.
5633 (vecdupssescalarmodesuffix): New mode attribute.
5634 (vec_dup<mode>): Use it.
5636 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5639 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5640 frame if argument is passed on stack.
5642 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5645 * ree.c (combine_reaching_defs): Optimize also
5646 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5647 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5649 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5652 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5654 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5657 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5658 after they are computed.
5660 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5662 PR tree-optimization/83695
5663 * gimple-loop-linterchange.cc
5664 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5665 reset cached scev information after interchange.
5666 (pass_linterchange::execute): Remove call to scev_reset_htab.
5668 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5670 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5671 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5672 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5673 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5674 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5675 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5676 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5677 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5678 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5679 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5680 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5681 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5682 (V_lane_reg): Likewise.
5683 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5685 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5686 (vfmal_lane_low<mode>_intrinsic,
5687 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5688 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5689 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5690 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5691 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5692 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5694 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5696 * config/arm/arm-cpus.in (fp16fml): New feature.
5697 (ALL_SIMD): Add fp16fml.
5698 (armv8.2-a): Add fp16fml as an option.
5699 (armv8.3-a): Likewise.
5700 (armv8.4-a): Add fp16fml as part of fp16.
5701 * config/arm/arm.h (TARGET_FP16FML): Define.
5702 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5704 * config/arm/arm-modes.def (V2HF): Define.
5705 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5706 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5707 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5708 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5709 vfmsl_low, vfmsl_high): New set of builtins.
5710 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5711 (vfml_op): New code attribute.
5712 (VFMLHALVES): New int iterator.
5713 (VFML, VFMLSEL): New mode attributes.
5714 (V_reg): Define mapping for V2HF.
5715 (V_hi, V_lo): New mode attributes.
5716 (VF_constraint): Likewise.
5717 (vfml_half, vfml_half_selector): New int attributes.
5718 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5720 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5721 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5723 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5724 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5725 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5726 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5728 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5729 Document new effective target and option set.
5731 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5733 * config/arm/arm-cpus.in (armv8_4): New feature.
5734 (ARMv8_4a): New fgroup.
5735 (armv8.4-a): New arch.
5736 * config/arm/arm-tables.opt: Regenerate.
5737 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5738 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5739 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5740 Add matching rules for -march=armv8.4-a and extensions.
5741 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5743 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5746 * config/rx/rx.md (BW): New mode attribute.
5747 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5749 2018-01-11 Richard Biener <rguenther@suse.de>
5751 PR tree-optimization/83435
5752 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5753 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5754 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5756 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5757 Alan Hayward <alan.hayward@arm.com>
5758 David Sherwood <david.sherwood@arm.com>
5760 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5762 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5763 (aarch64_print_address_internal): Use it to check for a zero offset.
5765 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5766 Alan Hayward <alan.hayward@arm.com>
5767 David Sherwood <david.sherwood@arm.com>
5769 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5770 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5771 Return a poly_int64 rather than a HOST_WIDE_INT.
5772 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5773 rather than a HOST_WIDE_INT.
5774 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5775 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5776 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5777 final_offset from HOST_WIDE_INT to poly_int64.
5778 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5779 to_constant when getting the number of units in an Advanced SIMD
5781 (aarch64_builtin_vectorized_function): Check for a constant number
5783 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5785 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5786 attribute instead of GET_MODE_NUNITS.
5787 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5788 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5789 GET_MODE_SIZE for fixed-size registers.
5790 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5791 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5792 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5793 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5794 (aarch64_print_operand, aarch64_print_address_internal)
5795 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5796 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5797 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5798 Handle polynomial GET_MODE_SIZE.
5799 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5800 wider than SImode without modification.
5801 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5802 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5803 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5804 passing and returning SVE modes.
5805 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5806 rather than GEN_INT.
5807 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5808 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5809 (aarch64_allocate_and_probe_stack_space): Likewise.
5810 (aarch64_layout_frame): Cope with polynomial offsets.
5811 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5812 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5814 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5815 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5816 poly_int64 rather than a HOST_WIDE_INT.
5817 (aarch64_get_separate_components, aarch64_process_components)
5818 (aarch64_expand_prologue, aarch64_expand_epilogue)
5819 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5820 (aarch64_anchor_offset): New function, split out from...
5821 (aarch64_legitimize_address): ...here.
5822 (aarch64_builtin_vectorization_cost): Handle polynomial
5823 TYPE_VECTOR_SUBPARTS.
5824 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5826 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5827 number of elements from the PARALLEL rather than the mode.
5828 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5829 rather than GET_MODE_BITSIZE.
5830 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5831 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5832 (aarch64_expand_vec_perm_const_1): Handle polynomial
5833 d->perm.length () and d->perm elements.
5834 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5835 Apply to_constant to d->perm elements.
5836 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5837 polynomial CONST_VECTOR_NUNITS.
5838 (aarch64_move_pointer): Take amount as a poly_int64 rather
5840 (aarch64_progress_pointer): Avoid temporary variable.
5841 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5842 the mode attribute instead of GET_MODE.
5844 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5845 Alan Hayward <alan.hayward@arm.com>
5846 David Sherwood <david.sherwood@arm.com>
5848 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5849 x exists before using it.
5850 (aarch64_add_constant_internal): Rename to...
5851 (aarch64_add_offset_1): ...this. Replace regnum with separate
5852 src and dest rtxes. Handle the case in which they're different,
5853 including when the offset is zero. Replace scratchreg with an rtx.
5854 Use 2 additions if there is no spare register into which we can
5855 move a 16-bit constant.
5856 (aarch64_add_constant): Delete.
5857 (aarch64_add_offset): Replace reg with separate src and dest
5858 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5859 Use aarch64_add_offset_1.
5860 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5861 an rtx rather than an int. Take the delta as a poly_int64
5862 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5863 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5864 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5865 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5866 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5868 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5869 aarch64_add_constant.
5871 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5873 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5874 Use scalar_float_mode.
5876 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5878 * config/aarch64/aarch64-simd.md
5879 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5880 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5881 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5882 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5883 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5884 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5885 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5886 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5887 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5888 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5890 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5893 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5894 targ_options->x_arm_arch_string is non NULL.
5896 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5898 * config/aarch64/aarch64.h
5899 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5901 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5904 * expmed.c (emit_store_flag_force): Swap if const op0
5905 and change VOIDmode to mode of op0.
5907 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5909 PR rtl-optimization/83761
5910 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5911 than bytes to mode_for_size.
5913 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5916 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5917 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5920 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5923 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5924 when in layout mode.
5925 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5926 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5929 2018-01-10 Michael Collison <michael.collison@arm.com>
5931 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5932 * config/aarch64/aarch64-option-extension.def: Add
5933 AARCH64_OPT_EXTENSION of 'fp16fml'.
5934 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5935 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5936 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5937 * config/aarch64/constraints.md (Ui7): New constraint.
5938 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5939 (VFMLA_SEL_W): Ditto.
5942 (VFMLA16_LOW): New int iterator.
5943 (VFMLA16_HIGH): Ditto.
5944 (UNSPEC_FMLAL): New unspec.
5945 (UNSPEC_FMLSL): Ditto.
5946 (UNSPEC_FMLAL2): Ditto.
5947 (UNSPEC_FMLSL2): Ditto.
5948 (f16mac): New code attribute.
5949 * config/aarch64/aarch64-simd-builtins.def
5950 (aarch64_fmlal_lowv2sf): Ditto.
5951 (aarch64_fmlsl_lowv2sf): Ditto.
5952 (aarch64_fmlalq_lowv4sf): Ditto.
5953 (aarch64_fmlslq_lowv4sf): Ditto.
5954 (aarch64_fmlal_highv2sf): Ditto.
5955 (aarch64_fmlsl_highv2sf): Ditto.
5956 (aarch64_fmlalq_highv4sf): Ditto.
5957 (aarch64_fmlslq_highv4sf): Ditto.
5958 (aarch64_fmlal_lane_lowv2sf): Ditto.
5959 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5960 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5961 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5962 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5963 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5964 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5965 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5966 (aarch64_fmlal_lane_highv2sf): Ditto.
5967 (aarch64_fmlsl_lane_highv2sf): Ditto.
5968 (aarch64_fmlal_laneq_highv2sf): Ditto.
5969 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5970 (aarch64_fmlalq_lane_highv4sf): Ditto.
5971 (aarch64_fmlsl_lane_highv4sf): Ditto.
5972 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5973 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5974 * config/aarch64/aarch64-simd.md:
5975 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5976 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5977 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5978 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5979 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5980 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5981 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5982 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5983 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5984 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5985 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5986 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5987 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5988 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5989 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5990 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5991 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5992 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5993 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5994 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5995 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5996 (vfmlsl_low_u32): Ditto.
5997 (vfmlalq_low_u32): Ditto.
5998 (vfmlslq_low_u32): Ditto.
5999 (vfmlal_high_u32): Ditto.
6000 (vfmlsl_high_u32): Ditto.
6001 (vfmlalq_high_u32): Ditto.
6002 (vfmlslq_high_u32): Ditto.
6003 (vfmlal_lane_low_u32): Ditto.
6004 (vfmlsl_lane_low_u32): Ditto.
6005 (vfmlal_laneq_low_u32): Ditto.
6006 (vfmlsl_laneq_low_u32): Ditto.
6007 (vfmlalq_lane_low_u32): Ditto.
6008 (vfmlslq_lane_low_u32): Ditto.
6009 (vfmlalq_laneq_low_u32): Ditto.
6010 (vfmlslq_laneq_low_u32): Ditto.
6011 (vfmlal_lane_high_u32): Ditto.
6012 (vfmlsl_lane_high_u32): Ditto.
6013 (vfmlal_laneq_high_u32): Ditto.
6014 (vfmlsl_laneq_high_u32): Ditto.
6015 (vfmlalq_lane_high_u32): Ditto.
6016 (vfmlslq_lane_high_u32): Ditto.
6017 (vfmlalq_laneq_high_u32): Ditto.
6018 (vfmlslq_laneq_high_u32): Ditto.
6019 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6020 (AARCH64_FL_FOR_ARCH8_4): New.
6021 (AARCH64_ISA_F16FML): New ISA flag.
6022 (TARGET_F16FML): New feature flag for fp16fml.
6023 (doc/invoke.texi): Document new fp16fml option.
