1 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10 (dw_val_node): Add val_symbolic_view.
11 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
12 (symview_upper_bound): New.
13 (new_line_info_table): Initialize symviews_since_reset.
14 (dwarf2out_source_line): Count symviews_since_reset and set
16 (dw_val_equal_p): Handle symview.
17 (add_AT_symview): New.
18 (print_dw_val): Handle symview.
19 (attr_checksum, attr_checksum_ordered): Likewise.
20 (same_dw_val_p, size_of_die): Likewise.
21 (value_format, output_die): Likewise.
22 (add_high_low_attributes): Use add_AT_symview for entry_view.
23 (dwarf2out_finish): Reset symview_upper_bound, clear
26 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
29 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
30 Add strict argument and use it.
31 (rs6000_split_multireg_move): Update for new strict argument.
32 (mem_operand_gpr): Disallow all non-offsettable addresses.
33 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
35 2018-03-09 Jakub Jelinek <jakub@redhat.com>
38 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
39 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
40 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
43 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
44 decl, use remap_type if we want to use the type.
46 2018-03-09 Martin Sebor <msebor@redhat.com>
48 PR tree-optimization/84526
49 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
51 (builtin_access::generic_overlap): Be prepared to handle non-array
54 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
56 PR rtl-optimization/84682
57 * lra-constraints.c (process_address_1): Check is_address flag
58 for address constraints.
59 (process_alt_operands): Likewise.
60 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
61 preprocess_constraints.
62 * recog.h (preprocess_constraints): Add oploc parameter.
64 * recog.c (preprocess_constraints): Test address_operand for
65 CT_ADDRESS constraints.
67 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
70 * lra-assigns.c (assign_by_spills): Return a flag of reload
71 assignment failure. Do not process the reload assignment
72 failures. Do not spill other reload pseudos if they has the same
74 (lra_assign): Add a return arg. Set up from the result of
75 assign_by_spills call.
76 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
77 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
78 usage_insns if it is not NULL.
79 (spill_hard_reg_in_range): New function.
80 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
81 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
83 (lra_assign): Change prototype.
84 * lra.c (lra): Add code to deal with fails by splitting hard reg
87 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
90 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
91 Accept complain bool parameter. Only emit errors if it is true.
92 (arm_parse_cpu_option_name): Likewise.
93 (arm_target_thumb_only): Adjust callers of the above.
94 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
95 prototype to take a default true bool parameter.
96 (arm_parse_arch_option_name): Likewise.
98 2018-03-09 David Malcolm <dmalcolm@redhat.com>
99 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
103 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
104 * configure: Regenerate.
105 * configure.ac ("linker --version-script option"): New.
106 ("linker soname option"): New.
108 2018-03-09 Richard Biener <rguenther@suse.de>
110 PR tree-optimization/84775
111 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
112 immediate uses of predicate stmts and mark them modified.
115 PR tree-optimization/84178
116 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
118 (version_loop_for_if_conversion): Delay update_ssa call.
119 (tree_if_conversion): Delay update_ssa until after predicate
122 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
125 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
126 when the function accesses prior frames.
128 2018-03-08 Jakub Jelinek <jakub@redhat.com>
131 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
132 gen_llsym, otherwise call maybe_gen_llsym.
135 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
136 has ',' character inside of it.
138 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
141 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
142 as clobbering CC_REGNUM.
144 2018-03-08 Richard Biener <rguenther@suse.de>
147 * tree-scalar-evolution.c: Include tree-into-ssa.h.
148 (follow_copies_to_constant): Do not follow SSA names registered
151 2018-03-08 Richard Biener <rguenther@suse.de>
153 PR tree-optimization/84178
154 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
156 (version_loop_for_if_conversion): Delay update_ssa call.
157 (tree_if_conversion): Delay update_ssa until after predicate
160 2018-03-08 David Malcolm <dmalcolm@redhat.com>
162 PR tree-optimization/84178
163 * tree-if-conv.c (release_bb_predicate): Remove the
164 the assertion that the stmts have NULL use_ops.
165 Discard the statements, asserting that they haven't
166 yet been added to a BB.
168 2018-03-08 Richard Biener <rguenther@suse.de>
170 PR tree-optimization/84746
171 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
172 (phi_translate): Pass in destination ANTIC_OUT set.
173 (phi_translate_1): Likewise. For a simplified result lookup
174 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
175 (phi_translate_set): Adjust.
176 (do_pre_regular_insertion): Likewise.
177 (do_pre_partial_partial_insertion): Likewise.
179 2018-03-08 Martin Liska <mliska@suse.cz>
181 PR gcov-profile/84735
182 * doc/gcov.texi: Document usage of profile files.
183 * gcov-io.h: Document changes in the format.
185 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
189 * dwarf2out.c (struct dw_line_info_table): Update comments for
191 (FORCE_RESET_NEXT_VIEW): New.
192 (FORCE_RESETTING_VIEW_P): New.
193 (RESETTING_VIEW_P): Check for -1 too.
194 (ZERO_VIEW_P): Likewise.
195 (new_line_info_table): Force-reset next view.
196 (dwarf2out_begin_function): Likewise.
197 (dwarf2out_source_line): Simplify zero_view_p initialization.
198 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
199 view directly. Omit view when omitting .loc at line 0.
201 2018-03-08 Jakub Jelinek <jakub@redhat.com>
203 PR tree-optimization/84740
204 * tree-switch-conversion.c (process_switch): Call build_constructors
205 only if info.phi_count is non-zero.
207 PR tree-optimization/84739
208 * tree-tailcall.c (find_tail_calls): Check call arguments against
209 DECL_ARGUMENTS (current_function_decl) rather than
210 DECL_ARGUMENTS (func) when checking for tail recursion.
212 2018-03-07 Jakub Jelinek <jakub@redhat.com>
214 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
215 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
216 Volker Reichelt's entry and add entries for people that perform
217 GCC fuzzy testing and report numerous bugs.
219 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
222 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
223 readonly data in sdata, if that is disabled.
224 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
225 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
226 -mreadonly-in-sdata option.
228 2018-03-07 Martin Sebor <msebor@redhat.com>
230 PR tree-optimization/84468
231 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
232 basic block when looking for nul assignment.
234 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
237 * except.h (output_function_exception_table): Adjust prototype.
238 * except.c (output_function_exception_table): Remove FNNAME parameter
239 and add SECTION parameter. Ouput one part of the table at a time.
240 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
241 the first part of the exception table and emit unwind directives.
242 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
243 (i386_pe_seh_cold_init): Likewise.
244 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
245 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
246 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
247 (ix86_output_call_insn): Emit a nop in one more case for SEH.
248 * config/i386/winnt.c: Include except.h.
249 (struct seh_frame_state): Add reg_offset, after_prologue and
250 in_cold_section fields.
251 (i386_pe_seh_end_prologue): Set seh->after_prologue.
252 (i386_pe_seh_cold_init): New function.
253 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
254 to seh->in_cold_section.
255 (seh_emit_push): Record the offset of the push.
256 (seh_emit_save): Record the offet of the save.
257 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
258 Test seh->after_prologue to disregard the epilogue.
259 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
260 (i386_pe_end_cold_function): New function.
262 2018-03-07 Jakub Jelinek <jakub@redhat.com>
265 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
266 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
269 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
271 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
272 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
275 * multiple_target.c: Include tree-inline.h and intl.h.
276 (expand_target_clones): Diagnose and fail if node->definition and
277 !tree_versionable_function_p (node->decl).
279 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
281 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
283 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
284 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
285 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
287 2018-03-06 Jakub Jelinek <jakub@redhat.com>
290 * combine.c (try_combine): Use reg_or_subregno instead of handling
291 just paradoxical SUBREGs and REGs.
293 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
295 * config/arc/arc.c (arc_finalize_pic): Remove function.
296 (arc_must_save_register): We use single base PIC register, remove
297 checks to save/restore the PIC register.
298 (arc_expand_prologue): Likewise.
299 * config/arc/arc-protos.h (arc_set_default_type_attributes):
301 (arc_verify_short): Likewise.
302 (arc_attr_type): Likewise.
303 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
304 (walk_stores): Likewise.
305 (arc_address_cost): Make it static.
306 (arc_verify_short): Likewise.
307 (branch_dest): Likewise.
308 (arc_attr_type): Likewise.
309 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
310 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
311 (arc_final_prescan_insn): Remove inserting the nops due to
312 hardware hazards. It is done in reorg step.
313 (insn_length_variant_t): Remove.
314 (insn_length_parameters_t): Likewise.
315 (arc_insn_length_parameters): Likewise.
316 (arc_get_insn_variants): Likewise.
317 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
319 2018-03-06 Jakub Jelinek <jakub@redhat.com>
322 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
325 PR tree-optimization/84687
326 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
328 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
330 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
332 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
333 Rename to ppc_speculation_barrier.
334 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
335 __builtin_ppc_speculation_barrier.
337 2018-03-05 Jakub Jelinek <jakub@redhat.com>
340 * combine.c (combine_simplify_rtx): Don't try to simplify if
341 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
344 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
346 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
347 to 32 bytes when compiling for POWER9.
349 2018-03-05 Jakub Jelinek <jakub@redhat.com>
352 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
353 regparm >= 3 with no arg reg available also for calls with
354 flag_force_indirect_call. Pass decl to ix86_function_regparm.
357 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
359 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
361 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
364 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
366 2018-03-05 Richard Biener <rguenther@suse.de>
368 PR tree-optimization/84486
369 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
370 When inserting a __builtin_assume_aligned call set the LHS
371 SSA name alignment info accordingly.
373 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
375 PR tree-optimization/84114
376 * config/aarch64/aarch64.c (aarch64_reassociation_width)
377 Avoid reassociation of FLOAT_MODE addition.
379 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
381 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
382 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
383 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
384 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
385 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
386 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
387 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
389 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
390 __builtin_ia32_wbinvd): New builtins.
391 (SPECIAL_ARGS2): New.
392 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
393 (SPECIAL_ARGS2): New.
394 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
395 (ix86_valid_target_attribute_inner_p): Ditto.
396 (ix86_init_mmx_sse_builtins): Add special_args2.
397 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
398 TARGET_WBNOINVD_P): New.
399 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
400 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
401 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
402 * config/i386/immintrin.h (_wbinvd): New intrinsic.
403 * config/i386/pconfigintrin.h: New file.
404 * config/i386/wbnoinvdintrin.h: Ditto.
405 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
406 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
408 2018-03-05 Richard Biener <rguenther@suse.de>
410 PR tree-optimization/84670
411 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
413 (BB_VISITED_WITH_VISITED_SUCCS): New define.
414 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
415 (compute_antic_aux): Only assert the number of values in ANTIC_IN
416 doesn't grow if all successors (recursively) were visited at least
419 2018-03-05 Richard Biener <rguenther@suse.de>
421 PR tree-optimization/84650
422 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
423 if executed in the loop pipeline.
425 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
427 * doc/configfiles.texi (Configuration Files): Move info about
428 conditionalizing $target-protos.h to...
429 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
430 differs from $target-protos.h.
432 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
433 Chung-Ju Wu <jasonwucj@gmail.com>
435 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
436 * config/nds32/nds32-multiple.md (setmemsi): Define.
437 * config/nds32/nds32-memory-manipulation.c
438 (nds32_gen_dup_4_byte_to_word_value): New.
439 (emit_setmem_word_loop): New.
440 (emit_setmem_byte_loop): New.
441 (nds32_expand_setmem_loop): New.
442 (nds32_expand_setmem_loop_v3m): New.
443 (nds32_expand_setmem_unroll): New.
444 (nds32_expand_setmem): New.
446 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
447 Chung-Ju Wu <jasonwucj@gmail.com>
449 * config/nds32/nds32-memory-manipulation.c
450 (nds32_emit_load_store): New.
451 (nds32_emit_post_inc_load_store): New.
452 (nds32_emit_mem_move): New.
453 (nds32_emit_mem_move_block): New.
454 (nds32_expand_movmemsi_loop_unknown_size): New.
455 (nds32_expand_movmemsi_loop_known_size): New.
456 (nds32_expand_movmemsi_loop): New.
457 (nds32_expand_movmemsi_unroll): New.
458 (nds32_expand_movmemqi): Rename ...
459 (nds32_expand_movmemsi): ... to this.
460 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
461 (movmemsi): ... to this.
462 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
463 (nds32_expand_movmemsi): ... to this.
465 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
466 Monk Chiang <sh.chiang04@gmail.com>
467 Chung-Ju Wu <jasonwucj@gmail.com>
469 * config/nds32/nds32-protos.h
470 (nds32_expand_load_multiple): New arguments.