6025 2018-01-10 Michael Collison <michael.collison@arm.com>
6027 * config/aarch64/aarch64-builtins.c:
6028 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6029 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6030 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6031 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6032 (AARCH64_ISA_SHA3): New ISA flag.
6033 (TARGET_SHA3): New feature flag for sha3.
6034 * config/aarch64/iterators.md (sha512_op): New int attribute.
6035 (CRYPTO_SHA512): New int iterator.
6036 (UNSPEC_SHA512H): New unspec.
6037 (UNSPEC_SHA512H2): Ditto.
6038 (UNSPEC_SHA512SU0): Ditto.
6039 (UNSPEC_SHA512SU1): Ditto.
6040 * config/aarch64/aarch64-simd-builtins.def
6041 (aarch64_crypto_sha512hqv2di): New builtin.
6042 (aarch64_crypto_sha512h2qv2di): Ditto.
6043 (aarch64_crypto_sha512su0qv2di): Ditto.
6044 (aarch64_crypto_sha512su1qv2di): Ditto.
6045 (aarch64_eor3qv8hi): Ditto.
6046 (aarch64_rax1qv2di): Ditto.
6047 (aarch64_xarqv2di): Ditto.
6048 (aarch64_bcaxqv8hi): Ditto.
6049 * config/aarch64/aarch64-simd.md:
6050 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6051 (aarch64_crypto_sha512su0qv2di): Ditto.
6052 (aarch64_crypto_sha512su1qv2di): Ditto.
6053 (aarch64_eor3qv8hi): Ditto.
6054 (aarch64_rax1qv2di): Ditto.
6055 (aarch64_xarqv2di): Ditto.
6056 (aarch64_bcaxqv8hi): Ditto.
6057 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6058 (vsha512h2q_u64): Ditto.
6059 (vsha512su0q_u64): Ditto.
6060 (vsha512su1q_u64): Ditto.
6061 (veor3q_u16): Ditto.
6062 (vrax1q_u64): Ditto.
6064 (vbcaxq_u16): Ditto.
6065 * config/arm/types.md (crypto_sha512): New type attribute.
6066 (crypto_sha3): Ditto.
6067 (doc/invoke.texi): Document new sha3 option.
6069 2018-01-10 Michael Collison <michael.collison@arm.com>
6071 * config/aarch64/aarch64-builtins.c:
6072 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6073 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6074 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6075 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6076 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6077 (AARCH64_ISA_SM4): New ISA flag.
6078 (TARGET_SM4): New feature flag for sm4.
6079 * config/aarch64/aarch64-simd-builtins.def
6080 (aarch64_sm3ss1qv4si): Ditto.
6081 (aarch64_sm3tt1aq4si): Ditto.
6082 (aarch64_sm3tt1bq4si): Ditto.
6083 (aarch64_sm3tt2aq4si): Ditto.
6084 (aarch64_sm3tt2bq4si): Ditto.
6085 (aarch64_sm3partw1qv4si): Ditto.
6086 (aarch64_sm3partw2qv4si): Ditto.
6087 (aarch64_sm4eqv4si): Ditto.
6088 (aarch64_sm4ekeyqv4si): Ditto.
6089 * config/aarch64/aarch64-simd.md:
6090 (aarch64_sm3ss1qv4si): Ditto.
6091 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6092 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6093 (aarch64_sm4eqv4si): Ditto.
6094 (aarch64_sm4ekeyqv4si): Ditto.
6095 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6096 (sm3part_op): Ditto.
6097 (CRYPTO_SM3TT): Ditto.
6098 (CRYPTO_SM3PART): Ditto.
6099 (UNSPEC_SM3SS1): New unspec.
6100 (UNSPEC_SM3TT1A): Ditto.
6101 (UNSPEC_SM3TT1B): Ditto.
6102 (UNSPEC_SM3TT2A): Ditto.
6103 (UNSPEC_SM3TT2B): Ditto.
6104 (UNSPEC_SM3PARTW1): Ditto.
6105 (UNSPEC_SM3PARTW2): Ditto.
6106 (UNSPEC_SM4E): Ditto.
6107 (UNSPEC_SM4EKEY): Ditto.
6108 * config/aarch64/constraints.md (Ui2): New constraint.
6109 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6110 * config/arm/types.md (crypto_sm3): New type attribute.
6111 (crypto_sm4): Ditto.
6112 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6113 (vsm3tt1aq_u32): Ditto.
6114 (vsm3tt1bq_u32): Ditto.
6115 (vsm3tt2aq_u32): Ditto.
6116 (vsm3tt2bq_u32): Ditto.
6117 (vsm3partw1q_u32): Ditto.
6118 (vsm3partw2q_u32): Ditto.
6119 (vsm4eq_u32): Ditto.
6120 (vsm4ekeyq_u32): Ditto.
6121 (doc/invoke.texi): Document new sm4 option.
6123 2018-01-10 Michael Collison <michael.collison@arm.com>
6125 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6126 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6127 (AARCH64_FL_FOR_ARCH8_4): New.
6128 (AARCH64_FL_V8_4): New flag.
6129 (doc/invoke.texi): Document new armv8.4-a option.
6131 2018-01-10 Michael Collison <michael.collison@arm.com>
6133 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6134 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6135 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6136 * config/aarch64/aarch64-option-extension.def: Add
6137 AARCH64_OPT_EXTENSION of 'sha2'.
6138 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6139 (crypto): Disable sha2 and aes if crypto disabled.
6140 (crypto): Enable aes and sha2 if enabled.
6141 (simd): Disable sha2 and aes if simd disabled.
6142 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6144 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6145 (TARGET_SHA2): New feature flag for sha2.
6146 (TARGET_AES): New feature flag for aes.
6147 * config/aarch64/aarch64-simd.md:
6148 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6149 conditional on TARGET_AES.
6150 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6151 (aarch64_crypto_sha1hsi): Make pattern conditional
6153 (aarch64_crypto_sha1hv4si): Ditto.
6154 (aarch64_be_crypto_sha1hv4si): Ditto.
6155 (aarch64_crypto_sha1su1v4si): Ditto.
6156 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6157 (aarch64_crypto_sha1su0v4si): Ditto.
6158 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6159 (aarch64_crypto_sha256su0v4si): Ditto.
6160 (aarch64_crypto_sha256su1v4si): Ditto.
6161 (doc/invoke.texi): Document new aes and sha2 options.
6163 2018-01-10 Martin Sebor <msebor@redhat.com>
6165 PR tree-optimization/83781
6166 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6169 2018-01-11 Martin Sebor <msebor@gmail.com>
6170 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6172 PR tree-optimization/83501
6173 PR tree-optimization/81703
6175 * tree-ssa-strlen.c (get_string_cst): Rename...
6176 (get_string_len): ...to this. Handle global constants.
6177 (handle_char_store): Adjust.
6179 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6180 Jim Wilson <jimw@sifive.com>
6182 * config/riscv/riscv-protos.h (riscv_output_return): New.
6183 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6184 (riscv_attribute_table, riscv_output_return),
6185 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6186 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6187 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6189 (riscv_expand_prologue): Add early return for naked function.
6190 (riscv_expand_epilogue): Likewise.
6191 (riscv_function_ok_for_sibcall): Return false for naked function.
6192 (riscv_set_current_function): New.
6193 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6194 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6195 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6196 * doc/extend.texi (RISC-V Function Attributes): New.
6198 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6200 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6201 check for 128-bit long double before checking TCmode.
6202 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6203 128-bit long doubles before checking TFmode or TCmode.
6204 (FLOAT128_IBM_P): Likewise.
6206 2018-01-10 Martin Sebor <msebor@redhat.com>
6208 PR tree-optimization/83671
6209 * builtins.c (c_strlen): Unconditionally return zero for the empty
6211 Use -Warray-bounds for warnings.
6212 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6213 for non-constant array indices with COMPONENT_REF, arrays of
6214 arrays, and pointers to arrays.
6215 (gimple_fold_builtin_strlen): Determine and set length range for
6216 non-constant character arrays.
6218 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6221 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6224 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6226 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6228 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6231 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6232 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6233 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6234 indexed_or_indirect_operand predicate.
6235 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6236 (*vsx_le_perm_load_v8hi): Likewise.
6237 (*vsx_le_perm_load_v16qi): Likewise.
6238 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6239 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6240 (*vsx_le_perm_store_v8hi): Likewise.
6241 (*vsx_le_perm_store_v16qi): Likewise.
6242 (eight unnamed splitters): Likewise.
6244 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6246 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6247 * config/rs6000/emmintrin.h: Likewise.
6248 * config/rs6000/mmintrin.h: Likewise.
6249 * config/rs6000/xmmintrin.h: Likewise.
6251 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6254 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6256 * tree.c (tree_nop_conversion): Return true for location wrapper
6258 (maybe_wrap_with_location): New function.
6259 (selftest::check_strip_nops): New function.
6260 (selftest::test_location_wrappers): New function.
6261 (selftest::tree_c_tests): Call it.
6262 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6263 (maybe_wrap_with_location): New decl.
6264 (EXPR_LOCATION_WRAPPER_P): New macro.
6265 (location_wrapper_p): New inline function.
6266 (tree_strip_any_location_wrapper): New inline function.
6268 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6271 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6272 stack_realign_offset for the largest alignment of stack slot
6274 (ix86_find_max_used_stack_alignment): New function.
6275 (ix86_finalize_stack_frame_flags): Use it. Set
6276 max_used_stack_alignment if we don't realign stack.
6277 * config/i386/i386.h (machine_function): Add
6278 max_used_stack_alignment.
6280 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6282 * config/arm/arm.opt (-mbranch-cost): New option.
6283 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6286 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6289 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6290 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6292 2018-01-10 Richard Biener <rguenther@suse.de>
6295 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6296 early out so it also covers the case where we have a non-NULL
6299 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6301 PR tree-optimization/83753
6302 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6303 for non-strided grouped accesses if the number of elements is 1.
6305 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6308 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6309 * i386.h (TARGET_USE_GATHER): Define.
6310 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6312 2018-01-10 Martin Liska <mliska@suse.cz>
6315 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6316 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6318 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6319 CLEANUP_NO_PARTITIONING is not set.
6321 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6323 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6324 for vectors, as a partial revert of r254296.
6325 * rtl.h (const_vec_p): Delete.
6326 (const_vec_duplicate_p): Don't test for vector CONSTs.
6327 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6328 * expmed.c (make_tree): Likewise.
6331 * common.md (E, F): Use CONSTANT_P instead of checking for
6333 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6334 checking for CONST_VECTOR.
6336 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6339 * predict.c (force_edge_cold): Handle in more sane way edges
6342 2018-01-09 Carl Love <cel@us.ibm.com>
6344 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6346 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6347 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6348 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6349 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6350 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6351 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6352 * config/rs6000/rs6000-protos.h: Add extern defition for
6353 rs6000_generate_float2_double_code.
6354 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6356 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6357 (float2_v2df): Add define_expand.
6359 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6362 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6363 op_mode in the force_to_mode call.
6365 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6367 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6368 instead of checking each element individually.
6369 (aarch64_evpc_uzp): Likewise.
6370 (aarch64_evpc_zip): Likewise.
6371 (aarch64_evpc_ext): Likewise.
6372 (aarch64_evpc_rev): Likewise.
6373 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6374 instead of checking each element individually. Return true without
6376 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6377 whether all selected elements come from the same input, instead of
6378 checking each element individually. Remove calls to gen_rtx_REG,
6379 start_sequence and end_sequence and instead assert that no rtl is
6382 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6384 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6385 order of HIGH and CONST checks.
6387 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6389 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6390 if the destination isn't an SSA_NAME.
6392 2018-01-09 Richard Biener <rguenther@suse.de>
6394 PR tree-optimization/83668
6395 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6397 (canonicalize_loop_form): ... here, renamed from ...
6398 (canonicalize_loop_closed_ssa_form): ... this and amended to
6399 swap successor edges for loop exit blocks to make us use
6400 the RPO order we need for initial schedule generation.
6402 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6404 PR tree-optimization/64811
6405 * match.pd: When optimizing comparisons with Inf, avoid
6406 introducing or losing exceptions from comparisons with NaN.
6408 2018-01-09 Martin Liska <mliska@suse.cz>
6411 * asan.c (shadow_mem_size): Add gcc_assert.
6413 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6415 Don't save registers in main().
6418 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6419 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6420 * config/avr/avr.c (avr_set_current_function): Don't error if
6421 naked, OS_task or OS_main are specified at the same time.
6422 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6424 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6426 * common/config/avr/avr-common.c (avr_option_optimization_table):
6427 Switch on -mmain-is-OS_task for optimizing compilations.
6429 2018-01-09 Richard Biener <rguenther@suse.de>
6431 PR tree-optimization/83572
6432 * graphite.c: Include cfganal.h.
6433 (graphite_transform_loops): Connect infinite loops to exit
6434 and remove fake edges at the end.
6436 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6438 * ipa-inline.c (edge_badness): Revert accidental checkin.
6440 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6443 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6444 symbols; not inline clones.
6446 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6449 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6450 hard registers. Formatting fixes.
6452 PR preprocessor/83722
6453 * gcc.c (try_generate_repro): Pass
6454 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6455 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6458 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6459 Kito Cheng <kito.cheng@gmail.com>
6461 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6462 (riscv_leaf_function_p): Delete.
6463 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6465 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6467 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6469 (do_ifelse): New function.
6470 (do_isel): New function.
6471 (do_sub3): New function.
6472 (do_add3): New function.
6473 (do_load_mask_compare): New function.
6474 (do_overlap_load_compare): New function.
6475 (expand_compare_loop): New function.
6476 (expand_block_compare): Call expand_compare_loop() when appropriate.
6477 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6479 (-mblock-compare-inline-loop-limit): New option.
6481 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6484 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6485 Reverse order of second and third operands in first alternative.
6486 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6487 of first and second elements in UNSPEC_VPERMR vector.
6488 (altivec_expand_vec_perm_le): Likewise.
6490 2018-01-08 Jeff Law <law@redhat.com>
6492 PR rtl-optimizatin/81308
6493 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6494 (process_switch): If group_case_labels makes a change, then set
6496 (pass_convert_switch::execute): If a switch is converted, then
6497 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6499 PR rtl-optimization/81308
6500 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6503 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6505 PR target/83663 - Revert r255946
6506 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6507 generation for cases where splatting a value is not useful.
6508 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6509 across a vec_duplicate and a paradoxical subreg forming a vector
6510 mode to a vec_concat.
6512 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6514 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6515 -march=armv8.3-a variants.
6516 * config/arm/t-multilib: Likewise.
6517 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6519 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6521 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6523 (cceq_ior_compare_complement): Give it a name so I can use it, and
6524 change boolean_or_operator predicate to boolean_operator so it can
6525 be used to generate a crand.
6526 (eqne): New code iterator.
6527 (bd/bd_neg): New code_attrs.
6528 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6529 a single define_insn.
6530 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6531 decrement (bdnzt/bdnzf/bdzt/bdzf).
6532 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6533 with the new names of the branch decrement patterns, and added the
6534 names of the branch decrement conditional patterns.
6536 2018-01-08 Richard Biener <rguenther@suse.de>
6538 PR tree-optimization/83563
6539 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6542 2018-01-08 Richard Biener <rguenther@suse.de>
6545 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6547 2018-01-08 Richard Biener <rguenther@suse.de>
6549 PR tree-optimization/83685
6550 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6551 references to abnormals.
6553 2018-01-08 Richard Biener <rguenther@suse.de>
6556 * dwarf2out.c (output_indirect_strings): Handle empty
6557 skeleton_debug_str_hash.
6558 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6560 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6562 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6563 (emit_store_direct): Likewise.
6564 (arc_trampoline_adjust_address): Likewise.
6565 (arc_asm_trampoline_template): New function.
6566 (arc_initialize_trampoline): Use asm_trampoline_template.
6567 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6568 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6569 * config/arc/arc.md (flush_icache): Delete pattern.
6571 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6573 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6574 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6577 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6580 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6581 by not USED_FOR_TARGET.
6582 (make_pass_resolve_sw_modes): Likewise.
6584 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6586 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6589 2018-01-08 Richard Biener <rguenther@suse.de>
6592 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6594 2018-01-08 Richard Biener <rguenther@suse.de>
6597 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6599 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6602 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6603 basic blocks with a small number of successors.
6604 (convert_control_dep_chain_into_preds): Improve handling of
6606 (dump_predicates): Split apart into...
6607 (dump_pred_chain): ...here...
6608 (dump_pred_info): ...and here.
6609 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6610 (can_chain_union_be_invalidated_p): Improve check for invalidation
6612 (uninit_uses_cannot_happen): Avoid unnecessary if
6613 convert_control_dep_chain_into_preds yielded nothing.
6615 2018-01-06 Martin Sebor <msebor@redhat.com>
6617 PR tree-optimization/83640
6618 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6619 subtracting negative offset from size.
6620 (builtin_access::overlap): Adjust offset bounds of the access to fall
6621 within the size of the object if possible.
6623 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6625 PR rtl-optimization/83699
6626 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6627 extract_bit_field_as_subreg to cases in which the extracted
6628 value is also a vector.
6630 * lra-constraints.c (process_alt_operands): Test for the equivalence
6631 substitutions when detecting a possible reload cycle.
6633 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6636 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6637 by default if flag_selective_schedling{,2}. Formatting fixes.
6639 PR rtl-optimization/83682
6640 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6641 if it has non-VECTOR_MODE element mode.
6642 (vec_duplicate_p): Likewise.
6645 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6646 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6648 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6651 * config/i386/i386-builtin.def
6652 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6653 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6654 Require also OPTION_MASK_ISA_AVX512F in addition to
6655 OPTION_MASK_ISA_GFNI.
6656 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6657 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6658 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6659 to OPTION_MASK_ISA_GFNI.
6660 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6661 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6662 OPTION_MASK_ISA_AVX512BW.
6663 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6664 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6665 addition to OPTION_MASK_ISA_GFNI.
6666 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6667 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6668 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6669 to OPTION_MASK_ISA_GFNI.
6670 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6671 a requirement for all ISAs rather than any of them with a few
6673 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6675 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6676 bitmasks to be enabled with 3 exceptions, instead of requiring any
6677 enabled ISA with lots of exceptions.
6678 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6679 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6680 Change avx512bw in isa attribute to avx512f.
6681 * config/i386/sgxintrin.h: Add license boilerplate.
6682 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6683 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6684 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6685 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6687 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6688 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6689 temporarily sse2 rather than sse if not enabled already.
6692 * config/i386/sse.md (VI248_VLBW): Rename to ...
6693 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6694 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6695 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6696 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6697 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6698 mode iterator instead of VI248_VLBW.
6700 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6702 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6703 (record_modified): Skip clobbers; add debug output.
6704 (param_change_prob): Use sreal frequencies.
6706 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6708 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6709 punt for user-aligned variables.
6711 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6713 * tree-chrec.c (chrec_contains_symbols): Return true for
6716 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6719 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6720 of (x|y) == x for BICS pattern.
6722 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6724 PR tree-optimization/83605
6725 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6726 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6729 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6731 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6732 * config/epiphany/rtems.h: New file.
6734 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6735 Uros Bizjak <ubizjak@gmail.com>
6738 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6739 QIreg_operand instead of register_operand predicate.
6740 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6741 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6742 comments instead of -fmitigate[-_]rop.
6744 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6747 * cgraphunit.c (symbol_table::compile): Switch to text_section
6748 before calling assembly_start debug hook.
6749 * run-rtl-passes.c (run_rtl_passes): Likewise.
6752 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6754 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6755 range_int_cst_p rather than !symbolic_range_p before calling
6756 extract_range_from_multiplicative_op_1.
6758 2018-01-04 Jeff Law <law@redhat.com>
6760 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6761 redundant test in assertion.
6763 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6765 * doc/rtl.texi: Document machine_mode wrapper classes.
6767 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6769 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6772 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6774 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6775 the VEC_PERM_EXPR fold to fail.
6777 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6780 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6781 to switched_sections.
6783 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6786 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6789 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6792 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6793 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6795 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6798 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6799 is BLKmode and bitpos not zero or mode change is needed.
6801 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6804 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6807 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6810 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6811 instead of MULT rtx. Update all corresponding splitters.
6813 (*ssub<modesuffix>): Ditto.