471 (nds32_expand_store_multiple): Ditto.
472 (nds32_valid_multiple_load_store): Rename ...
473 (nds32_valid_multiple_load_store_p): ... to this.
474 * config/nds32/nds32-memory-manipulation.c
475 (nds32_expand_load_multiple): Refine implementation.
476 (nds32_expand_store_multiple): Ditto.
477 * config/nds32/nds32-multiple.md
478 (load_multiple): Update nds32_expand_load_multiple interface.
479 (store_multiple): Update nds32_expand_store_multiple interface.
480 * config/nds32/nds32-predicates.c
481 (nds32_valid_multiple_load_store): Rename ...
482 (nds32_valid_multiple_load_store_p): ... to this and refine
484 * config/nds32/predicates.md
485 (nds32_load_multiple_and_update_address_operation): New predicate.
486 (nds32_store_multiple_and_update_address_operation): New predicate.
488 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
489 Chung-Ju Wu <jasonwucj@gmail.com>
491 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
492 (combo): New attribute.
493 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
495 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
497 * config/nds32/nds32.opt: Change -mcmodel= default value.
499 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
500 Monk Chiang <sh.chiang04@gmail.com>
501 Chung-Ju Wu <jasonwucj@gmail.com>
503 * config/nds32/constants.md (unspec_element): New enum.
504 * config/nds32/constraints.md (Umw): New constraint.
505 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
506 * config/nds32/nds32-intrinsic.md: Likewise.
507 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
508 (nds32_valid_smw_lwm_base_p): New.
509 (nds32_output_smw_single_word): New.
510 (nds32_output_lmw_single_word): New.
511 (nds32_expand_unaligned_load): New.
512 (nds32_expand_unaligned_store): New.
513 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
514 (nds32_output_smw_single_word): Declare.
515 (nds32_output_lmw_single_word): Declare.
516 (nds32_expand_unaligned_load): Declare.
517 (nds32_expand_unaligned_store): Declare.
518 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
519 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
520 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
521 NDS32_BUILTIN_UASTORE_DW.
522 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
525 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
526 Kito Cheng <kito.cheng@gmail.com>
527 Chung-Ju Wu <jasonwucj@gmail.com>
529 * config/nds32/nds32-intrinsic.c
530 (nds32_expand_builtin_null_ftype_reg): Delete.
531 (nds32_expand_builtin_reg_ftype_imm): Ditto.
532 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
533 (nds32_read_argument): New.
534 (nds32_legitimize_target): Ditto.
535 (nds32_legitimize_argument): Ditto.
536 (nds32_check_constant_argument): Ditto.
537 (nds32_expand_unop_builtin): Ditto.
538 (nds32_expand_unopimm_builtin): Ditto.
539 (nds32_expand_binop_builtin): Ditto.
540 (nds32_builtin_decl_impl): Ditto.
541 (builtin_description): Ditto.
542 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
543 (nds32_init_builtins_impl): Ditto.
544 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
545 (nds32_builtin_decl): New.
546 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
547 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
549 2018-03-02 Jeff Law <law@redhat.com>
551 * reorg.c (stop_search_p): Handle DEBUG_INSN.
552 (redundant_insn, fill_simple_delay_slots): Likewise.
553 (fill_slots_from_thread): Likewise.
554 * resource.c (mark_referenced_resources): Likewise.
555 (mark_set_resources, find_dead_or_set_registers): Likewise.
557 2018-03-02 Jakub Jelinek <jakub@redhat.com>
559 * substring-locations.h (format_warning_va): Formatting fix for
561 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
563 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
564 * substring-locations.c: Include intl.h.
565 (format_warning_va): Turned into small wrapper around
566 format_warning_n_va, renamed to ...
567 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
568 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
570 (format_warning_at_substring_n): New function.
571 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
572 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
573 format_warning_at_substring with just a shorter name instead of
574 const function pointer.
575 (fmtwarn_n): New function.
576 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
577 appropriate, get rid of all the fmtstr temporaries, move conditionals
578 with G_() wrapped string literals directly into fmtwarn arguments,
579 cast dir.len to (int), formatting fixes.
581 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
583 * doc/invoke.texi: Remove "Cilk Plus" references.
585 2018-03-02 Jakub Jelinek <jakub@redhat.com>
586 Richard Biener <rguenther@suse.de>
589 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
590 for error or warning attributes if CALL_FROM_THUNK_P is set.
593 2018-03-02 Jakub Jelinek <jakub@redhat.com>
596 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
597 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
600 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
601 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
603 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
604 instead of -1U in last predictors element's probability member.
606 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
609 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
610 arguments if they are comparable.
612 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
614 PR tree-optimization/84634
615 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
616 masks and masked_loop_p with a single loop_masks, making sure it's
617 null for bb vectorization.
619 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
621 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
622 (vect_analyze_data_ref_access): Use loop->safe_len rather than
623 loop->force_vectorize to check whether there is no alias.
625 2018-03-02 Jakub Jelinek <jakub@redhat.com>
628 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
630 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
632 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
633 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
634 instead of a loop around prev_real_insn.
635 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
639 * config/i386/i386.c (ix86_print_operand): Use conditional
640 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
643 2018-03-02 Richard Biener <rguenther@suse.de>
645 PR tree-optimization/84427
646 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
647 (bitmap_set_subtract_values): Rewrite to handle multiple
650 (prune_clobbered_mems): Likewise.
651 (phi_translate): Take edge instead of pred/phiblock.
652 (phi_translate_1): Likewise.
653 (phi_translate_set): Likewise. Insert all translated
654 exprs for a value into the set, keeping possibly multiple
655 expressions per value.
656 (compute_antic_aux): Adjust for phi_translate changes.
657 When intersecting union the expressions and prune those
658 not in the final value set, keeping possibly multiple
659 expressions per value. Do not use value-insertion
660 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
661 all expressions. Add verification that the value-sets
662 only shrink during iteration.
663 (compute_partial_antic_aux): Adjust for the phi_translate changes.
664 (do_pre_regular_insertion): Likewise.
665 (do_pre_partial_partial_insertion): Likewise.
667 2018-03-02 Richard Biener <rguenther@suse.de>
670 * config/darwin.c (saved_debug_info_level): New static global.
671 (darwin_asm_lto_start): Disable debug info generation for LTO out.
672 (darwin_asm_lto_end): Restore debug info generation settings.
674 2018-03-01 Martin Liska <mliska@suse.cz>
677 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
680 2018-03-01 Richard Biener <rguenther@suse.de>
683 * dwarf2out.c (gen_variable_die): Properly handle late VLA
684 type annotation with LTO when debug was disabled at compile-time.
686 2018-03-01 Matthew Fortune <mfortune@gmail.com>
688 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
690 (mips_final_postscan_insn): Likewise.
692 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
694 PR rtl-optimization/84528
695 * alias.c (init_alias_target): Add commentary.
696 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
697 a unique base value if the frame pointer is not eliminated
698 to the stack pointer.
700 2018-03-01 Tom de Vries <tom@codesourcery.com>
702 PR rtl-optimization/83327
703 * lra-int.h (hard_regs_spilled_into): Declare.
704 * lra.c (hard_regs_spilled_into): Define.
705 (init_reg_info): Init hard_regs_spilled_into.
706 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
707 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
708 (process_bb_lives): Handle hard_regs_spilled_into.
709 (lra_create_live_ranges_1): Before doing liveness propagation, clear
710 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
712 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
714 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
715 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
716 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
717 * config/rs6000/aix72.h: New file.
719 2018-02-28 Jakub Jelinek <jakub@redhat.com>
721 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
722 instead of warning_at with conditional singular and plural messages
726 * stor-layout.c (update_alignment_for_field): For
727 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
728 && !DECL_PACKED (field), do the alignment update, just use
729 only desired_align instead of MAX (type_align, desired_align)
731 (place_field): Don't do known_align < desired_align handling
732 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
733 is non-NULL, instead do it after rli->prev_field handling and
734 only if not within a bitfield word. For DECL_PACKED (field)
735 use type_align of BITS_PER_UNIT.
737 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
739 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
740 superfluous parentheses and trailing spaces.
742 2018-02-28 Richard Biener <rguenther@suse.de>
744 PR tree-optimization/84584
745 * graphite-scop-detection.c (scop_detection::add_scop): Discard
746 SCoPs with fake exit edge.
748 2018-02-28 Martin Liska <mliska@suse.cz>
751 * timevar.c (timer::print): Fix format to properly print 100%
754 2018-02-28 Richard Biener <rguenther@suse.de>
757 * genmatch.c (capture_info::walk_match): Do not mark
758 captured expressions without operands as expr_p given
759 they act more like predicates and should be subject to
760 "lost tail" side-effect preserving.
762 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
764 PR rtl-optimization/81611
765 * auto-inc-dec.c (attempt_change): Move dead note from
766 mem_insn if it's the next use of regno
767 (find_address): Take address use of reg holding
768 non-incremented value. Add parm to limit search to the named
770 (merge_in_block): Attempt to use a mem insn that is the next
771 use of the original regno.
773 2018-02-27 Martin Sebor <msebor@redhat.com>
776 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
777 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
779 2018-02-27 Martin Sebor <msebor@redhat.com>
782 * diagnostic-core.h (warning_n, error_n, inform_n): Change
783 n argument to unsigned HOST_WIDE_INT.
784 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
785 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
786 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
787 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
789 2018-02-27 Richard Biener <rguenther@suse.de>
791 PR tree-optimization/84512
792 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
793 Do not use the estimate returned from record_stmt_cost for
794 the scalar iteration cost but sum properly using add_stmt_cost.
796 2018-02-27 Richard Biener <rguenther@suse.de>
798 PR tree-optimization/84466
799 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
800 Adjust last change to less strictly validate use operands.
802 2018-02-27 Martin Liska <mliska@suse.cz>
804 PR gcov-profile/84548
805 * gcov.c (process_file): Allow partial overlap and consider it
806 also as group functions.
807 (output_lines): Properly calculate range of lines for a group.
809 2018-02-27 Martin Liska <mliska@suse.cz>
811 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
812 'ggc' suffixes. Change first column width.
813 (timer::print): Fix formatting of the column.
815 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
817 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
818 preserve inline entry blocks for the sake of debug inline
819 entry point markers alone.
820 (remove_unused_locals): Suggest in comments a better place to
821 force the preservation of inline entry blocks that are
822 otherwise unused, but do not preserve them.
824 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
826 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
828 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
831 * config/i386/constraints.md (Bs): Replace
832 ix86_indirect_branch_register with
833 TARGET_INDIRECT_BRANCH_REGISTER.
835 * config/i386/i386.md (indirect_jump): Likewise.
836 (tablejump): Likewise.
837 (*sibcall_memory): Likewise.
838 (*sibcall_value_memory): Likewise.
839 Peepholes of indirect call and jump via memory: Likewise.
840 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
841 (*sibcall_value_GOT_32): Likewise.
842 * config/i386/predicates.md (indirect_branch_operand): Likewise.
843 (GOT_memory_operand): Likewise.
844 (call_insn_operand): Likewise.
845 (sibcall_insn_operand): Likewise.
846 (GOT32_symbol_operand): Likewise.
847 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
849 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
851 PR rtl-optimization/83496
852 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
853 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
854 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
855 redundant insn, if any.
856 (relax_delay_slots): Likewise.
857 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
859 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
861 PR tree-optimization/83965
862 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
863 that grouped statements are part of a reduction chain. Return
864 true if the statement is not marked as a reduction itself but
866 (vect_recog_dot_prod_pattern): Don't check whether the statement
867 is part of a group here.
868 (vect_recog_sad_pattern): Likewise.
869 (vect_recog_widen_sum_pattern): Likewise.
871 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
874 * final.c (rest_of_clean_state): Also look for calls inside sequences.
876 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
879 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
881 (ix86_output_indirect_function_return): New prototype.
882 (ix86_split_simple_return_pop_internal): Likewise.
883 * config/i386/i386.c (indirect_return_via_cx): New.
884 (indirect_return_via_cx_bnd): Likewise.
885 (indirect_thunk_name): Handle return va CX_REG.
886 (output_indirect_thunk_function): Create alias for
887 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
888 (ix86_output_indirect_jmp): Remove the bool argument.
889 (ix86_output_indirect_function_return): New function.
890 (ix86_split_simple_return_pop_internal): Likewise.
891 * config/i386/i386.md (*indirect_jump): Don't pass false
892 to ix86_output_indirect_jmp.
893 (*tablejump_1): Likewise.
894 (simple_return_pop_internal): Change it to define_insn_and_split.