6815 (*cmp_sadd_di): Update split patterns.
6816 (*cmp_sadd_si): Ditto.
6817 (*cmp_sadd_sidi): Ditto.
6818 (*cmp_ssub_di): Ditto.
6819 (*cmp_ssub_si): Ditto.
6820 (*cmp_ssub_sidi): Ditto.
6821 * config/alpha/predicates.md (const23_operand): New predicate.
6822 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6823 Look for ASHIFT, not MULT inner operand.
6824 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6826 2018-01-04 Martin Liska <mliska@suse.cz>
6828 PR gcov-profile/83669
6829 * gcov.c (output_intermediate_file): Add version to intermediate
6831 * doc/gcov.texi: Document new field 'version' in intermediate
6832 file format. Fix location of '-k' option of gcov command.
6834 2018-01-04 Martin Liska <mliska@suse.cz>
6837 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6839 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6841 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6843 2018-01-03 Martin Sebor <msebor@redhat.com>
6845 PR tree-optimization/83655
6846 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6847 checking calls with invalid arguments.
6849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6851 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6852 (vectorizable_mask_load_store): Delete.
6853 (vectorizable_call): Return false for masked loads and stores.
6854 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6855 instead of gimple_assign_rhs1.
6856 (vectorizable_load): Handle IFN_MASK_LOAD.
6857 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6861 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6863 (vectorizable_mask_load_store): ...here.
6864 (vectorizable_load): ...and here.
6866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6868 * tree-vect-stmts.c (vect_build_all_ones_mask)
6869 (vect_build_zero_merge_argument): New functions, split out from...
6870 (vectorizable_load): ...here.
6872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6874 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6876 (vectorizable_mask_load_store): ...here.
6877 (vectorizable_store): ...and here.
6879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6881 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6883 (vectorizable_mask_load_store): ...here.
6885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6887 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6888 (vect_model_store_cost): Take a vec_load_store_type instead of a
6890 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6891 (vect_model_store_cost): Take a vec_load_store_type instead of a
6893 (vectorizable_mask_load_store): Update accordingly.
6894 (vectorizable_store): Likewise.
6895 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6899 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6900 IFN_MASK_LOAD calls here rather than...
6901 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6904 Alan Hayward <alan.hayward@arm.com>
6905 David Sherwood <david.sherwood@arm.com>
6907 * expmed.c (extract_bit_field_1): For vector extracts,
6908 fall back to extract_bit_field_as_subreg if vec_extract
6911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6912 Alan Hayward <alan.hayward@arm.com>
6913 David Sherwood <david.sherwood@arm.com>
6915 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6916 they are variable or constant sized.
6917 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6918 slots for constant-sized data.
6920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6921 Alan Hayward <alan.hayward@arm.com>
6922 David Sherwood <david.sherwood@arm.com>
6924 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6925 handling COND_EXPRs with boolean comparisons, try to find a better
6926 basis for the mask type than the boolean itself.
6928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6930 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6931 is calculated and how it can be overridden.
6932 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6933 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6935 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6939 Alan Hayward <alan.hayward@arm.com>
6940 David Sherwood <david.sherwood@arm.com>
6942 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6943 Remove the mode argument.
6944 (aarch64_simd_valid_immediate): Remove the mode and inverse
6946 * config/aarch64/iterators.md (bitsize): New iterator.
6947 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6948 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6949 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6950 aarch64_simd_valid_immediate.
6951 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6952 (aarch64_reg_or_bic_imm): Likewise.
6953 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6954 with an insn_type enum and msl with a modifier_type enum.
6955 Replace element_width with a scalar_mode. Change the shift
6956 to unsigned int. Add constructors for scalar_float_mode and
6957 scalar_int_mode elements.
6958 (aarch64_vect_float_const_representable_p): Delete.
6959 (aarch64_can_const_movi_rtx_p)
6960 (aarch64_simd_scalar_immediate_valid_for_move)
6961 (aarch64_simd_make_constant): Update call to
6962 aarch64_simd_valid_immediate.
6963 (aarch64_advsimd_valid_immediate_hs): New function.
6964 (aarch64_advsimd_valid_immediate): Likewise.
6965 (aarch64_simd_valid_immediate): Remove mode and inverse
6966 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6967 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6968 and aarch64_float_const_representable_p on the result.
6969 (aarch64_output_simd_mov_immediate): Remove mode argument.
6970 Update call to aarch64_simd_valid_immediate and use of
6971 simd_immediate_info.
6972 (aarch64_output_scalar_simd_mov_immediate): Update call
6975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6976 Alan Hayward <alan.hayward@arm.com>
6977 David Sherwood <david.sherwood@arm.com>
6979 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6980 (mode_nunits): Likewise CONST_MODE_NUNITS.
6981 * machmode.def (ADJUST_NUNITS): Document.
6982 * genmodes.c (mode_data::need_nunits_adj): New field.
6983 (blank_mode): Update accordingly.
6984 (adj_nunits): New variable.
6985 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6987 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6988 listed in adj_nunits.
6989 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6990 listed in adj_nunits. Don't emit case statements for such modes.
6991 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6992 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6993 nothing if adj_nunits is nonnull.
6994 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6995 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6996 (emit_mode_fbit): Update use of print_maybe_const_decl.
6997 (emit_move_size): Likewise. Treat the array as non-const
6999 (emit_mode_adjustments): Handle adj_nunits.
7001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7003 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7004 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7005 (VECTOR_MODES): Use it.
7006 (make_vector_modes): Take the prefix as an argument.
7008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7009 Alan Hayward <alan.hayward@arm.com>
7010 David Sherwood <david.sherwood@arm.com>
7012 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7013 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7014 for MODE_VECTOR_BOOL.
7015 * machmode.def (VECTOR_BOOL_MODE): Document.
7016 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7017 (make_vector_bool_mode): New function.
7018 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7020 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7021 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7023 * stor-layout.c (int_mode_for_mode): Likewise.
7024 * tree.c (build_vector_type_for_mode): Likewise.
7025 * varasm.c (output_constant_pool_2): Likewise.
7026 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7027 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7028 for MODE_VECTOR_BOOL.
7029 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7030 of mode class checks.
7031 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7032 instead of a list of mode class checks.
7033 (expand_vector_scalar_condition): Likewise.
7034 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7037 Alan Hayward <alan.hayward@arm.com>
7038 David Sherwood <david.sherwood@arm.com>
7040 * machmode.h (mode_size): Change from unsigned short to
7042 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7043 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7044 or if measurement_type is not polynomial.
7045 (fixed_size_mode::includes_p): Check for constant-sized modes.
7046 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7047 return a poly_uint16 rather than an unsigned short.
7048 (emit_mode_size): Change the type of mode_size from unsigned short
7049 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7050 (emit_mode_adjustments): Cope with polynomial vector sizes.
7051 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7053 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7055 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7056 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7057 * caller-save.c (setup_save_areas): Likewise.
7058 (replace_reg_with_saved_mem): Likewise.
7059 * calls.c (emit_library_call_value_1): Likewise.
7060 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7061 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7062 (gen_lowpart_for_combine): Likewise.
7063 * convert.c (convert_to_integer_1): Likewise.
7064 * cse.c (equiv_constant, cse_insn): Likewise.
7065 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7066 (cselib_subst_to_values): Likewise.
7067 * dce.c (word_dce_process_block): Likewise.
7068 * df-problems.c (df_word_lr_mark_ref): Likewise.
7069 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7070 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7071 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7072 (rtl_for_decl_location): Likewise.
7073 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7074 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7075 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7076 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7077 (expand_expr_real_1): Likewise.
7078 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7079 (pad_below): Likewise.
7080 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7081 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7082 * ira.c (get_subreg_tracking_sizes): Likewise.
7083 * ira-build.c (ira_create_allocno_objects): Likewise.
7084 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7085 (ira_sort_regnos_for_alter_reg): Likewise.
7086 * ira-costs.c (record_operand_costs): Likewise.
7087 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7088 (resolve_simple_move): Likewise.
7089 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7090 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7091 (lra_constraints): Likewise.
7092 (CONST_POOL_OK_P): Reject variable-sized modes.
7093 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7094 (add_pseudo_to_slot, lra_spill): Likewise.
7095 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7096 * optabs-query.c (get_best_extraction_insn): Likewise.
7097 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7098 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7099 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7100 * recog.c (offsettable_address_addr_space_p): Likewise.
7101 * regcprop.c (maybe_mode_change): Likewise.
7102 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7103 * regrename.c (build_def_use): Likewise.
7104 * regstat.c (dump_reg_info): Likewise.
7105 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7106 (find_reloads, find_reloads_subreg_address): Likewise.
7107 * reload1.c (eliminate_regs_1): Likewise.
7108 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7109 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7110 (simplify_binary_operation_1, simplify_subreg): Likewise.
7111 * targhooks.c (default_function_arg_padding): Likewise.
7112 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7113 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7114 (verify_gimple_assign_ternary): Likewise.
7115 * tree-inline.c (estimate_move_cost): Likewise.
7116 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7117 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7118 (get_address_cost_ainc): Likewise.
7119 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7120 (vect_supportable_dr_alignment): Likewise.
7121 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7122 (vectorizable_reduction): Likewise.
7123 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7124 (vectorizable_operation, vectorizable_load): Likewise.
7125 * tree.c (build_same_sized_truth_vector_type): Likewise.
7126 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7127 * var-tracking.c (emit_note_insn_var_location): Likewise.
7128 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7129 (ADDR_VEC_ALIGN): Likewise.
7131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7132 Alan Hayward <alan.hayward@arm.com>
7133 David Sherwood <david.sherwood@arm.com>
7135 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7137 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7138 or if measurement_type is polynomial.
7139 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7140 * combine.c (make_extraction): Likewise.
7141 * dse.c (find_shift_sequence): Likewise.
7142 * dwarf2out.c (mem_loc_descriptor): Likewise.
7143 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7144 (extract_bit_field, extract_low_bits): Likewise.
7145 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7146 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7147 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7148 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7149 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7150 * reload.c (find_reloads): Likewise.
7151 * reload1.c (alter_reg): Likewise.
7152 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7153 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7154 * tree-if-conv.c (predicate_mem_writes): Likewise.
7155 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7156 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7157 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7158 * valtrack.c (dead_debug_insert_temp): Likewise.
7159 * varasm.c (mergeable_constant_section): Likewise.
7160 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7163 Alan Hayward <alan.hayward@arm.com>
7164 David Sherwood <david.sherwood@arm.com>
7166 * expr.c (expand_assignment): Cope with polynomial mode sizes
7167 when assigning to a CONCAT.
7169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7170 Alan Hayward <alan.hayward@arm.com>
7171 David Sherwood <david.sherwood@arm.com>
7173 * machmode.h (mode_precision): Change from unsigned short to
7175 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7177 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7178 or if measurement_type is not polynomial.
7179 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7180 in which the mode is already known to be a scalar_int_mode.
7181 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7182 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7184 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7185 for GET_MODE_PRECISION.
7186 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7187 for GET_MODE_PRECISION.
7188 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7190 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7191 (expand_field_assignment, make_extraction): Likewise.
7192 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7193 (get_last_value): Likewise.
7194 * convert.c (convert_to_integer_1): Likewise.
7195 * cse.c (cse_insn): Likewise.
7196 * expr.c (expand_expr_real_1): Likewise.
7197 * lra-constraints.c (simplify_operand_subreg): Likewise.
7198 * optabs-query.c (can_atomic_load_p): Likewise.
7199 * optabs.c (expand_atomic_load): Likewise.
7200 (expand_atomic_store): Likewise.
7201 * ree.c (combine_reaching_defs): Likewise.
7202 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7203 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7204 * tree.h (type_has_mode_precision_p): Likewise.
7205 * ubsan.c (instrument_si_overflow): Likewise.
7207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7208 Alan Hayward <alan.hayward@arm.com>
7209 David Sherwood <david.sherwood@arm.com>
7211 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7212 polynomial numbers of units.
7213 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7214 (valid_vector_subparts_p): New function.
7215 (build_vector_type): Remove temporary shim and take the number
7216 of units as a poly_uint64 rather than an int.
7217 (build_opaque_vector_type): Take the number of units as a
7218 poly_uint64 rather than an int.
7219 * tree.c (build_vector_from_ctor): Handle polynomial
7220 TYPE_VECTOR_SUBPARTS.
7221 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7222 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7223 (build_vector_from_val): If the number of units is variable,
7224 use build_vec_duplicate_cst for constant operands and
7225 VEC_DUPLICATE_EXPR otherwise.
7226 (make_vector_type): Remove temporary is_constant ().
7227 (build_vector_type, build_opaque_vector_type): Take the number of
7228 units as a poly_uint64 rather than an int.
7229 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7231 * cfgexpand.c (expand_debug_expr): Likewise.
7232 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7233 (store_constructor, expand_expr_real_1): Likewise.
7234 (const_scalar_mask_from_tree): Likewise.
7235 * fold-const-call.c (fold_const_reduction): Likewise.
7236 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7237 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7238 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7239 (fold_relational_const): Likewise.
7240 (native_interpret_vector): Likewise. Change the size from an
7241 int to an unsigned int.
7242 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7243 TYPE_VECTOR_SUBPARTS.
7244 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7245 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7246 duplicating a non-constant operand into a variable-length vector.
7247 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7248 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7249 * ipa-icf.c (sem_variable::equals): Likewise.
7250 * match.pd: Likewise.
7251 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7252 * print-tree.c (print_node): Likewise.
7253 * stor-layout.c (layout_type): Likewise.
7254 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7255 * tree-cfg.c (verify_gimple_comparison): Likewise.
7256 (verify_gimple_assign_binary): Likewise.
7257 (verify_gimple_assign_ternary): Likewise.
7258 (verify_gimple_assign_single): Likewise.
7259 * tree-pretty-print.c (dump_generic_node): Likewise.
7260 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7261 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7262 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7263 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7264 (vect_shift_permute_load_chain): Likewise.
7265 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7266 (expand_vector_condition, optimize_vector_constructor): Likewise.
7267 (lower_vec_perm, get_compute_type): Likewise.
7268 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7269 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7270 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7271 (vect_recog_mask_conversion_pattern): Likewise.
7272 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7273 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7274 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7275 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7276 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7277 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7278 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7279 (supportable_widening_operation): Likewise.
7280 (supportable_narrowing_operation): Likewise.
7281 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7283 * varasm.c (output_constant): Likewise.
7285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7286 Alan Hayward <alan.hayward@arm.com>
7287 David Sherwood <david.sherwood@arm.com>
7289 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7290 so that both the length == 3 and length != 3 cases set up their
7291 own permute vectors. Add comments explaining why we know the
7292 number of elements is constant.
7293 (vect_permute_load_chain): Likewise.
7295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7296 Alan Hayward <alan.hayward@arm.com>
7297 David Sherwood <david.sherwood@arm.com>
7299 * machmode.h (mode_nunits): Change from unsigned char to
7301 (ONLY_FIXED_SIZE_MODES): New macro.
7302 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7303 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7304 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7306 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7307 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7308 or if measurement_type is not polynomial.
7309 * genmodes.c (ZERO_COEFFS): New macro.
7310 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7312 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7313 Use ZERO_COEFFS when emitting initializers.
7314 * data-streamer.h (bp_pack_poly_value): New function.
7315 (bp_unpack_poly_value): Likewise.
7316 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7317 for GET_MODE_NUNITS.
7318 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7319 for GET_MODE_NUNITS.
7320 * tree.c (make_vector_type): Remove temporary shim and make
7321 the real function take the number of units as a poly_uint64
7323 (build_vector_type_for_mode): Handle polynomial nunits.
7324 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7325 * emit-rtl.c (const_vec_series_p_1): Likewise.
7326 (gen_rtx_CONST_VECTOR): Likewise.
7327 * fold-const.c (test_vec_duplicate_folding): Likewise.
7328 * genrecog.c (validate_pattern): Likewise.
7329 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7330 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7331 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7332 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7333 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7334 * rtlanal.c (subreg_get_info): Likewise.
7335 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7336 (vect_grouped_load_supported): Likewise.
7337 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7338 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7339 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7340 (simplify_const_unary_operation, simplify_binary_operation_1)
7341 (simplify_const_binary_operation, simplify_ternary_operation)
7342 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7343 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7344 instead of CONST_VECTOR_NUNITS.
7345 * varasm.c (output_constant_pool_2): Likewise.
7346 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7347 explicit-encoded elements in the XVEC for variable-length vectors.
7349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7351 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7354 Alan Hayward <alan.hayward@arm.com>
7355 David Sherwood <david.sherwood@arm.com>
7357 * coretypes.h (fixed_size_mode): Declare.
7358 (fixed_size_mode_pod): New typedef.
7359 * builtins.h (target_builtins::x_apply_args_mode)
7360 (target_builtins::x_apply_result_mode): Change type to
7361 fixed_size_mode_pod.
7362 * builtins.c (apply_args_size, apply_result_size, result_vector)
7363 (expand_builtin_apply_args_1, expand_builtin_apply)
7364 (expand_builtin_return): Update accordingly.
7366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7368 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7369 * cselib.c (cselib_hash_rtx): Likewise.
7370 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7371 CONST_VECTOR encoding.
7373 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7374 Jeff Law <law@redhat.com>
7377 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7378 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7379 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7380 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7383 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7384 explicitly probe *sp in a noreturn function if there were any callee
7385 register saves or frame pointer is needed.
7387 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7390 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7391 BLKmode for ternary, binary or unary expressions.
7394 * var-tracking.c (delete_vta_debug_insn): New inline function.
7395 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7396 insns from get_insns () to NULL instead of each bb separately.
7397 Use delete_vta_debug_insn. No longer static.
7398 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7399 delete_vta_debug_insns callers.
7400 * rtl.h (delete_vta_debug_insns): Declare.
7401 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7402 instead of variable_tracking_main.
7404 2018-01-03 Martin Sebor <msebor@redhat.com>
7406 PR tree-optimization/83603
7407 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7408 arguments past the endof the argument list in functions declared
7409 without a prototype.
7410 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7411 Avoid checking when arguments are null.
7413 2018-01-03 Martin Sebor <msebor@redhat.com>
7416 * doc/extend.texi (attribute const): Fix a typo.
7417 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7418 issuing -Wsuggest-attribute for void functions.
7420 2018-01-03 Martin Sebor <msebor@redhat.com>
7422 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7423 offset_int::from instead of wide_int::to_shwi.
7424 (maybe_diag_overlap): Remove assertion.
7425 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7426 * gimple-ssa-sprintf.c (format_directive): Same.
7427 (parse_directive): Same.
7428 (sprintf_dom_walker::compute_format_length): Same.
7429 (try_substitute_return_value): Same.
7431 2018-01-03 Jeff Law <law@redhat.com>
7434 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7435 non-constant residual for zero at runtime and avoid probing in
7436 that case. Reorganize code for trailing problem to mirror handling
7439 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7441 PR tree-optimization/83501
7442 * tree-ssa-strlen.c (get_string_cst): New.
7443 (handle_char_store): Call get_string_cst.
7445 2018-01-03 Martin Liska <mliska@suse.cz>
7447 PR tree-optimization/83593
7448 * tree-ssa-strlen.c: Include tree-cfg.h.
7449 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7450 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7451 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7453 (strlen_dom_walker::before_dom_children): Call
7454 gimple_purge_dead_eh_edges. Dump tranformation with details
7456 (strlen_dom_walker::before_dom_children): Update call by adding
7457 new argument cleanup_eh.
7458 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7460 2018-01-03 Martin Liska <mliska@suse.cz>
7463 * cif-code.def (VARIADIC_THUNK): New enum value.
7464 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7467 2018-01-03 Jan Beulich <jbeulich@suse.com>
7469 * sse.md (mov<mode>_internal): Tighten condition for when to use
7470 vmovdqu<ssescalarsize> for TI and OI modes.
7472 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7474 Update copyright years.
7476 2018-01-03 Martin Liska <mliska@suse.cz>
7479 * ipa-visibility.c (function_and_variable_visibility): Skip
7480 functions with noipa attribure.
7482 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7484 * gcc.c (process_command): Update copyright notice dates.
7485 * gcov-dump.c (print_version): Ditto.
7486 * gcov.c (print_version): Ditto.
7487 * gcov-tool.c (print_version): Ditto.
7488 * gengtype.c (create_file): Ditto.