895 Call ix86_split_simple_return_pop_internal to split it for
897 (simple_return_indirect_internal): Call
898 ix86_output_indirect_function_return instead of
899 ix86_output_indirect_jmp.
901 2018-02-26 Jakub Jelinek <jakub@redhat.com>
904 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
905 memset and value initialization afterwards.
907 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
909 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
911 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
914 * common/config/aarch64/aarch64-common.c
915 (aarch_option_optimization_table[]): Switch
916 off fomit-frame-pointer
918 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
919 Chung-Ju Wu <jasonwucj@gmail.com>
921 * config/nds32/nds32-multiple.md (load_multiple): Disallow
923 (store_multiple): Ditto.
925 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
927 * config.gcc: Add --with-cpu support for nds32 target.
928 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
929 * config/nds32/nds32.opt: Add -mcpu= option.
931 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
933 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
934 isel=yes): Warn for these deprecated options.
936 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
938 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
939 ISA_2_5_MASKS_EMBEDDED.
941 2018-02-23 Jakub Jelinek <jakub@redhat.com>
943 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
944 p->max as pointers rather than using iterative_hash_expr.
946 2018-02-23 Carl Love <cel@us.ibm.com>
948 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
949 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
951 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
952 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
953 P8V_BUILTIN_VEC_VUNSIGNED2.
955 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
958 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
959 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
961 * lra-constraints.c (curr_insn_transform): Set up
962 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
964 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
965 * lra-eliminations.c (spill_pseudos): Ditto.
966 (process_insn_for_elimination): Ditto.
967 * lra-lives.c (reg_early_clobber_p): Use the new macros.
968 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
969 LRA_NON_CLOBBERED_ALT.
971 2018-02-22 Martin Sebor <msebor@redhat.com>
973 PR tree-optimization/84480
974 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
975 to maybe_diag_stxncpy_trunc. Call it.
976 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
977 from gimple_fold_builtin_strcpy. Print inlining stack.
978 (handle_builtin_stxncpy): Print inlining stack.
979 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
981 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
984 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
985 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
986 and -fcheck-pointer-bounds are used together.
987 (indirect_thunk_prefix): New enum.
988 (indirect_thunk_need_prefix): New function.
989 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
990 "_nt" instead of "_bnd" for NOTRACK prefix.
991 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
992 (output_indirect_thunk_function): Likewise.
994 (ix86_code_end): Update output_indirect_thunk_function calls.
995 (ix86_output_indirect_branch_via_reg): Replace
996 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
997 (ix86_output_indirect_branch_via_push): Likewise.
998 (ix86_output_function_return): Likewise.
999 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1000 incompatible with -fcf-protection=branch and
1001 -fcheck-pointer-bounds.
1003 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1006 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1007 Change gcc_assert call to output_operand_lossage.
1009 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1011 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1013 2018-02-22 DJ Delorie <dj@redhat.com>
1014 Sebastian Perta <sebastian.perta@renesas.com>
1015 Oleg Endo <olegendo@gcc.gnu.org>
1017 * config/rx/rx.c (rx_rtx_costs): New function.
1018 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1020 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1022 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1024 2018-02-22 Martin Liska <mliska@suse.cz>
1027 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1028 Add "native" as a possible value.
1030 2018-02-22 Martin Liska <mliska@suse.cz>
1033 * config/i386/i386.c (ix86_option_override_internal):
1034 Add "native" as a possible value for -march and -mtune.
1036 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1039 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1040 to all type variants.
1042 PR tree-optimization/84503
1043 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1044 width as info->bitpos + info->bitsize - start.
1045 (merged_store_group::merge_overlapping): Simplify width computation.
1046 (check_no_overlap): New function.
1047 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1048 start + width and last_order of the group, fail if check_no_overlap
1050 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1051 to group if check_no_overlap fails.
1053 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1055 * config/rs6000/altivec.md: Delete contraint arguments to
1056 define_expand, define_split, and define_peephole2, and in
1057 define_insn_and_split if always unused.
1058 * config/rs6000/darwin.md: Ditto.
1059 * config/rs6000/dfp.md: Ditto.
1060 * config/rs6000/rs6000.md: Ditto.
1061 * config/rs6000/sync.md: Ditto.
1062 * config/rs6000/vector.md: Ditto.
1063 * config/rs6000/vsx.md: Ditto.
1065 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1067 * config/rs6000/altivec.md: Write output control strings as braced
1068 blocks instead of double-quoted strings.
1069 * config/rs6000/darwin.md: Ditto.
1070 * config/rs6000/rs6000.md: Ditto.
1071 * config/rs6000/vector.md: Ditto.
1072 * config/rs6000/vsx.md: Ditto.
1074 2018-02-21 Jason Merrill <jason@redhat.com>
1076 PR c++/84314 - ICE with templates and fastcall attribute.
1077 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1079 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1081 * ipa-cp.c (determine_versionability): Fix comment typos.
1083 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1086 * ipa-cp.c (determine_versionability): Do not version functions caling
1089 2018-02-21 Martin Liska <mliska@suse.cz>
1092 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1093 Add "native" as a possible value.
1094 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1095 the macro when native cpu detection is available.
1097 2018-02-21 Martin Liska <mliska@suse.cz>
1100 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1101 Add "native" as a possible value.
1102 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1103 when native cpu detection is available.
1105 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1106 Martin Sebor <msebor@redhat.com>
1108 PR tree-optimization/84478
1109 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1111 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1112 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1113 support which is conservatively correct, for 2 only stay conservative
1114 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1115 argument to the 2 argument get_range_strlen, adjust 6 arg
1116 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1118 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1119 (gimple_fold_builtin_strlen): Pass true as last argument to
1122 2018-02-20 Martin Sebor <msebor@redhat.com>
1125 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1126 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1127 (builtin_memref::builtin_memref): Factor out parts into
1128 set_base_and_offset and call it.
1130 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1133 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1134 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1135 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1136 search at the associated MODE_INT.
1138 2018-02-20 Jeff Law <law@redhat.com>
1141 PR tree-optimization/81592
1143 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1144 for range data rather than using global data.
1145 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1146 range data rather than using global data.
1147 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1148 pass it to children as needed.
1149 (struct directive::fmtresult): Similarly.
1150 (struct directive::set_width): Similarly.
1151 (struct directive::set_precision): Similarly.
1152 (format_integer, format_directive, parse_directive): Similarly.
1153 (format_none): Accept unnamed vr_values parameter.
1154 (format_percent, format_floating, format_character): Similarly.
1155 (format_string, format_plain): Similarly.
1156 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1157 the EVRP range analyzer for range data rather than using global data.
1158 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1159 gimple-ssa-evrp-analyze.h
1160 (class sprintf_dom_walker): Add after_dom_children member function.
1161 Add evrp_range_analyzer member.
1162 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1163 range analyzer as needed.
1164 (sprintf_dom_walker::after_dom_children): New member function.
1165 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1167 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1168 (evrp_range_analyzer::pop_to_marker): Likewise.
1170 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1172 PR tree-optimization/84419
1173 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1174 with the required type if its current type is compatible but
1177 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1180 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1181 after vectorization.
1183 2018-02-20 Martin Liska <mliska@suse.cz>
1186 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1187 possible values if we don't have a hint.
1189 2018-02-20 Martin Liska <mliska@suse.cz>
1193 * final.c (shorten_branches): Build align_tab array with one
1195 * opts.c (finish_options): Add alignment option limit check.
1196 (MAX_CODE_ALIGN): Likewise.
1197 (MAX_CODE_ALIGN_VALUE): Likewise.
1198 * doc/invoke.texi: Document maximum allowed option value for
1199 all -falign-* options.
1201 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1204 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1205 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1206 * var-tracking.c (emit_note_insn_var_location): Remove all references
1207 to NOTE_INSN_CALL_ARG_LOCATION.
1208 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1209 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1210 Use copy_rtx_if_shared.
1211 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1212 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1213 (dwarf2out_var_location): Remove handling of
1214 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1216 * final.c (final_scan_insn): Remove all references to
1217 NOTE_INSN_CALL_ARG_LOCATION.
1218 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1219 before dumping final insns.
1220 * except.c (emit_note_eh_region_end): Remove all references to
1221 NOTE_INSN_CALL_ARG_LOCATION.
1222 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1223 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1224 * config/arc/arc.c (hwloop_optimize): Likewise.
1225 * config/arm/arm.c (create_fix_barrier): Likewise.
1226 * config/s390/s390.c (s390_chunkify_start): Likewise.
1227 * config/sh/sh.c (find_barrier): Likewise.
1228 * config/i386/i386.c (rest_of_insert_endbranch,
1229 ix86_seh_fixup_eh_fallthru): Likewise.
1230 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1231 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1232 * config/frv/frv.c (frv_function_prologue): Likewise.
1233 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1235 (note_outside_basic_block_p): Remove all references to
1236 NOTE_INSN_CALL_ARG_LOCATION.
1237 * gengtype.c (adjust_field_rtx_def): Likewise.
1238 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1240 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1241 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1244 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1247 PR tree-optimization/84452
1248 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1249 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1252 2018-02-19 Martin Liska <mliska@suse.cz>
1255 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1257 2018-02-19 Martin Liska <mliska@suse.cz>
1258 Richard Sandiford <richard.sandiford@linaro.org>
1260 PR tree-optimization/82491
1261 * gimple-fold.c (get_base_constructor): Make earlier bail out
1264 2018-02-19 Carl Love <cel@us.ibm.com>
1266 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1267 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1269 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1270 P8V_BUILTIN_VEC_NEG.
1272 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1274 * config/rl78/rl78.md (movdf): New define expand.
1276 2018-02-19 Martin Liska <mliska@suse.cz>
1279 * doc/invoke.texi: Fix typo.
1280 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1282 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1284 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1285 handle rs6000_single_float and rs6000_double_float specially for
1288 2018-02-16 Jeff Law <law@redhat.com>
1290 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1293 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1296 * ipa-inline.c (inline_small_functions): Fix a typo.
1298 2018-02-16 Nathan Sidwell <nathan@acm.org>
1300 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1302 2018-02-16 Carl Love <cel@us.ibm.com>
1304 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1305 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1306 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1307 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1308 expansion to P8V_BUILTIN_VEC_FLOAT2.
1310 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1312 PR rtl-optimization/70023
1313 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1314 src_regno into account.
1316 2018-02-16 Carl Love <cel@us.ibm.com>
1318 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1319 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1320 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1321 * config/rs6000/rs6000.c: Remove case statements for
1322 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1323 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1324 and P9V_BUILTIN_VEC_VINSERT4B.
1325 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1326 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1327 * config/rs6000/vsx.md:
1328 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1331 2018-02-16 Carl Love <cel@us.ibm.com>
1333 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1335 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1337 * config/rs6000/rs6000-c.c: Add the definitions for
1338 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1339 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1340 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1341 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1342 definition for insert4b and define insn *insert3b_internal.
1343 * doc/extend.texi: Add documentation for vec_extract4b.
1345 2018-02-16 Nathan Sidwell <nathan@acm.org>
1347 * doc/extend.texi (Backwards Compatibility): Mention friend
1348 injection. Note for-scope is deprecated.
1349 * doc/invoke.texi (-ffriend-injection): Deprecate.
1351 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1353 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1354 that moved to I2, also allow destinations that are a paradoxical
1355 subreg (instead of a normal reg).
1357 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1360 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1363 2018-02-16 Richard Biener <rguenther@suse.de>
1365 PR tree-optimization/84037
1366 PR tree-optimization/84016
1368 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1369 Adjust vec_construct for the fact we need additional higher latency
1370 128bit inserts for AVX256 and AVX512 vector builds.
1371 (ix86_add_stmt_cost): Scale vector construction cost for
1374 2018-02-16 Richard Biener <rguenther@suse.de>
1376 PR tree-optimization/84417
1377 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1378 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1379 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1381 2018-02-16 Martin Liska <mliska@suse.cz>
1384 * internal-fn.def (ASAN_CHECK): Set proper flags.
1385 (ASAN_MARK): Likewise.
1387 2018-02-16 Julia Koval <julia.koval@intel.com>
1389 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1390 from PTA_CANNONLAKE.
1392 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1395 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1396 Use ++iter rather than iter++ for std::list iterators.
1397 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1398 defer deleting them until all nodes in the forest are processed. Do
1399 free even leaf nodes. Change to_process into auto_vec.
1402 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1403 * vec.h (vec_default_construct): Use memset instead of placement new
1404 if BROKEN_VALUE_INITIALIZATION is defined.
1405 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1406 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1409 PR rtl-optimization/83723
1410 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1411 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1412 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1414 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1416 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1418 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1420 PR rtl-optimization/81443
1421 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1422 from inner REGs to paradoxical SUBREGs.