7489 * doc/cpp.texi: Bump @copying's copyright year.
7490 * doc/cppinternals.texi: Ditto.
7491 * doc/gcc.texi: Ditto.
7492 * doc/gccint.texi: Ditto.
7493 * doc/gcov.texi: Ditto.
7494 * doc/install.texi: Ditto.
7495 * doc/invoke.texi: Ditto.
7497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7499 * vector-builder.h (vector_builder::m_full_nelts): Change from
7500 unsigned int to poly_uint64.
7501 (vector_builder::full_nelts): Update prototype accordingly.
7502 (vector_builder::new_vector): Likewise.
7503 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7504 (vector_builder::operator ==): Likewise.
7505 (vector_builder::finalize): Likewise.
7506 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7507 Take the number of elements as a poly_uint64 rather than an
7509 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7510 from unsigned int to poly_uint64.
7511 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7512 (vec_perm_indices::new_vector): Likewise.
7513 (vec_perm_indices::length): Likewise.
7514 (vec_perm_indices::nelts_per_input): Likewise.
7515 (vec_perm_indices::input_nelts): Likewise.
7516 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7517 number of elements per input as a poly_uint64 rather than an
7518 unsigned int. Use the original encoding for variable-length
7519 vectors, rather than clamping each individual element.
7520 For the second and subsequent elements in each pattern,
7521 clamp the step and base before clamping their sum.
7522 (vec_perm_indices::series_p): Handle polynomial element counts.
7523 (vec_perm_indices::all_in_range_p): Likewise.
7524 (vec_perm_indices_to_tree): Likewise.
7525 (vec_perm_indices_to_rtx): Likewise.
7526 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7527 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7528 (tree_vector_builder::new_binary_operation): Handle polynomial
7529 element counts. Return false if we need to know the number
7530 of elements at compile time.
7531 * fold-const.c (fold_vec_perm): Punt if the number of elements
7532 isn't known at compile time.
7534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7536 * vec-perm-indices.h (vec_perm_builder): Change element type
7537 from HOST_WIDE_INT to poly_int64.
7538 (vec_perm_indices::element_type): Update accordingly.
7539 (vec_perm_indices::clamp): Handle polynomial element_types.
7540 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7541 (vec_perm_indices::all_in_range_p): Likewise.
7542 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7544 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7545 polynomial vec_perm_indices element types.
7546 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7547 * fold-const.c (fold_vec_perm): Likewise.
7548 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7549 * tree-vect-generic.c (lower_vec_perm): Likewise.
7550 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7551 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7552 element type to HOST_WIDE_INT.
7554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7555 Alan Hayward <alan.hayward@arm.com>
7556 David Sherwood <david.sherwood@arm.com>
7558 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7559 rather than an int. Use plus_constant.
7560 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7561 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7564 Alan Hayward <alan.hayward@arm.com>
7565 David Sherwood <david.sherwood@arm.com>
7567 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7568 a HOST_WIDE_INT to a poly_int64.
7570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7571 Alan Hayward <alan.hayward@arm.com>
7572 David Sherwood <david.sherwood@arm.com>
7574 * calls.c (load_register_parameters): Cope with polynomial
7575 mode sizes. Require a constant size for BLKmode parameters
7576 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7577 forces a parameter to be padded at the lsb end in order to
7578 fill a complete number of words, require the parameter size
7579 to be ordered wrt UNITS_PER_WORD.
7581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7582 Alan Hayward <alan.hayward@arm.com>
7583 David Sherwood <david.sherwood@arm.com>
7585 * reload1.c (spill_stack_slot_width): Change element type
7586 from unsigned int to poly_uint64_pod.
7587 (alter_reg): Treat mode sizes as polynomial.
7589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7590 Alan Hayward <alan.hayward@arm.com>
7591 David Sherwood <david.sherwood@arm.com>
7593 * reload.c (complex_word_subreg_p): New function.
7594 (reload_inner_reg_of_subreg, push_reload): Use it.
7596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7597 Alan Hayward <alan.hayward@arm.com>
7598 David Sherwood <david.sherwood@arm.com>
7600 * lra-constraints.c (process_alt_operands): Reject matched
7601 operands whose sizes aren't ordered.
7602 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7609 that the mode size is in the set {1, 2, 4, 8, 16}.
7611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7612 Alan Hayward <alan.hayward@arm.com>
7613 David Sherwood <david.sherwood@arm.com>
7615 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7616 Use plus_constant instead of gen_rtx_PLUS.
7618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7619 Alan Hayward <alan.hayward@arm.com>
7620 David Sherwood <david.sherwood@arm.com>
7622 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7623 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7624 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7625 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7626 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7627 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7628 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7629 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7630 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7631 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7633 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7634 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7635 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7636 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7637 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7638 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7639 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7640 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7641 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7642 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7644 * expr.c (emit_move_resolve_push): Treat the input and result
7645 of PUSH_ROUNDING as a poly_int64.
7646 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7647 (emit_push_insn): Likewise.
7648 * lra-eliminations.c (mark_not_eliminable): Likewise.
7649 * recog.c (push_operand): Likewise.
7650 * reload1.c (elimination_effects): Likewise.
7651 * rtlanal.c (nonzero_bits1): Likewise.
7652 * calls.c (store_one_arg): Likewise. Require the padding to be
7653 known at compile time.
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 (emit_single_push_insn_1): Treat mode sizes as polynomial.
7660 Use plus_constant instead of gen_rtx_PLUS.
7662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7663 Alan Hayward <alan.hayward@arm.com>
7664 David Sherwood <david.sherwood@arm.com>
7666 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7670 Alan Hayward <alan.hayward@arm.com>
7671 David Sherwood <david.sherwood@arm.com>
7673 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7674 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7675 via stack temporaries. Treat the mode size as polynomial too.
7677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7678 Alan Hayward <alan.hayward@arm.com>
7679 David Sherwood <david.sherwood@arm.com>
7681 * expr.c (expand_expr_real_2): When handling conversions involving
7682 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7683 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7684 as a poly_uint64 too.
7686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7687 Alan Hayward <alan.hayward@arm.com>
7688 David Sherwood <david.sherwood@arm.com>
7690 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
7697 REGMODE_NATURAL_SIZE.
7698 * expmed.c (store_bit_field_1): Likewise.
7699 * expr.c (store_constructor): Likewise.
7700 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7701 and polynomial REGMODE_NATURAL_SIZE.
7702 (gen_lowpart_common): Likewise.
7703 * reginfo.c (record_subregs_of_mode): Likewise.
7704 * rtlanal.c (read_modify_subreg_p): Likewise.
7706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7707 Alan Hayward <alan.hayward@arm.com>
7708 David Sherwood <david.sherwood@arm.com>
7710 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7711 numbers of elements.
7713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7714 Alan Hayward <alan.hayward@arm.com>
7715 David Sherwood <david.sherwood@arm.com>
7717 * match.pd: Cope with polynomial numbers of vector elements.
7719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7720 Alan Hayward <alan.hayward@arm.com>
7721 David Sherwood <david.sherwood@arm.com>
7723 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7724 in a POINTER_PLUS_EXPR.
7726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7727 Alan Hayward <alan.hayward@arm.com>
7728 David Sherwood <david.sherwood@arm.com>
7730 * omp-simd-clone.c (simd_clone_subparts): New function.
7731 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7732 (ipa_simd_modify_function_body): Likewise.
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-generic.c (nunits_for_known_piecewise_op): New function.
7739 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7740 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7741 (expand_vector_condition, vector_element): Likewise.
7742 (subparts_gt): New function.
7743 (get_compute_type): Use subparts_gt.
7744 (count_type_subparts): Delete.
7745 (expand_vector_operations_1): Use subparts_gt instead of
7746 count_type_subparts.
7748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7749 Alan Hayward <alan.hayward@arm.com>
7750 David Sherwood <david.sherwood@arm.com>
7752 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7753 (vect_compile_time_alias): ...this new function. Do the calculation
7754 on poly_ints rather than trees.
7755 (vect_prune_runtime_alias_test_list): Update call accordingly.
7757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7758 Alan Hayward <alan.hayward@arm.com>
7759 David Sherwood <david.sherwood@arm.com>
7761 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7763 (vect_schedule_slp_instance): Likewise.
7765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7766 Alan Hayward <alan.hayward@arm.com>
7767 David Sherwood <david.sherwood@arm.com>
7769 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7770 constant and extern definitions for variable-length vectors.
7771 (vect_get_constant_vectors): Note that the number of units
7772 is known to be constant.
7774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7775 Alan Hayward <alan.hayward@arm.com>
7776 David Sherwood <david.sherwood@arm.com>
7778 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7779 of units as polynomial. Choose between WIDE and NARROW based
7782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7783 Alan Hayward <alan.hayward@arm.com>
7784 David Sherwood <david.sherwood@arm.com>
7786 * tree-vect-stmts.c (simd_clone_subparts): New function.
7787 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7790 Alan Hayward <alan.hayward@arm.com>
7791 David Sherwood <david.sherwood@arm.com>
7793 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7794 vectors as polynomial. Use build_index_vector for
7797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7798 Alan Hayward <alan.hayward@arm.com>
7799 David Sherwood <david.sherwood@arm.com>
7801 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7802 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7803 for variable-length vectors.
7804 (vectorizable_mask_load_store): Treat the number of units as
7805 polynomial, asserting that it is constant if the condition has
7806 already been enforced.
7807 (vectorizable_store, vectorizable_load): Likewise.
7809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7810 Alan Hayward <alan.hayward@arm.com>
7811 David Sherwood <david.sherwood@arm.com>
7813 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7814 of units as polynomial. Punt if we can't tell at compile time
7815 which vector contains the final result.
7817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7818 Alan Hayward <alan.hayward@arm.com>
7819 David Sherwood <david.sherwood@arm.com>
7821 * tree-vect-loop.c (vectorizable_induction): Treat the number
7822 of units as polynomial. Punt on SLP inductions. Use an integer
7823 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7824 cast of such a series for variable-length floating-point
7827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7828 Alan Hayward <alan.hayward@arm.com>
7829 David Sherwood <david.sherwood@arm.com>
7831 * tree.h (build_index_vector): Declare.
7832 * tree.c (build_index_vector): New function.
7833 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7834 of units as polynomial, forcibly converting it to a constant if
7835 vectorizable_reduction has already enforced the condition.
7836 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7837 to create a {1,2,3,...} vector.
7838 (vectorizable_reduction): Treat the number of units as polynomial.
7839 Choose vectype_in based on the largest scalar element size rather
7840 than the smallest number of units. Enforce the restrictions
7843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7844 Alan Hayward <alan.hayward@arm.com>
7845 David Sherwood <david.sherwood@arm.com>
7847 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7848 number of units as polynomial.
7850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7851 Alan Hayward <alan.hayward@arm.com>
7852 David Sherwood <david.sherwood@arm.com>
7854 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7855 * target.def (autovectorize_vector_sizes): Return the vector sizes
7856 by pointer, using vector_sizes rather than a bitmask.
7857 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7858 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7859 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7861 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7862 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7863 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7864 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7865 * omp-general.c (omp_max_vf): Likewise.
7866 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7867 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7868 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7869 * tree-vect-slp.c (vect_slp_bb): Likewise.
7870 * doc/tm.texi: Regenerate.
7871 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7873 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7874 the vector size as a poly_uint64 rather than an unsigned int.
7875 (current_vector_size): Change from an unsigned int to a poly_uint64.
7876 (get_vectype_for_scalar_type): Update accordingly.
7877 * tree.h (build_truth_vector_type): Take the size and number of
7878 units as a poly_uint64 rather than an unsigned int.
7879 (build_vector_type): Add a temporary overload that takes
7880 the number of units as a poly_uint64 rather than an unsigned int.
7881 * tree.c (make_vector_type): Likewise.
7882 (build_truth_vector_type): Take the number of units as a poly_uint64
7883 rather than an unsigned int.
7885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7886 Alan Hayward <alan.hayward@arm.com>
7887 David Sherwood <david.sherwood@arm.com>
7889 * target.def (get_mask_mode): Take the number of units and length
7890 as poly_uint64s rather than unsigned ints.
7891 * targhooks.h (default_get_mask_mode): Update accordingly.
7892 * targhooks.c (default_get_mask_mode): Likewise.
7893 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7894 * doc/tm.texi: Regenerate.
7896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7897 Alan Hayward <alan.hayward@arm.com>
7898 David Sherwood <david.sherwood@arm.com>
7900 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7901 * omp-general.c (omp_max_vf): Likewise.
7902 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7903 (expand_omp_simd): Handle polynomial safelen.
7904 * omp-low.c (omplow_simd_context): Add a default constructor.
7905 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7906 (lower_rec_simd_input_clauses): Update accordingly.
7907 (lower_rec_input_clauses): Likewise.
7909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7910 Alan Hayward <alan.hayward@arm.com>
7911 David Sherwood <david.sherwood@arm.com>
7913 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7914 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7915 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7916 (vect_analyze_slp_cost): Likewise.
7917 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7918 (vect_model_load_cost): Likewise.
7920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7921 Alan Hayward <alan.hayward@arm.com>
7922 David Sherwood <david.sherwood@arm.com>
7924 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7925 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7926 from an unsigned int * to a poly_uint64_pod *.
7927 (calculate_unrolling_factor): New function.
7928 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7931 Alan Hayward <alan.hayward@arm.com>
7932 David Sherwood <david.sherwood@arm.com>
7934 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7935 from an unsigned int to a poly_uint64.
7936 (_loop_vec_info::slp_unrolling_factor): Likewise.
7937 (_loop_vec_info::vectorization_factor): Change from an int
7939 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7940 (vect_get_num_vectors): New function.
7941 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7942 (vect_get_num_copies): Use vect_get_num_vectors.
7943 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7944 to an unsigned int *.
7945 (vect_analyze_data_refs): Change min_vf from an int * to a
7947 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7948 than an unsigned HOST_WIDE_INT.
7949 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7950 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7951 to an unsigned int *.
7952 (vect_analyze_data_ref_dependences): Likewise.
7953 (vect_compute_data_ref_alignment): Handle polynomial vf.
7954 (vect_enhance_data_refs_alignment): Likewise.
7955 (vect_prune_runtime_alias_test_list): Likewise.
7956 (vect_shift_permute_load_chain): Likewise.
7957 (vect_supportable_dr_alignment): Likewise.
7958 (dependence_distance_ge_vf): Take the vectorization factor as a
7959 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7960 (vect_analyze_data_refs): Change min_vf from an int * to a
7962 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7963 vfm1 as a poly_uint64 rather than an int. Make the same change
7964 for the returned bound_scalar.
7965 (vect_gen_vector_loop_niters): Handle polynomial vf.
7966 (vect_do_peeling): Likewise. Update call to
7967 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7968 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7970 * tree-vect-loop.c (vect_determine_vectorization_factor)
7971 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7972 (vect_get_known_peeling_cost): Likewise.
7973 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7974 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7975 (vect_transform_loop): Likewise. Use the lowest possible VF when
7976 updating the upper bounds of the loop.
7977 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7979 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7980 polynomial unroll factors.
7981 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7982 (vect_make_slp_decision): Likewise.
7983 (vect_supported_load_permutation_p): Likewise, and polynomial
7985 (vect_analyze_slp_cost): Handle polynomial vf.
7986 (vect_slp_analyze_node_operations): Likewise.
7987 (vect_slp_analyze_bb_1): Likewise.
7988 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7989 than an unsigned HOST_WIDE_INT.
7990 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7991 (vectorizable_load): Handle polynomial vf.
7992 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7994 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7997 Alan Hayward <alan.hayward@arm.com>
7998 David Sherwood <david.sherwood@arm.com>
8000 * match.pd: Handle bit operations involving three constants
8001 and try to fold one pair.
8003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8005 * tree-vect-loop-manip.c: Include gimple-fold.h.
8006 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8007 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8008 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8009 Add a path that uses a step of VF instead of 1, but disable it
8011 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8012 and niters_no_overflow parameters. Update calls to
8013 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8014 Create a new SSA name if the latter choses to use a ste other
8015 than zero, and return it via niters_vector_mult_vf_var.
8016 * tree-vect-loop.c (vect_transform_loop): Update calls to
8017 vect_do_peeling, vect_gen_vector_loop_niters and
8018 slpeel_make_loop_iterate_ntimes.
8019 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8020 (vect_gen_vector_loop_niters): Update declarations after above changes.
8022 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8024 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8025 128-bit round to integer instructions.
8026 (ceil<mode>2): Likewise.
8027 (btrunc<mode>2): Likewise.
8028 (round<mode>2): Likewise.
8030 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8032 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8033 unaligned VSX load/store on P8/P9.
8034 (expand_block_clear): Allow the use of unaligned VSX
8035 load/store on P8/P9.
8037 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8039 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8041 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8042 swap associated with both a load and a store.
8044 2018-01-02 Andrew Waterman <andrew@sifive.com>
8046 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8047 * config/riscv/riscv.md (clear_cache): Use it.
8049 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8051 * web.c: Remove out-of-date comment.
8053 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8055 * expr.c (fixup_args_size_notes): Check that any existing
8056 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8057 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8058 (emit_single_push_insn): ...here.
8060 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8062 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8063 (const_vector_encoded_nelts): New function.
8064 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8065 (const_vector_int_elt, const_vector_elt): Declare.
8066 * emit-rtl.c (const_vector_int_elt_1): New function.
8067 (const_vector_elt): Likewise.
8068 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8069 of CONST_VECTOR_ELT.
8071 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8073 * expr.c: Include rtx-vector-builder.h.
8074 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8075 directly on the tree encoding.
8076 (const_vector_from_tree): Likewise.
8077 * optabs.c: Include rtx-vector-builder.h.
8078 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8079 sequence of "u" values.
8080 * vec-perm-indices.c: Include rtx-vector-builder.h.
8081 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8082 directly on the vec_perm_indices encoding.
8084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8086 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8087 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8088 * rtx-vector-builder.h: New file.
8089 * rtx-vector-builder.c: Likewise.
8090 * rtl.h (rtx_def::u2): Add a const_vector field.
8091 (CONST_VECTOR_NPATTERNS): New macro.
8092 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8093 (CONST_VECTOR_DUPLICATE_P): Likewise.
8094 (CONST_VECTOR_STEPPED_P): Likewise.
8095 (CONST_VECTOR_ENCODED_ELT): Likewise.
8096 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8097 (unwrap_const_vec_duplicate): Likewise.
8098 (const_vec_series_p): Check for a non-duplicated vector encoding.
8099 Say that the function only returns true for integer vectors.
8100 * emit-rtl.c: Include rtx-vector-builder.h.
8101 (gen_const_vec_duplicate_1): Delete.
8102 (gen_const_vector): Call gen_const_vec_duplicate instead of
8103 gen_const_vec_duplicate_1.
8104 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8105 (gen_const_vec_duplicate): Use rtx_vector_builder.
8106 (gen_const_vec_series): Likewise.
8107 (gen_rtx_CONST_VECTOR): Likewise.
8108 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8109 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8110 Build a new vector rather than modifying a CONST_VECTOR in-place.
8111 (handle_special_swappables): Update call accordingly.
8112 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8113 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8114 Build a new vector rather than modifying a CONST_VECTOR in-place.
8115 (handle_special_swappables): Update call accordingly.
8117 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8119 * simplify-rtx.c (simplify_const_binary_operation): Use
8120 CONST_VECTOR_ELT instead of XVECEXP.
8122 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8124 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8125 the selector elements to be different from the data elements
8126 if the selector is a VECTOR_CST.
8127 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8128 ssizetype for the selector.
8130 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8132 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8133 before testing each element individually.
8134 * tree-vect-generic.c (lower_vec_perm): Likewise.
8136 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8138 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8139 * selftest-run-tests.c (selftest::run_tests): Call it.
8140 * vector-builder.h (vector_builder::operator ==): New function.
8141 (vector_builder::operator !=): Likewise.
8142 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8143 (vec_perm_indices::all_from_input_p): New function.
8144 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8145 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8146 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8147 instead of reading the VECTOR_CST directly. Detect whether both
8148 vector inputs are the same before constructing the vec_perm_indices,
8149 and update the number of inputs argument accordingly. Use the
8150 utility functions added above. Only construct sel2 if we need to.
8152 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8154 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8155 the broadcast of the low byte.