1424 2018-02-16 Richard Biener <rguenther@suse.de>
1426 PR tree-optimization/84399
1427 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1428 For operands we can analyze at their definition make sure we can
1429 analyze them at each use as well.
1431 2018-02-16 Richard Biener <rguenther@suse.de>
1433 PR tree-optimization/84190
1434 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1435 volatile accesses if the decl isn't volatile.
1437 2018-02-15 Jason Merrill <jason@redhat.com>
1439 PR c++/84314 - ICE with templates and fastcall attribute.
1440 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1441 TYPE_CANONICAL on an existing type.
1443 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1445 PR tree-optimization/84383
1446 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1447 dstoff nor call operand_equal_p if dstbase is NULL.
1449 PR tree-optimization/84334
1450 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1451 also a CONSTANT_CLASS_P, punt.
1453 2018-02-14 Jim Wilson <jimw@sifive.com>
1455 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1456 first SMALL_OPERAND check. New local min_second_step. Move assert
1457 to where locals are set. Add TARGET_RVC support.
1458 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1460 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1462 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1464 2018-02-14 Martin Sebor <msebor@redhat.com>
1466 PR tree-optimization/83698
1467 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1468 arrays constrain the offset range to their bounds.
1469 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1470 (builtin_access::overlap): Avoid setting the size of overlap if it's
1472 (maybe_diag_overlap): Also consider arrays when deciding what values
1473 of offsets to include in diagnostics.
1475 2018-02-14 Martin Sebor <msebor@redhat.com>
1478 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1479 that correspond to the kind of a declaration.
1481 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1484 * config/pa/pa.md: Load address of PIC label using the linkage table
1485 if the label is nonlocal.
1487 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1489 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1490 warning message if user requests -maltivec=be.
1491 * doc/invoke.texi: Document deprecation of -maltivec=be.
1493 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1496 * config/rs6000/rs6000-c.c: Update definitions for
1497 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1498 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1500 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1503 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1504 add _get_ssp intrinsics. Remove argument from
1505 __builtin_ia32_rdssp[d|q].
1506 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1507 * config/i386/i386-builtin.def: Remove argument from
1508 __builtin_ia32_rdssp[d|q].
1509 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1510 ix86_expand_special_args_builtin for _rdssp[d|q].
1511 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1512 Clear register before usage.
1513 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1514 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1516 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1518 PR tree-optimization/84357
1519 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1520 operand 1 of an ARRAY_REF too.
1522 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1525 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1526 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1528 (set_of_reg): New struct.
1529 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1530 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1531 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1533 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1534 Split into bitclr, bitset, bitinvert patterns if appropriate.
1535 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1536 use rx_fuse_in_memory_bitop.
1537 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1538 to named insn, correct maximum insn length.
1540 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1543 * machmode.def: Define a complex mode for PARTIAL_INT.
1544 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1546 * doc/rtl.texi: Document CSPImode.
1547 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1549 (msp430_hard_regno_nregs_with_padding): Likewise.
1551 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1554 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1556 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1558 PR rtl-optimization/84169
1559 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1560 we generated a parallel as new i3 and we split that to new i2 and i3
1561 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1562 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1563 those to i2, not i1. Partially rewrite this scan code.
1565 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1568 * stor-layout.c (place_field): For variable length fields, adjust
1569 offset_align afterwards not just based on the field's alignment,
1570 but also on the size.
1573 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1574 of exps and logs in the use_exp2 case.
1576 2018-02-13 Jeff Law <law@redhat.com>
1578 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1581 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1584 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1588 * common.opt (gas-loc-support, gas-locview-support): New.
1589 (ginline-points, ginternal-reset-location-views): New.
1590 * doc/invoke.texi: Document them. Use @itemx where intended.
1591 (gvariable-location-views): Adjust.
1592 * target.def (reset_location_view): New.
1593 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1594 (TARGET_RESET_LOCATION_VIEW): New.
1595 * doc/tm.texi: Rebuilt.
1596 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1597 (dwarf2out_default_as_locview_support): New.
1598 (output_asm_line_debug_info): Use option variables.
1599 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1600 (output_loc_list): Likewise.
1601 (add_high_low_attributes): Check option variables.
1602 Don't output entry view attribute in strict mode.
1603 (gen_inlined_subroutine_die): Check option variables.
1604 (dwarf2out_inline_entry): Likewise.
1605 (init_sections_and_labels): Likewise.
1606 (dwarf2out_early_finish): Likewise.
1607 (maybe_reset_location_view): New, from...
1608 (dwarf2out_var_location): ... here. Call it.
1609 * debug.h (dwarf2out_default_as_loc_support): Declare.
1610 (dwarf2out_default_as_locview_support): Declare.
1611 * hooks.c (hook_int_rtx_insn_0): New.
1612 * hooks.h (hook_int_rtx_insn_0): Declare.
1613 * toplev.c (process_options): Take -gas-loc-support and
1614 -gas-locview-support from dwarf2out. Enable
1615 -gvariable-location-views by default only with locview
1616 assembler support. Enable -ginternal-reset-location-views by
1617 default only if the target defines the corresponding hook.
1618 Enable -ginline-points by default if location views are
1619 enabled; force it disabled if statement frontiers are
1621 * tree-inline.c (expand_call_inline): Check option variables.
1622 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1624 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1626 PR tree-optimization/84321
1627 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1628 handling. Also check whether the anti-range contains any values
1629 that satisfy the mask; switch to a VR_RANGE if not.
1631 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1634 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1636 2018-02-13 Martin Jambor <mjambor@suse.cz>
1639 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1640 of call statements, also set location of a load to a temporary.
1642 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1644 * config/rl78/rl78.c (add_vector_labels): New function.
1645 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1646 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1647 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1648 which checks that no arguments are passed.
1649 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1650 * doc/extend.texi: Documentation for the new attribute.
1652 2018-02-13 Andreas Schwab <schwab@suse.de>
1654 * config/riscv/linux.h (CPP_SPEC): Define.
1656 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1659 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1660 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1661 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1662 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1663 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1664 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1665 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1666 temporarily for AES and PCLMUL builtins.
1668 PR tree-optimization/84339
1669 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1670 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1674 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1675 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1676 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1678 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1680 * omp-simd-clone.h: New file.
1681 * omp-simd-clone.c: Include omp-simd-clone.h.
1682 (expand_simd_clones): No longer static.
1683 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1684 cgraph.h and omp-simd-clone.h.
1685 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1686 (vect_recog_widen_shift_pattern): Formatting fix.
1687 (vect_pattern_recog_1): Don't check optab for calls.
1690 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1691 operands[2] into a REG before using gen_lowpart on it.
1693 2018-02-12 Jeff Law <law@redhat.com>
1696 * config/sh/sh.c (find_barrier): Consider a sibling call
1699 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1700 successfully back substituting a reg.
1702 2018-02-12 Richard Biener <rguenther@suse.de>
1704 PR tree-optimization/84037
1705 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1706 parameter, move visited init to caller.
1707 (vect_slp_analyze_operations): Separate cost from validity
1708 check, initialize visited once for all instances.
1709 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1711 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1713 (vect_model_promotion_demotion_cost): Likewise.
1714 (vectorizable_bswap): Guard cost modeling with !slp_node
1715 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1717 (vectorizable_call): Likewise.
1718 (vectorizable_conversion): Likewise.
1719 (vectorizable_assignment): Likewise.
1720 (vectorizable_shift): Likewise.
1721 (vectorizable_operation): Likewise.
1722 (vectorizable_store): Likewise.
1723 (vectorizable_load): Likewise.
1724 (vectorizable_condition): Likewise.
1725 (vectorizable_comparison): Likewise.
1727 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1730 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1731 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1732 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1734 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1737 * params.def (inline-min-speedup): Increase from 8 to 15.
1738 (max-inline-insns-auto): Decrease from 40 to 30.
1739 * ipa-split.c (consider_split): Add some buffer for function to
1740 be considered inlining candidate.
1741 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1744 2018-02-12 Richard Biener <rguenther@suse.de>
1746 PR tree-optimization/84037
1747 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1748 matched stmts if we cannot swap the non-matched ones.
1750 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1752 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1753 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1754 _mm_maskz_scalef_round_ss): New intrinsics.
1755 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1756 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1757 __builtin_ia32_scalefss_round): Remove.
1758 (__builtin_ia32_scalefsd_mask_round,
1759 __builtin_ia32_scalefss_mask_round): New intrinsics.
1760 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1761 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1762 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1763 "<round_constraint>")): Changed to ...
1764 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1765 "<round_scalar_constraint>")): ... this.
1766 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1767 %0, %1, %2<round_op3>}"): Changed to ...
1768 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1769 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1770 %2<round_scalar_mask_op3>}"): ... this.
1771 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1773 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1775 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1776 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1777 (_mm_maskz_sqrt_round_ss): New intrinsics.
1778 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1779 (__builtin_ia32_sqrtsd_mask_round)
1780 (__builtin_ia32_sqrtss_mask_round): New builtins.
1781 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1782 (__builtin_ia32_sqrtss_round): Remove.
1783 (__builtin_ia32_sqrtsd_mask_round)
1784 (__builtin_ia32_sqrtss_mask_round): New builtins.
1785 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1786 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1787 ((match_operand:VF_128 1 "vector_operand"
1788 "xBm,<round_constraint>")): Changed to ...
1789 ((match_operand:VF_128 1 "vector_operand"
1790 "xBm,<round_scalar_constraint>")): ... this.
1791 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1792 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1793 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1794 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1795 %<iptr>1<round_scalar_mask_op3>}): ... this.
1796 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1797 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1799 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1802 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1803 Cast vec_cmpeq result to correct type.
1804 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1805 Cast vec_cmpgt result to correct type.
1807 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1809 * final.c (final_scan_insn_1): Renamed from...
1810 (final_scan_insn): ... this. New wrapper, to recover
1811 seen from the outermost call in recursive ones.
1812 * config/sparc/sparc.c (output_return): Drop seen from call.
1813 (output_sibcall): Likewise.
1814 * config/visium/visium.c (output_branch): Likewise.
1816 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1818 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1821 2018-02-10 Alan Modra <amodra@gmail.com>
1824 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1825 Specify LR as an input.
1827 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1830 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1831 remove_member_access_dummy_vars): New functions.
1832 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1833 lower_omp_1, execute_lower_omp): Use them.
1835 PR rtl-optimization/84308
1836 * shrink-wrap.c (spread_components): Release todo vector.
1838 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1840 PR rtl-optimization/57193
1841 * ira-color.c (struct allocno_color_data): Add member
1842 conflict_allocno_hard_prefs.
1843 (update_conflict_allocno_hard_prefs): New.
1844 (bucket_allocno_compare_func): Add a preference based on
1845 conflict_allocno_hard_prefs.
1846 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1847 (color_allocnos): Remove a dead code. Initiate
1848 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1850 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1853 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1854 constraint from =wa to wa. Avoid a subreg on the output operand,
1855 instead use a pseudo and subreg it in a move.
1856 (p9_xxbrd_<mode>): Changed to ...
1857 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1858 (p9_xxbrd_v2df): New expander.
1859 (p9_xxbrw_<mode>): Changed to ...
1860 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1861 (p9_xxbrw_v4sf): New expander.
1863 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1865 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1867 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1870 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1871 multiply in 32-bit mode.
1872 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1873 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1876 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1878 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1879 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1880 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1881 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1883 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1886 * dwarf2out.c (is_trivial_indirect_ref): New function.
1887 (dwarf2out_late_global_decl): Do not generate a location
1888 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1889 and that are not defined in the current unit.
1891 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1893 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1894 instead of a libcall for UNORDERED.
1896 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1899 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1900 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1902 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1905 * config/s390/s390.c (s390_set_current_function): Invoke
1906 s390_indirect_branch_settings also if fndecl didn't change.
1908 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1910 * config/rs6000/rs6000.md (blockage): Set length to zero.
1912 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1914 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1916 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1919 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1920 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1924 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1925 PARALLEL incoming that failed vt_get_decl_and_offset check.
1928 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1929 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1931 (get_variable_section): For decls in named .bss* sections pass true as
1932 second argument to bss_initializer_p.
1934 2018-02-09 Marek Polacek <polacek@redhat.com>
1935 Jakub Jelinek <jakub@redhat.com>
1938 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1939 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1940 Sync some changes from cxx_fold_indirect_ref.
1942 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1944 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1946 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1947 (BLOCK_INLINE_ENTRY_LABEL): New.
1948 (dwarf2out_var_location): Disregard inline entry markers.
1949 (inline_entry_data): New struct.
1950 (inline_entry_data_hasher): New hashtable type.
1951 (inline_entry_data_hasher::hash): New.
1952 (inline_entry_data_hasher::equal): New.
1953 (inline_entry_data_table): New variable.
1954 (add_high_low_attributes): Add DW_AT_entry_pc and
1955 DW_AT_GNU_entry_view attributes if a pending entry is found
1956 in inline_entry_data_table. Add old entry_pc attribute only
1957 if debug nonbinding markers are disabled.
1958 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1959 markers are enabled.
1960 (block_within_block_p, dwarf2out_inline_entry): New.
1961 (dwarf2out_finish): Check that no entries remained in
1962 inline_entry_data_table.
1963 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1964 (final_scan_insn, notice_source_line): Likewise.
1965 (rest_of_clean_state): Skip inline entry markers.
1966 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1968 * gimple.c (gimple_build_debug_inline_entry): New.
1969 * gimple.h (enum gimple_debug_subcode): Add
1970 GIMPLE_DEBUG_INLINE_ENTRY.
1971 (gimple_build_debug_inline_entry): Declare.
1972 (gimple_debug_inline_entry_p): New.
1973 (gimple_debug_nonbind_marker_p): Adjust.
1974 * insn-notes.def (INLINE_ENTRY): New.
1975 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1976 inline entry marker notes.
1977 (print_insn): Likewise.
1978 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1979 (INSN_DEBUG_MARKER_KIND): Likewise.
1980 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1981 * tree-inline.c (expand_call_inline): Build and insert
1982 debug_inline_entry stmt.
1983 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1984 inline entry blocks early, if nonbind markers are enabled.
1985 (dump_scope_block): Dump fragment info.
1986 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1987 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1988 (gimple_build_debug_inline_entry): New.
1989 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1990 Enable/disable inline entry points too.
1991 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1992 (DEBUG_INSN): Describe inline entry markers.
1994 * common.opt (gvariable-location-views): New.
1995 (gvariable-location-views=incompat5): New.
1996 * config.in: Rebuilt.
1997 * configure: Rebuilt.
1998 * configure.ac: Test assembler for view support.
1999 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2000 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2001 * dwarf2out.c (var_loc_view): New typedef.
2002 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2003 (dwarf2out_locviews_in_attribute): New.
2004 (dwarf2out_locviews_in_loclist): New.
2005 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2006 (enum dw_line_info_opcode): Add LI_adv_address.
2007 (struct dw_line_info_table): Add view.
2008 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2009 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2010 (zero_view_p): New variable.
2011 (ZERO_VIEW_P): New macro.
2012 (output_asm_line_debug_info): New.
2013 (struct var_loc_node): Add view.
2014 (add_AT_view_list, AT_loc_list): New.
2015 (add_var_loc_to_decl): Add view param. Test it against last.
2016 (new_loc_list): Add view params. Record them.
2017 (AT_loc_list_ptr): Handle loc and view lists.
2018 (view_list_to_loc_list_val_node): New.
2019 (print_dw_val): Handle dw_val_class_view_list.
2020 (size_of_die): Likewise.
2021 (value_format): Likewise.
2022 (loc_list_has_views): New.
2023 (gen_llsym): Set vl_symbol too.
2024 (maybe_gen_llsym, skip_loc_list_entry): New.
2025 (dwarf2out_maybe_output_loclist_view_pair): New.
2026 (output_loc_list): Output view list or entries too.
2027 (output_view_list_offset): New.
2028 (output_die): Handle dw_val_class_view_list.
2029 (output_dwarf_version): New.
2030 (output_compilation_unit_header): Use it.
2031 (output_skeleton_debug_sections): Likewise.
2032 (output_rnglists, output_line_info): Likewise.
2033 (output_pubnames, output_aranges): Update version comments.
2034 (output_one_line_info_table): Output view numbers in asm comments.
2035 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2036 Call maybe_gen_llsym.
2037 (loc_list_from_tree_1): Adjust.
2038 (add_AT_location_description): Create view list attribute if
2039 needed, check it's absent otherwise.
2040 (convert_cfa_to_fb_loc_list): Adjust.
2041 (maybe_emit_file): Call output_asm_line_debug_info for test.
2042 (dwarf2out_var_location): Reset views as needed. Precompute
2043 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2044 attribute. Set view.
2045 (new_line_info_table): Reset next view.
2046 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2047 (dwarf2out_source_line): Likewise. Output view resets and labels to
2048 the assembler, or select appropriate line info opcodes.
2049 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2050 (optimize_string_length): Catch it. Adjust.
2051 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2052 dw_val_class_view_list, and remove it if no longer needed.
2053 (hash_loc_list): Hash view numbers.
2054 (loc_list_hasher::equal): Compare them.
2055 (optimize_location_lists): Check whether a view list symbol is
2056 needed, and whether the locview attribute is present, and
2057 whether they match. Remove the locview attribute if no longer
2059 (index_location_lists): Call skip_loc_list_entry for test.
2060 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2061 Use output_dwarf_version.
2062 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2063 (struct dw_val_node): Add val_view_list.
2064 * final.c (SEEN_NEXT_VIEW): New.
2065 (set_next_view_needed): New.
2066 (clear_next_view_needed): New.
2067 (maybe_output_next_view): New.
2068 (final_start_function): Rename to...
2069 (final_start_function_1): ... this. Take pointer to FIRST,
2070 add SEEN parameter. Emit param bindings in the initial view.
2071 (final_start_function): Reintroduce SEEN-less interface.
2072 (final): Rename to...
2073 (final_1): ... this. Take SEEN parameter. Output final pending
2074 next view at the end.
2075 (final): Reintroduce seen-less interface.
2076 (final_scan_insn): Output pending next view before switching
2077 sections or ending a block. Mark the next view as needed when
2078 outputting variable locations. Notify debug backend of section
2079 changes, and of location view changes.
2080 (rest_of_handle_final): Adjust.
2081 * toplev.c (process_options): Autodetect value for debug variable
2082 location views option. Warn on incompat5 without -gdwarf-5.
2083 * doc/invoke.texi (gvariable-location-views): New.
2084 (gvariable-location-views=incompat5): New.
2085 (gno-variable-location-views): New.
2087 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2089 PR tree-optimization/84136
2090 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2091 that the result of find_edge is non-NULL.
2093 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2096 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2097 storing integer register in SImode. Fix cost of 256 and 512
2098 byte aligned SSE register store.
2100 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2102 * config/i386/i386.c (ix86_multiplication_cost): Fix
2103 multiplication cost for TARGET_AVX512DQ.
2105 2018-02-08 Marek Polacek <polacek@redhat.com>
2107 PR tree-optimization/84238
2108 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2111 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2113 PR tree-optimization/84265
2114 * tree-vect-stmts.c (vectorizable_store): Don't treat
2115 VMAT_CONTIGUOUS accesses as grouped.
2116 (vectorizable_load): Likewise.
2118 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2120 PR tree-optimization/81635
2121 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2122 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2123 (test_round_for_mask): New functions.
2124 (wide_int_cc_tests): Call test_round_for_mask.
2125 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2126 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2127 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2128 range returned by get_range_info.
2130 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2133 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2134 * symtab.c: Include builtins.h
2135 (symtab_node::output_to_lto_symbol_table_p): Move here
2136 from lto-streamer-out.c:output_symbol_p.
2137 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2138 (output_symbol_p): Move all logic to symtab.c
2139 (produce_symtab): Update.
2141 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2143 * config/s390/s390-opts.h (enum indirect_branch): Define.
2144 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2145 (s390_indirect_branch_via_thunk)
2146 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2147 (enum s390_indirect_branch_type): Define.
2148 * config/s390/s390.c (struct s390_frame_layout, struct
2149 machine_function): Remove.
2150 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2151 (indirect_branch_table_label_no, indirect_branch_table_name):
2153 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2154 (enum s390_indirect_branch_option): Define.
2155 (s390_return_addr_from_memory): New function.
2156 (s390_handle_string_attribute): New function.
2157 (s390_attribute_table): Add new attribute handler.
2158 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2159 (s390_indirect_branch_via_thunk): New function.
2160 (s390_indirect_branch_via_inline_thunk): New function.
2161 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2162 sibling call optimization for non z10 compiles.
2163 (s390_emit_call): Force indirect branch target to be a single
2164 register. Add r1 clobber for non-z10 compiles.
2165 (s390_emit_epilogue): Emit return jump via return_use expander.
2166 (s390_reorg): Handle JUMP_INSNs as execute targets.
2167 (s390_option_override_internal): Perform validity checks for the
2168 new command line options.
2169 (s390_indirect_branch_attrvalue): New function.
2170 (s390_indirect_branch_settings): New function.
2171 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2172 (s390_output_indirect_thunk_function): New function.
2173 (s390_code_end): Implement target hook.
2174 (s390_case_values_threshold): Implement target hook.
2175 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2177 * config/s390/s390.h (struct s390_frame_layout)
2178 (struct machine_function): Move here from s390.c.
2179 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2180 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2181 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2182 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2183 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2184 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2185 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2186 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2187 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2188 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2189 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2190 (mnemonic attribute): Add values which aren't recognized
2192 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2193 pattern for branch conversion. Fix mnemonic attribute.
2194 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2195 indirect branch via thunk if requested.
2196 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2197 ("*indirect_jump"): Disable for branch conversion using out of
2199 ("indirect_jump_via_thunk<mode>_z10")
2200 ("indirect_jump_via_thunk<mode>")
2201 ("indirect_jump_via_inlinethunk<mode>_z10")
2202 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2203 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2204 ("casesi_jump_via_inlinethunk<mode>_z10")
2205 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2206 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2207 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2208 ("*indirect2_jump"): Disable for branch conversion.
2209 ("casesi_jump"): Turn into expander and expand patterns for branch
2211 ("return_use"): New expander.
2212 ("*return"): Emit return via thunk and rename it to ...
2213 ("*return<mode>"): ... this one.
2214 * config/s390/s390.opt: Add new options and and enum for the
2217 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2219 * lra-constraints.c (match_reload): Unconditionally use
2220 gen_lowpart_SUBREG, rather than selecting between that
2221 and equivalent gen_rtx_SUBREG code.
2223 2018-02-08 Richard Biener <rguenther@suse.de>
2225 PR tree-optimization/84233
2226 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2227 changed flag instead of boguously re-using phi_inserted.
2229 2018-02-08 Martin Jambor <mjambor@suse.cz>
2231 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2232 static local variables.
2234 2018-02-08 Richard Biener <rguenther@suse.de>
2236 PR tree-optimization/84278
2237 * tree-vect-stmts.c (vectorizable_store): When looking for
2238 smaller vector types to perform grouped strided loads/stores
2239 make sure the mode is supported by the target.
2240 (vectorizable_load): Likewise.
2242 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2244 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2245 Increase LDP/STP opportunities by adding adjacent callee-saves.
2247 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2249 PR rtl-optimization/84068
2250 PR rtl-optimization/83459
2251 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2253 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2255 PR tree-optimization/84224
2256 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2257 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2260 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2263 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2264 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2265 count, remove one USE.
2267 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2269 * doc/install.texi (Configuration): Document the
2270 --with-long-double-format={ibm,ieee} PowerPC configuration
2274 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2275 Convert from define_expand to be define_insn_and_split. Rework
2276 float/double/_Float128 conversions to QI/HI/SImode to work with
2277 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2278 conversions to QI/HImode types did a store and then a load to
2279 truncate the value. For conversions to VSX registers, don't split
2280 the insn, instead emit the code directly. Use the code iterator
2281 any_fix to combine signed and unsigned conversions.
2282 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2283 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2284 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2285 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2286 (fix_<mode>di2_hw): Likewise.
2287 (fixuns_<mode>di2_hw): Likewise.
2288 (fix_<mode>si2_hw): Likewise.
2289 (fixuns_<mode>si2_hw): Likewise.
2290 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2291 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2292 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2293 fix<uns>_trunc<SFDF:mode>si2_p8.
2294 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2296 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2297 (fix<uns>_<mode>_mem): Likewise.
2298 (fctiw<u>z_<mode>_mem): Likewise.
2299 (fix<uns>_<mode>_mem): Likewise.
2300 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2301 the register allocator from doing a direct move to the GPRs to do
2302 a store, and instead use the ISA 3.0 store byte/half-word from
2303 vector register instruction. For IEEE 128-bit floating point,
2304 also optimize stores of 32-bit ints.
2305 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2307 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2309 * genextract.c (push_pathstr_operand): New function to support
2311 (walk_rtx): Call push_pathstr_operand.
2312 (print_path): Support [a-zA-Z].