8156 (expand_mult_highpart): Use an explicit encoding for the permutes.
8157 * optabs-query.c (can_mult_highpart_p): Likewise.
8158 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8159 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8160 (vectorizable_bswap): Likewise.
8161 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8162 explicit encoding for the power-of-2 permutes.
8163 (vect_permute_store_chain): Likewise.
8164 (vect_grouped_load_supported): Likewise.
8165 (vect_permute_load_chain): Likewise.
8167 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8169 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8170 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8171 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8172 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8173 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8174 (vect_gen_perm_mask_any): Likewise.
8176 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8178 * int-vector-builder.h: New file.
8179 * vec-perm-indices.h: Include int-vector-builder.h.
8180 (vec_perm_indices): Redefine as an int_vector_builder.
8181 (auto_vec_perm_indices): Delete.
8182 (vec_perm_builder): Redefine as a stand-alone class.
8183 (vec_perm_indices::vec_perm_indices): New function.
8184 (vec_perm_indices::clamp): Likewise.
8185 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8186 (vec_perm_indices::new_vector): New function.
8187 (vec_perm_indices::new_expanded_vector): Update for new
8188 vec_perm_indices class.
8189 (vec_perm_indices::rotate_inputs): New function.
8190 (vec_perm_indices::all_in_range_p): Operate directly on the
8191 encoded form, without computing elided elements.
8192 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8193 encoding. Update for new vec_perm_indices class.
8194 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8195 the given vec_perm_builder.
8196 (expand_vec_perm_var): Update vec_perm_builder constructor.
8197 (expand_mult_highpart): Use vec_perm_builder instead of
8198 auto_vec_perm_indices.
8199 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8200 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8201 or double series encoding as appropriate.
8202 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8203 vec_perm_indices instead of auto_vec_perm_indices.
8204 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8205 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8206 (vect_permute_store_chain): Likewise.
8207 (vect_grouped_load_supported): Likewise.
8208 (vect_permute_load_chain): Likewise.
8209 (vect_shift_permute_load_chain): Likewise.
8210 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8211 (vect_transform_slp_perm_load): Likewise.
8212 (vect_schedule_slp_instance): Likewise.
8213 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8214 (vectorizable_mask_load_store): Likewise.
8215 (vectorizable_bswap): Likewise.
8216 (vectorizable_store): Likewise.
8217 (vectorizable_load): Likewise.
8218 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8219 vec_perm_indices instead of auto_vec_perm_indices. Use
8220 tree_to_vec_perm_builder to read the vector from a tree.
8221 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8222 vec_perm_builder instead of a vec_perm_indices.
8223 (have_whole_vector_shift): Use vec_perm_builder and
8224 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8225 truncation to calc_vec_perm_mask_for_shift.
8226 (vect_create_epilog_for_reduction): Likewise.
8227 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8228 from auto_vec_perm_indices to vec_perm_indices.
8229 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8230 instead of changing individual elements.
8231 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8232 the vector in d.perm.
8233 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8234 from auto_vec_perm_indices to vec_perm_indices.
8235 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8236 instead of changing individual elements.
8237 (arm_vectorize_vec_perm_const): Use new_vector to install
8238 the vector in d.perm.
8239 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8240 Update vec_perm_builder constructor.
8241 (rs6000_expand_interleave): Likewise.
8242 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8243 (rs6000_expand_interleave): Likewise.
8245 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8247 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8248 to qimode could truncate the indices.
8249 * optabs.c (expand_vec_perm_var): Likewise.
8251 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8253 * Makefile.in (OBJS): Add vec-perm-indices.o.
8254 * vec-perm-indices.h: New file.
8255 * vec-perm-indices.c: Likewise.
8256 * target.h (vec_perm_indices): Replace with a forward class
8258 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8259 * optabs.h: Include vec-perm-indices.h.
8260 (expand_vec_perm): Delete.
8261 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8262 (expand_vec_perm_const): Declare.
8263 * target.def (vec_perm_const_ok): Replace with...
8264 (vec_perm_const): ...this new hook.
8265 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8266 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8267 * doc/tm.texi: Regenerate.
8268 * optabs.def (vec_perm_const): Delete.
8269 * doc/md.texi (vec_perm_const): Likewise.
8270 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8271 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8272 expand_vec_perm for constant permutation vectors. Assert that
8273 the mode of variable permutation vectors is the integer equivalent
8274 of the mode that is being permuted.
8275 * optabs-query.h (selector_fits_mode_p): Declare.
8276 * optabs-query.c: Include vec-perm-indices.h.
8277 (selector_fits_mode_p): New function.
8278 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8279 is defined, instead of checking whether the vec_perm_const_optab
8280 exists. Use targetm.vectorize.vec_perm_const instead of
8281 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8282 fit in the vector mode before using a variable permute.
8283 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8284 vec_perm_indices instead of an rtx.
8285 (expand_vec_perm): Replace with...
8286 (expand_vec_perm_const): ...this new function. Take the selector
8287 as a vec_perm_indices rather than an rtx. Also take the mode of
8288 the selector. Update call to shift_amt_for_vec_perm_mask.
8289 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8290 Use vec_perm_indices::new_expanded_vector to expand the original
8291 selector into bytes. Check whether the indices fit in the vector
8292 mode before using a variable permute.
8293 (expand_vec_perm_var): Make global.
8294 (expand_mult_highpart): Use expand_vec_perm_const.
8295 * fold-const.c: Includes vec-perm-indices.h.
8296 * tree-ssa-forwprop.c: Likewise.
8297 * tree-vect-data-refs.c: Likewise.
8298 * tree-vect-generic.c: Likewise.
8299 * tree-vect-loop.c: Likewise.
8300 * tree-vect-slp.c: Likewise.
8301 * tree-vect-stmts.c: Likewise.
8302 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8304 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8305 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8306 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8307 (aarch64_vectorize_vec_perm_const): ...this new function.
8308 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8309 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8310 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8311 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8312 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8313 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8314 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8316 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8317 check for NEON modes.
8318 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8319 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8320 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8321 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8323 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8324 the old VEC_PERM_CONST conditions.
8325 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8326 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8327 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8328 (ia64_vectorize_vec_perm_const_ok): Merge into...
8329 (ia64_vectorize_vec_perm_const): ...this new function.
8330 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8331 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8332 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8333 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8334 * config/mips/mips.c (mips_expand_vec_perm_const)
8335 (mips_vectorize_vec_perm_const_ok): Merge into...
8336 (mips_vectorize_vec_perm_const): ...this new function.
8337 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8338 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8339 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8340 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8341 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8342 (rs6000_expand_vec_perm_const): Delete.
8343 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8345 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8346 (altivec_expand_vec_perm_const_le): Take each operand individually.
8347 Operate on constant selectors rather than rtxes.
8348 (altivec_expand_vec_perm_const): Likewise. Update call to
8349 altivec_expand_vec_perm_const_le.
8350 (rs6000_expand_vec_perm_const): Delete.
8351 (rs6000_vectorize_vec_perm_const_ok): Delete.
8352 (rs6000_vectorize_vec_perm_const): New function.
8353 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8354 an element count and rtx array.
8355 (rs6000_expand_extract_even): Update call accordingly.
8356 (rs6000_expand_interleave): Likewise.
8357 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8358 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8359 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8360 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8361 (rs6000_expand_vec_perm_const): Delete.
8362 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8363 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8364 (altivec_expand_vec_perm_const_le): Take each operand individually.
8365 Operate on constant selectors rather than rtxes.
8366 (altivec_expand_vec_perm_const): Likewise. Update call to
8367 altivec_expand_vec_perm_const_le.
8368 (rs6000_expand_vec_perm_const): Delete.
8369 (rs6000_vectorize_vec_perm_const_ok): Delete.
8370 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8371 reference to the SPE evmerge intructions.
8372 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8373 an element count and rtx array.
8374 (rs6000_expand_extract_even): Update call accordingly.
8375 (rs6000_expand_interleave): Likewise.
8376 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8377 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8379 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8381 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8383 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8384 vector mode and that that mode matches the mode of the data
8386 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8387 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8388 directly using expand_vec_perm_1 when forcing selectors into
8390 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8392 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8394 * optabs-query.h (can_vec_perm_p): Delete.
8395 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8396 * optabs-query.c (can_vec_perm_p): Split into...
8397 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8398 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8399 particular selector is valid.
8400 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8401 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8402 (vect_grouped_load_supported): Likewise.
8403 (vect_shift_permute_load_chain): Likewise.
8404 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8405 (vect_transform_slp_perm_load): Likewise.
8406 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8407 (vectorizable_bswap): Likewise.
8408 (vect_gen_perm_mask_checked): Likewise.
8409 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8410 implementations of variable permutation vectors into account
8411 when deciding which selector to use.
8412 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8413 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8414 with a false third argument.
8415 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8416 to test whether the constant selector is valid and can_vec_perm_var_p
8417 to test whether a variable selector is valid.
8419 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8421 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8422 * optabs-query.c (can_vec_perm_p): Likewise.
8423 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8424 instead of vec_perm_indices.
8425 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8426 (vect_gen_perm_mask_checked): Likewise,
8427 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8428 (vect_gen_perm_mask_checked): Likewise,
8430 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8432 * optabs-query.h (qimode_for_vec_perm): Declare.
8433 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8434 (qimode_for_vec_perm): ...this new function.
8435 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8437 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8439 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8440 does not have a conditional at the top.
8442 2018-01-02 Richard Biener <rguenther@suse.de>
8444 * ipa-inline.c (big_speedup_p): Fix expression.
8446 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8449 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8452 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8456 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8457 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8458 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8459 cond_taken_branch_cost 3->4.
8461 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8463 PR tree-optimization/83581
8464 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8465 TODO_cleanup_cfg if any changes have been made.
8468 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8469 convert_modes if target mode has the right side, but different mode
8473 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8474 last argument when extracting from CONCAT. If either from_real or
8475 from_imag is NULL, use expansion through memory. If result is not
8476 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8477 the parts directly to inner mode, if even that fails, use expansion
8481 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8482 check for bswap in mode rather than HImode and use that in expand_unop
8485 Copyright (C) 2018 Free Software Foundation, Inc.
8487 Copying and distribution of this file, with or without modification,
8488 are permitted in any medium without royalty provided the copyright
8489 notice and this notice are preserved.