2314 2018-02-07 Richard Biener <rguenther@suse.de>
2316 PR tree-optimization/84037
2317 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2318 (cse_and_gimplify_to_preheader): Declare.
2319 (vect_get_place_in_interleaving_chain): Likewise.
2320 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2322 (_loop_vec_info::~_loop_vec_info): Delete it.
2323 (cse_and_gimplify_to_preheader): New function.
2324 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2325 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2326 (vectorizable_load): Likewise. For grouped stores always base
2327 the IV on the first element.
2328 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2329 condition before gimplifying.
2331 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2333 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2334 *DIV_EXPR and *MOD_EXPR.
2336 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2339 * config/i386/i386.c (ix86_option_override_internal): Mask out
2340 the CF_SET bit when checking -fcf-protection.
2342 2018-02-07 Tom de Vries <tom@codesourcery.com>
2345 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2348 2018-02-07 Richard Biener <rguenther@suse.de>
2350 PR tree-optimization/84204
2351 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2354 PR tree-optimization/84205
2355 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2356 special-case isl_ast_op_zdiv_r.
2358 PR tree-optimization/84223
2359 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2360 Only add conditions from within the region.
2361 (gather_bbs::after_dom_children): Adjust.
2363 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2366 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2367 * config/avr/avr.md: Only post-reload split REG-REG moves if
2368 either register is GENERAL_REG_P.
2370 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2372 PR tree-optimization/84235
2373 * tree-ssa-scopedtables.c
2374 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2375 if the subtraction is performed in floating point type where NaNs are
2376 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2377 build 1. Formatting fix.
2379 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2382 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2383 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2384 and skip it regardless of bb boundaries. Use CALL_P macro,
2385 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2386 unnecessarily, formatting fix.
2388 2018-02-06 Michael Collison <michael.collison@arm.com>
2390 * config/arm/thumb2.md:
2391 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2392 (*thumb_mov_notscc): Ditto.
2394 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2397 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2398 unsigned_fix, not "s".
2400 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2402 * configure.ac (gcc_fn_eh_frame_ro): New function.
2403 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2404 correct .eh_frame permissions.
2405 * configure: Regenerate.
2407 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2409 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2412 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2414 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2415 Display warning message for -mno-speculate-indirect-jumps.
2417 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2419 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2421 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2423 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2425 PR tree-optimization/84225
2426 * tree-eh.c (find_trapping_overflow): Only call
2427 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2429 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2432 * config/i386/i386.c: Reimplement the check of possible options
2433 -mibt/-mshstk conbination. Change error messages.
2434 * doc/invoke.texi: Fix a typo: remove extra '='.
2436 2018-02-06 Marek Polacek <polacek@redhat.com>
2438 PR tree-optimization/84228
2439 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2441 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2444 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2445 emitted arch directives.
2446 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2447 __ARM_FEATURE_COPROC before changing architectures.
2449 2018-02-06 Richard Biener <rguenther@suse.de>
2451 * config/i386/i386.c (print_reg): Fix typo.
2452 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2454 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2456 * configure: Regenerate.
2458 2018-02-05 Martin Sebor <msebor@redhat.com>
2460 PR tree-optimization/83369
2461 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2464 2018-02-05 Martin Liska <mliska@suse.cz>
2466 * doc/invoke.texi: Cherry-pick upstream r323995.
2468 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2470 * ira.c (ira_init_register_move_cost): Adjust comment.
2472 2018-02-05 Martin Liska <mliska@suse.cz>
2474 PR gcov-profile/84137
2475 * doc/gcov.texi: Fix typo in documentation.
2477 2018-02-05 Martin Liska <mliska@suse.cz>
2479 PR gcov-profile/83879
2480 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2481 using dlopen functionality.
2483 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2485 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2486 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2487 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2488 _mm_maskz_range_round_ss): New intrinsics.
2489 (__builtin_ia32_rangesd128_round)
2490 (__builtin_ia32_rangess128_round): Remove.
2491 (__builtin_ia32_rangesd128_mask_round,
2492 __builtin_ia32_rangess128_mask_round): New builtins.
2493 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2494 __builtin_ia32_rangess128_round): Remove.
2495 (__builtin_ia32_rangesd128_mask_round,
2496 __builtin_ia32_rangess128_mask_round): New builtins.
2497 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2498 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2499 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2500 "<round_saeonly_constraint>")): Changed to ...
2501 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2502 "<round_saeonly_scalar_constraint>")): ... this.
2503 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2504 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2505 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2506 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2507 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2509 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2511 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2513 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2514 Remove all values except native, 8540 and 8548.
2516 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2518 * config/i386/i386.c (ix86_output_function_return): Pass
2519 INVALID_REGNUM, instead of -1, as invalid register number to
2520 indirect_thunk_name and output_indirect_thunk.
2522 2018-02-02 Julia Koval <julia.koval@intel.com>
2524 * config.gcc: Add -march=icelake.
2525 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2526 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2527 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2528 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2529 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2530 (processor_target_table): Add icelake.
2531 (ix86_option_override_internal): Handle new PTAs.
2532 (get_builtin_code_for_version): Handle icelake.
2533 (M_INTEL_COREI7_ICELAKE): New.
2534 (fold_builtin_cpu): Handle icelake.
2535 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2536 * doc/invoke.texi: Add -march=icelake.
2538 2018-02-02 Julia Koval <julia.koval@intel.com>
2540 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2541 to wide_int_bitmask.
2542 * wide-int-bitmask.h: New.
2544 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2547 * config/i386/i386.md: Replace Pmode with word_mode in
2548 builtin_setjmp_setup and builtin_longjmp to support x32.
2550 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2554 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2556 (rs6000_supported_cpu_names): New static variable.
2557 (linux_cpu_translation_table): Likewise.
2558 (elf_platform) <cpu>: Define new static variable and use it.
2559 Translate kernel AT_PLATFORM name to canonical name if needed.
2560 Error if platform name is unknown.
2562 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2565 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2567 2018-02-01 Jeff Law <law@redhat.com>
2570 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2571 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2572 the scratch if RELEASE_VIA_POP is false.
2573 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2574 If we have to save a temporary register, decrement SIZE appropriately.
2575 Pass new arguments to release_scratch_register_on_entry.
2576 (ix86_adjust_stack_and_probe): Likewise.
2577 (ix86_emit_probe_stack_range): Pass new arguments to
2578 release_scratch_register_on_entry.
2580 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2582 PR rtl-optimization/84157
2583 * combine.c (change_zero_ext): Use REG_P predicate in
2584 front of HARD_REGISTER_P predicate.
2586 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2588 * config/avr/avr.c (avr_option_override): Move disabling of
2589 -fdelete-null-pointer-checks to...
2590 * common/config/avr/avr-common.c (avr_option_optimization_table):
2593 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2595 PR tree-optimization/81635
2596 * tree-data-ref.c (split_constant_offset_1): For types that
2597 wrap on overflow, try to use range info to prove that wrapping
2600 2018-02-01 Renlin Li <renlin.li@arm.com>
2603 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2605 (aarch64_register_move_cost): Likewise.
2606 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2608 (REG_CLASS_NAMES): Likewise.
2609 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2610 TAILCALL_ADDR_REGS. Remove IP registers.
2611 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2613 2018-02-01 Richard Biener <rguenther@suse.de>
2615 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2616 for specifying RPO order and allow NULL for that.
2617 * domwalk.c (dom_walker::dom_walker): Likewise.
2618 (dom_walker::walk): Handle NULL RPO order.
2619 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2621 (rewrite_update_dom_walker): Likewise.
2622 (mark_def_dom_walker): Likewise.
2624 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2626 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2627 (aarch64_maybe_expand_sve_subreg_move): Declare.
2628 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2629 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2631 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2632 that are semantically a reverse operation.
2633 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2634 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2635 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2637 (aarch64_can_change_mode_class): For big-endian, forbid changes
2638 between two SVE modes if they have different element sizes.
2640 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2642 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2643 the TImode handling for big-endian targets.
2645 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2647 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2648 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2650 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2651 Remove BSWAP handing for big-endian targets and use the form of
2652 LD1RQ appropariate for the mode.
2654 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2656 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2657 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2660 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2663 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2664 check for operands that need to go through aarch64_sve_reload_be.
2666 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2668 PR tree-optimization/81661
2669 PR tree-optimization/84117
2670 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2671 * tree-eh.c: Include gimplify.h.
2672 (find_trapping_overflow, replace_trapping_overflow,
2673 rewrite_to_non_trapping_overflow): New functions.
2674 * tree-vect-loop.c: Include tree-eh.h.
2675 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2676 * tree-data-ref.c: Include tree-eh.h.
2677 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2679 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2681 PR rtl-optimization/84123
2682 * combine.c (change_zero_ext): Check if hard register satisfies
2683 can_change_dest_mode before calling gen_lowpart_SUBREG.
2685 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2688 * ira.c (ira_init_register_move_cost): Remove assert.
2690 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2692 PR rtl-optimization/84071
2693 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2694 * doc/tm.texi: Regenerate.
2696 2018-01-31 Richard Biener <rguenther@suse.de>
2698 PR tree-optimization/84132
2699 * tree-data-ref.c (analyze_miv_subscript): Properly
2700 check whether evolution_function_is_affine_multivariate_p
2701 before calling gcd_of_steps_may_divide_p.
2703 2018-01-31 Julia Koval <julia.koval@intel.com>
2706 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2707 * config/i386/i386.md (rdpid_rex64) New.
2708 (rdpid): Make 32bit only.
2710 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2713 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2714 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2716 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2719 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2721 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2723 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2725 PR rtl-optimization/84071
2726 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2727 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2729 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2731 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2732 (arc_attribute_table): Add 'aux' attribute.
2733 (arc_in_small_data_p): Consider aux like variables.
2734 (arc_is_aux_reg_p): New function.
2735 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2736 (arc_get_aux_arg): New function.
2737 (prepare_move_operands): Handle aux-register access.
2738 (arc_handle_aux_attribute): New function.
2739 * doc/extend.texi (ARC Variable attributes): Add subsection.
2741 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2743 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2744 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2745 (arc_attribute_table): Add 'uncached' attribute.
2746 (arc_print_operand): Print '.di' flag for uncached memory
2748 (arc_in_small_data_p): Do not consider for small data the uncached
2750 (arc_is_uncached_mem_p): New function.
2751 * config/arc/predicates.md (compact_store_memory_operand): Check
2752 for uncached memory accesses.
2753 (nonvol_nonimm_operand): Likewise.
2754 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2756 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2759 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2760 falign-loops=): Add Optimization flag.
2762 2018-01-30 Jeff Law <law@redhat.com>
2765 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2766 INT_REGISTERS_SAVED. Check it prior to calling
2767 get_scratch_register_on_entry.
2768 (ix86_adjust_stack_and_probe): Similarly.
2769 (ix86_emit_probe_stack_range): Similarly.
2770 (ix86_expand_prologue): Corresponding changes.
2772 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2775 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2776 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2778 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2781 * lra-constraints.c (curr_insn_transform): Process AND in the
2784 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2786 PR rtl-optimization/83986
2787 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2788 dependence against last_pending_memory_flush in addition to
2791 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2793 PR tree-optimization/81611
2794 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2797 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2800 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2803 2018-01-30 Richard Biener <rguenther@suse.de>
2804 Jakub Jelinek <jakub@redhat.com>
2806 PR tree-optimization/84111
2807 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2808 inner loops added during recursion, as they don't have up-to-date
2811 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2814 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2815 (can_inline_edge_by_limits_p): ... here.
2816 (can_early_inline_edge_p, check_callers,
2817 update_caller_keys, update_callee_keys, recursive_inlining,
2818 add_new_edges_to_heap, speculation_useful_p,
2819 inline_small_functions,
2820 inline_small_functions, flatten_function,
2821 inline_to_all_callers_1): Update.
2823 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2825 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2828 2018-01-30 Richard Biener <rguenther@suse.de>
2830 PR tree-optimization/83008
2831 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2832 invariant and constant vector uses in stmts when they need
2835 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2838 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2839 * configure: Regenerate.
2841 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2843 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2845 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2846 Use gen_rtx_REG rather than gen_lowpart.
2848 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2850 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2851 rather than 0 when creating partial subregs.
2853 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2855 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2858 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2861 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2862 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2863 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2864 flags. This restores the settings used before the 2017-07-24.
2865 Turning off pre increment/decrement/modify allows IVOPTS to
2866 optimize DF/SF loops where the index is an int.
2868 2018-01-29 Richard Biener <rguenther@suse.de>
2869 Kelvin Nilsen <kelvin@gcc.gnu.org>
2872 * tree-vect-stmts.c (vectorizable_call): Don't call
2873 targetm.vectorize_builtin_md_vectorized_function if callee is
2876 2018-01-22 Carl Love <cel@us.ibm.com>
2878 * doc/extend.tex: Fix typo in second arg in
2879 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2881 2018-01-29 Richard Biener <rguenther@suse.de>
2883 PR tree-optimization/84086
2884 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2885 (flush_ssaname_freelist): When SSA names were released reset
2886 the SCEV hash table.
2888 2018-01-29 Richard Biener <rguenther@suse.de>
2890 PR tree-optimization/84057
2891 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2892 removed paths when removing edges.
2894 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2896 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2897 -mfunction-return=@var{choice}.
2899 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2902 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2904 (layout::print_source_line): Likewise.
2905 (test_get_line_width_without_trailing_whitespace): Add test cases.
2907 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2910 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2913 2018-01-26 Jim Wilson <jimw@sifive.com>
2915 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2917 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2920 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2922 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2923 and CMP + SUB-immediate -> SUBS.
2925 2018-01-26 Martin Sebor <msebor@redhat.com>
2927 PR tree-optimization/83896
2928 * tree-ssa-strlen.c (get_string_len): Rename...
2929 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2930 Avoid assuming length is constant.
2931 (handle_char_store): Use HOST_WIDE_INT for string length.
2933 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2936 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2937 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2939 2018-01-26 Richard Biener <rguenther@suse.de>
2941 PR rtl-optimization/84003
2942 * dse.c (record_store): Only record redundant stores when
2943 the earlier store aliases at least all accesses the later one does.
2945 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2947 PR rtl-optimization/83985
2948 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2949 REG_CFA_RESTORE insns.
2950 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2951 REG_CFA_RESTORE insns here.
2954 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2955 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2957 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2959 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2960 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2961 (arc_init): Likewise.
2962 (arc_override_options): Likewise.
2963 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2965 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2967 * config/arc/arc.h (TARGET_DBNZ): Define.
2968 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2969 properly set the tune attribute.
2970 (dbnz): Use TARGET_DBNZ guard.
2971 * config/arc/arc.opt (mtune): Add core3 option.
2973 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2975 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2976 recognize new pic like addresses.
2977 (arc_delegitimize_address): Clean up.
2979 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2981 * config/arc/arc-arches.def: Option mrf16 valid for all
2983 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2984 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2985 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2986 * config/arc/arc-tables.opt: Regenerate.
2987 * config/arc/arc.c (arc_conditional_register_usage): Handle
2988 reduced register file case.
2989 (arc_file_start): Set must have build attributes.
2990 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2992 * config/arc/arc.opt (mrf16): Add new option.
2993 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2994 * config/arc/genmultilib.awk: Handle new mrf16 option.
2995 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2996 * config/arc/t-multilib: Regenerate.
2997 * doc/invoke.texi (ARC Options): Document mrf16 option.
2999 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3001 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3002 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3003 (arc_attribute_table): Add 'secure_call' attribute.
3004 (arc_print_operand): Print secure call operand.
3005 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3007 (arc_is_secure_call_p): New function. * config/arc/arc.md
3008 (call_i): Add support for sjli instruction.
3009 (call_value_i): Likewise.
3010 * config/arc/constraints.md (Csc): New constraint.
3012 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3013 John Eric Martin <John.Martin@emmicro-us.com>
3015 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3016 * config/arc/arc.c (_arc_jli_section): New struct.
3017 (arc_jli_section): New type.
3018 (rc_jli_sections): New static variable.
3019 (arc_handle_jli_attribute): New function.
3020 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3021 (arc_file_end): New function.
3022 (TARGET_ASM_FILE_END): Define.
3023 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3024 (arc_add_jli_section): New function.
3025 (jli_call_scan): Likewise.
3026 (arc_reorg): Call jli_call_scan.
3027 (arc_output_addsi): Remove 'S' from printing asm operand.
3028 (arc_is_jli_call_p): New function.
3029 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3031 (movhi_insn): Likewise.
3032 (movsi_insn): Likewise.
3033 (movsi_set_cc_insn): Likewise.
3034 (loadqi_update): Likewise.
3035 (load_zeroextendqisi_update): Likewise.
3036 (load_signextendqisi_update): Likewise.
3037 (loadhi_update): Likewise.
3038 (load_zeroextendhisi_update): Likewise.
3039 (load_signextendhisi_update): Likewise.
3040 (loadsi_update): Likewise.
3041 (loadsf_update): Likewise.
3042 (movsicc_insn): Likewise.
3043 (bset_insn): Likewise.
3044 (bxor_insn): Likewise.
3045 (bclr_insn): Likewise.
3046 (bmsk_insn): Likewise.
3047 (bicsi3_insn): Likewise.
3048 (cmpsi_cc_c_insn): Likewise.
3049 (movsi_ne): Likewise.
3050 (movsi_cond_exec): Likewise.
3051 (clrsbsi2): Likewise.
3062 (call_i): Remove 'S' asm letter, add jli instruction.
3063 (call_value_i): Likewise.
3064 * config/arc/arc.op (mjli-always): New option.
3065 * config/arc/constraints.md (Cji): New constraint.
3066 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3068 (subsf3_fpx): Likewise.
3069 (mulsf3_fpx): Likewise.
3070 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3072 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3074 * doc/invoke.texi (ARC): Document mjli-always option.
3076 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3078 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3079 avoid addition with 0 and use incw and decw where possible.
3081 2018-01-26 Richard Biener <rguenther@suse.de>
3083 PR tree-optimization/81082
3084 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3085 association if it requires casting to unsigned.
3086 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3087 from fold_plusminus_mult_expr to catch important cases late when
3088 range info is available.
3090 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3092 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3093 * configure.ac (hidden_linkonce): New test.
3094 * configure: Regenerate.
3095 * config.in: Regenerate.
3097 2018-01-26 Julia Koval <julia.koval@intel.com>
3099 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3100 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3101 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3102 _mm_mask_bitshuffle_epi64_mask): Fix type.
3103 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3104 USI_FTYPE_V4DI_V4DI_USI): Remove.
3105 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3106 __builtin_ia32_vpshufbitqmb256_mask,
3107 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3108 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3109 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3111 2018-01-26 Alan Modra <amodra@gmail.com>
3114 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3115 UNSPEC_VBPERMQ. Sort other unspecs.
3117 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3119 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3121 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3124 * predict.c (drop_profile): Do not push/pop cfun; update also
3126 (handle_missing_profiles): Fix logic looking for zero profiles.
3128 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3131 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3132 on functions with #pragma omp declare simd or functions with simd
3134 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3135 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3136 Remove trailing \n from warning_at calls.
3138 2018-01-25 Tom de Vries <tom@codesourcery.com>
3141 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3142 for neutered workers.
3144 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3147 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3148 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3150 2018-01-24 Jeff Law <law@redhat.com>
3153 * i386.c (get_probe_interval): Move to earlier point.
3154 (ix86_compute_frame_layout): If -fstack-clash-protection and
3155 the frame is larger than the probe interval, then use pushes
3156 to save registers rather than reg->mem moves.
3157 (ix86_expand_prologue): Remove conditional for int_registers_saved
3160 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3163 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3164 min/max for never referenced object.
3166 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3169 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3171 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3172 attributes from DECL_ATTRIBUTES (decl) without affecting
3173 DECL_ATTRIBUTES (current_function_decl).
3174 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3175 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3177 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3179 PR tree-optimization/83979
3180 * fold-const.c (fold_comparison): Use constant_boolean_node
3181 instead of boolean_{true,false}_node.
3183 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3185 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3188 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3190 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3191 Simplify the clause that sets the length attribute.
3192 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3193 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3194 clause that sets the length attribute.
3195 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3197 2018-01-24 Tom de Vries <tom@codesourcery.com>
3200 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3201 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3202 Add strict parameter.
3203 (prevent_branch_around_nothing): Insert dummy insn between branch to
3204 label and label with no ptx insn inbetween.
3205 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3207 2018-01-24 Tom de Vries <tom@codesourcery.com>
3210 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3211 for neutered threads in warp.
3212 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3214 2018-01-24 Richard Biener <rguenther@suse.de>
3216 PR tree-optimization/83176
3217 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3220 2018-01-24 Richard Biener <rguenther@suse.de>
3222 PR tree-optimization/82819
3223 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3224 code generating pluses that are no-ops in the target precision.
3226 2018-01-24 Richard Biener <rguenther@suse.de>
3229 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3231 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3233 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3234 to merge probabilities.
3235 * predict.c (probably_never_executed): Also mark as cold functions
3236 with global 0 profile and guessed local profile.
3237 * profile-count.c (profile_probability::combine_with_count): New
3239 * profile-count.h (profile_probability::operator*,
3240 profile_probability::operator*=, profile_probability::operator/,
3241 profile_probability::operator/=): Reduce precision to adjusted
3242 and set value to guessed on contradictory divisions.
3243 (profile_probability::combine_with_freq): Remove.
3244 (profile_probability::combine_wiht_count): Declare.
3245 (profile_count::force_nonzero):: Set to adjusted.
3246 (profile_count::probability_in):: Set quality to adjusted.
3247 * tree-ssa-tail-merge.c (replace_block_by): Use
3250 2018-01-23 Andrew Waterman <andrew@sifive.com>
3251 Jim Wilson <jimw@sifive.com>
3253 * config/riscv/riscv.c (riscv_stack_boundary): New.
3254 (riscv_option_override): Set riscv_stack_boundary. Handle
3255 riscv_preferred_stack_boundary_arg.
3256 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3257 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3258 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3259 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3260 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3261 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3263 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3266 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3267 of struct ix86_frame.
3268 (ix86_expand_epilogue): Likewise. Add a local variable for
3269 the reg_save_offset field in struct ix86_frame.
3271 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3273 PR tree-optimization/82604
3274 * tree-loop-distribution.c (enum partition_kind): New enum item
3275 PKIND_PARTIAL_MEMSET.
3276 (partition_builtin_p): Support above new enum item.
3277 (generate_code_for_partition): Ditto.
3278 (compute_access_range): Differentiate cases that equality can be
3279 proven at all loops, the innermost loops or no loops.
3280 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3281 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3282 (finalize_partitions, distribute_loop): Don't fuse partition of
3283 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3284 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3287 2018-01-23 Martin Liska <mliska@suse.cz>
3289 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3290 order to ignore the predictor.
3291 (PRED_POLYMORPHIC_CALL): Likewise.
3292 (PRED_RECURSIVE_CALL): Likewise.
3294 2018-01-23 Martin Liska <mliska@suse.cz>
3296 * tree-profile.c (tree_profiling): Print function header to
3297 aware reader which function we are working on.
3298 * value-prof.c (gimple_find_values_to_profile): Do not print
3299 not interesting value histograms.
3301 2018-01-23 Martin Liska <mliska@suse.cz>
3303 * profile-count.h (enum profile_quality): Add
3304 profile_uninitialized as the first value. Do not number values
3305 as they are zero based.
3306 (profile_count::verify): Update sanity check.
3307 (profile_probability::verify): Likewise.
3309 2018-01-23 Nathan Sidwell <nathan@acm.org>
3311 * doc/invoke.texi (ffor-scope): Deprecate.
3313 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3315 PR tree-optimization/83510
3316 * domwalk.c (set_all_edges_as_executable): New function.
3317 (dom_walker::dom_walker): Convert bool param
3318 "skip_unreachable_blocks" to enum reachability. Move setup of
3319 edge flags to set_all_edges_as_executable and only do it when
3320 reachability is REACHABLE_BLOCKS.
3321 * domwalk.h (enum dom_walker::reachability): New enum.
3322 (dom_walker::dom_walker): Convert bool param
3323 "skip_unreachable_blocks" to enum reachability.
3324 (set_all_edges_as_executable): New decl.
3325 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3326 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3328 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3329 but converting true to REACHABLE_BLOCKS.
3330 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3332 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3333 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3334 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3336 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3337 if check_all_array_refs will be called.
3339 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3341 * tree.c (selftest::test_location_wrappers): Add more test
3344 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3346 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3347 (selftest::test_bit_in_range): Likewise.
3349 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3352 * doc/sourcebuild.texi (vect_float): Say that the selector
3353 only describes the situation when -funsafe-math-optimizations is on.
3354 (vect_float_strict): Document.
3356 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3358 PR tree-optimization/83965
3359 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3360 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3361 instead of checking only for a reduction.
3362 (vect_recog_widen_sum_pattern): Likewise.
3364 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3366 * predict.c (probably_never_executed): Only use precise profile info.
3367 (compute_function_frequency): Skip after inlining hack since we now
3368 have quality checking.
3370 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3372 * profile-count.h (profile_probability::very_unlikely,
3373 profile_probability::unlikely, profile_probability::even): Set
3374 precision to guessed.
3376 2018-01-23 Richard Biener <rguenther@suse.de>
3378 PR tree-optimization/83963
3379 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3380 Properly terminate dominator walk when crossing the exit edge not
3381 when visiting its source block.
3383 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3386 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3387 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3389 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3391 PR tree-optimization/83957
3392 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3393 semicolon after for body surrounded by braces.
3395 PR tree-optimization/83081
3396 * profile-count.h (profile_probability::split): New method.
3397 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3398 Use profile_probability::split.
3399 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3400 when splitting a single conditional jump into 2.
3402 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3404 PR tree-optimization/69452
3405 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3408 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3410 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3411 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3412 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3414 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3416 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3417 * config/rl78/rl78.md (movdi): New define_expand.
3418 * config/rl78/rl78.c (rl78_split_movdi): New function.
3420 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3423 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3425 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3426 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3427 128-bit to produce an UNSPEC move to get the double word with the
3428 signbit and then a shift directly to do signbit.
3429 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3430 implementation with a new version that just does either a direct
3431 move or a regular move. Move memory interface to separate insns.
3432 Move insns so they are next to the expander.
3433 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3434 with signbit move. Split big and little endian case.
3435 (signbit<mode>2_dm_mem_le): Likewise.
3436 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3437 (signbit<mode>2_dm2): Likewise.
3439 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3441 * config/rl78/rl78.md (anddi3): New define_expand.
3443 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3445 * config/rl78/rl78.md (umindi3): New define_expand.
3447 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3449 * config/rl78/rl78.md (smindi3): New define_expand.
3451 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3453 * config/rl78/rl78.md (smaxdi3): New define_expand.
3455 2018-01-22 Carl Love <cel@us.ibm.com>
3457 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3458 LVX_V1TI): Add macro expansion.
3459 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3460 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3461 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3462 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3463 Change check to determine if the instruction is a byte reversing
3464 entry. Fix typo in comment.
3465 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3466 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3467 Add def_builtin calls for new builtins.
3468 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3469 Add define_insn expansion.
3471 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3473 * config/rl78/rl78.md (umaxdi3): New define_expand.
3475 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3477 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3478 for non-QImode registers.
3480 2018-01-22 Richard Biener <rguenther@suse.de>
3482 PR tree-optimization/83963
3483 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3484 including the loop exit block.
3485 (scop_detection::merge_sese): Likewise.
3486 (scop_detection::add_scop): Do it here instead.
3488 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3490 * doc/sourcebuild.texi (arm_softfloat): Document.
3492 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3495 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3496 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3497 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3499 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3500 David Edelsohn <dje.gcc@gmail.com>
3503 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3504 Change "crset eq" to "crset 2".
3505 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3506 (*call_indirect_aix<mode>_nospec): Likewise.
3507 (*call_value_indirect_aix<mode>_nospec): Likewise.
3508 (*call_indirect_elfv2<mode>_nospec): Likewise.
3509 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3510 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3511 change assembly output from . to $.
3512 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3513 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3514 (*tablejump<mode>_internal1_nospec): Likewise.
3516 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3519 * config/sh/sh_optimize_sett_clrt.cc:
3520 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3522 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3524 PR tree-optimization/83940
3525 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3526 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3527 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3528 use it to pass back the definition type.
3529 (vect_check_store_rhs): Likewise rhs_dt_out.
3530 (vect_build_gather_load_calls): Add a mask_dt argument and use
3531 it instead of a call to vect_is_simple_use.
3532 (vectorizable_store): Update calls to vect_check_load_store_mask
3533 and vect_check_store_rhs. Use the dt returned by the latter instead
3534 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3535 instead of calls to vect_is_simple_use. Pass the scalar rather
3536 than the vector operand to vect_is_simple_use when handling
3537 second and subsequent copies of an rhs value.
3538 (vectorizable_load): Update calls to vect_check_load_store_mask
3539 and vect_build_gather_load_calls. Use the cached mask_dt and
3540 gs_info.offset_dt instead of calls to vect_is_simple_use.
3542 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3545 * tree-emutls.c: Include gimplify.h.
3546 (lower_emutls_2): New function.
3547 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3548 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3549 it before further processing.
3552 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3553 UINTVAL (trueop1) instead of INTVAL (op1).
3555 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3559 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3560 INCOMING_FRAME_SP_OFFSET if not defined.
3561 (scan_trace): Add ENTRY argument. If true and
3562 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3563 emit a note to adjust the CFA offset.
3564 (create_cfi_notes): Adjust scan_trace callers.
3565 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3566 INCOMING_FRAME_SP_OFFSET in the CIE.
3567 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3568 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3570 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3571 * doc/tm.texi: Regenerated.
3573 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3575 PR rtl-optimization/83147
3576 * lra-constraints.c (remove_inheritance_pseudos): Use
3577 lra_substitute_pseudo_within_insn.
3579 2018-01-19 Tom de Vries <tom@codesourcery.com>
3580 Cesar Philippidis <cesar@codesourcery.com>
3583 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3585 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3588 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3589 spaces for function labels.
3591 2018-01-19 Martin Liska <mliska@suse.cz>
3593 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3594 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3595 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3596 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3597 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3598 (PRED_CONST_RETURN): Change from 69 to 65.
3599 (PRED_NULL_RETURN): Change from 91 to 71.
3600 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3601 (PRED_LOOP_GUARD): Change from 66 to 73.
3603 2018-01-19 Martin Liska <mliska@suse.cz>
3605 * predict.c (predict_insn_def): Add new assert.
3606 (struct branch_predictor): Change type to signed integer.
3607 (test_prediction_value_range): Amend test to cover
3609 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3610 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3611 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3612 (PRED_LOOP_IV_COMPARE): Likewise.
3613 * predict.h (PROB_UNINITIALIZED): Define new constant.
3615 2018-01-19 Martin Liska <mliska@suse.cz>
3617 * predict.c (dump_prediction): Add new format for
3618 analyze_brprob.py script which is enabled with -details
3620 * profile-count.h (precise_p): New function.
3622 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3624 PR tree-optimization/83922
3625 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3626 there are no statements that need masking.
3627 (vect_active_double_reduction_p): New function.
3628 (vect_analyze_loop_operations): Use it when handling phis that
3629 are not in the loop header.
3631 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3633 PR tree-optimization/83914
3634 * tree-vect-loop.c (vectorizable_induction): Don't convert
3635 init_expr or apply the peeling adjustment for inductions
3636 that are nested within the vectorized loop.
3638 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3640 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3643 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3647 * function.h (gimplify_parameters): Add gimple_seq * argument.
3648 * function.c: Include gimple.h and options.h.
3649 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3650 for the added local temporaries if needed.
3651 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3652 if there are any parameter cleanups, wrap whole body into a
3653 try/finally with the cleanups.
3655 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3658 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3659 Use GET_MODE_CLASS for scalar floating point.
3661 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3665 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3666 Fix call of call_cgraph_insertion_hooks.
3668 2018-01-18 Martin Sebor <msebor@redhat.com>
3670 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3672 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3675 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3678 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3681 * common.opt: (-ffile-prefix-map): New option.
3682 * opts.c (common_handle_option): Defer it.
3683 * opts-global.c (handle_common_deferred_options): Handle it.
3684 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3685 * file-prefix-map.h: New file.
3686 (remap_debug_filename, add_debug_prefix_map): ...here.
3687 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3688 * final.c (debug_prefix_map, add_debug_prefix_map
3689 remap_debug_filename): Move to...
3690 * file-prefix-map.c: New file.
3691 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3692 generalize, get rid of alloca(), use strrchr() instead of strchr().
3693 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3694 Implement in terms of add_prefix_map().
3695 (remap_macro_filename, remap_debug_filename): Implement in term of
3697 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3698 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3699 * dbxout.c: Include file-prefix-map.h.
3700 * varasm.c: Likewise.
3701 * vmsdbgout.c: Likewise.
3702 * xcoffout.c: Likewise.
3703 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3704 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3705 * doc/invoke.texi (-ffile-prefix-map): Document.
3706 (-fdebug-prefix-map): Update description.
3708 2018-01-18 Martin Liska <mliska@suse.cz>
3710 * config/i386/i386.c (indirect_thunk_name): Document that also
3712 (output_indirect_thunk): Document why both instructions
3713 (pause and lfence) are generated.
3715 2018-01-18 Richard Biener <rguenther@suse.de>
3717 PR tree-optimization/83887
3718 * graphite-scop-detection.c
3719 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3720 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3721 (scop_detection::merge_sese): Re-implement with a flood-fill
3722 algorithm that properly finds a SESE region if it exists.
3724 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3727 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3728 pointer_diff optimizations use view_convert instead of convert.
3730 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3732 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3733 Generate different code for -mno-speculate-indirect-jumps.
3734 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3735 (*call_indirect_aix<mode>): Disable for
3736 -mno-speculate-indirect-jumps.
3737 (*call_indirect_aix<mode>_nospec): New define_insn.
3738 (*call_value_indirect_aix<mode>): Disable for
3739 -mno-speculate-indirect-jumps.
3740 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3741 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3742 -mno-speculate-indirect-jumps.
3743 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3745 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3747 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3748 long double type, set the flags for noting the default long double
3749 type, even if we don't pass or return a long double type.
3751 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3754 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3757 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3759 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3760 support for merge[hl].
3761 (fold_mergehl_helper): New helper function.
3762 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3763 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3764 (altivec_vmrglw_direct): Add xxmrglw insn.
3766 2018-01-17 Andrew Waterman <andrew@sifive.com>
3768 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3769 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3771 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3774 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3775 call the lto_location_cache before reading the
3776 DECL_SOURCE_LOCATION of the types.
3778 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3779 Richard Sandiford <richard.sandiford@linaro.org>
3781 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3782 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3783 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3784 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3785 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3787 * config/aarch64/constraints.md (aarch64_movti_operand):
3789 * config/aarch64/predicates.md (Uti): Add new constraint.
3791 2018-01-17 Carl Love <cel@us.ibm.com>
3793 * config/rs6000/vsx.md (define_expand xl_len_r,
3794 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3795 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3797 (define_expand, define_insn): Move the shift left from the
3798 define_insn to the define_expand for lxvl and stxvl instructions.
3799 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3800 and XL_LEN_R definitions to PURE.
3802 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3804 * config/i386/i386.c (indirect_thunk_name): Declare regno
3805 as unsigned int. Compare regno with INVALID_REGNUM.
3806 (output_indirect_thunk): Ditto.
3807 (output_indirect_thunk_function): Ditto.
3808 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3809 in the call to output_indirect_thunk_function.
3811 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3814 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3815 rather than the size of inner_type to determine the stack slot size
3816 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3818 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3821 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3824 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3826 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3827 endian Linux systems to optionally enable multilibs for selecting
3828 the long double type if the user configured an explicit type.
3829 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3830 have no long double multilibs if not defined.
3831 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3832 warn if the user used -mabi={ieee,ibm}longdouble and we built
3833 multilibs for long double.
3834 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3835 appropriate multilib option.
3836 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3838 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3839 for building long double multilibs.
3840 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3842 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3844 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3847 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3849 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3852 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3855 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3858 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3860 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3861 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3863 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3865 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3866 different rtl trees depending on TARGET_64BIT.
3867 (rs6000_gen_lvx): Likewise.
3869 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3871 * config/visium/visium.md (nop): Tweak comment.
3872 (hazard_nop): Likewise.
3874 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3876 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3877 -mspeculate-indirect-jumps.
3878 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3879 for -mno-speculate-indirect-jumps.
3880 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3881 (*call_value_indirect_elfv2<mode>): Disable for
3882 -mno-speculate-indirect-jumps.
3883 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3884 (indirect_jump): Emit different RTL for
3885 -mno-speculate-indirect-jumps.
3886 (*indirect_jump<mode>): Disable for
3887 -mno-speculate-indirect-jumps.
3888 (*indirect_jump<mode>_nospec): New define_insn.
3889 (tablejump): Emit different RTL for
3890 -mno-speculate-indirect-jumps.
3891 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3892 (tablejumpsi_nospec): New define_expand.
3893 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3894 (tablejumpdi_nospec): New define_expand.
3895 (*tablejump<mode>_internal1): Disable for
3896 -mno-speculate-indirect-jumps.
3897 (*tablejump<mode>_internal1_nospec): New define_insn.
3898 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3901 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3903 * caller-save.c (insert_save): Drop unnecessary parameter. All
3906 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3907 Richard Biener <rguenth@suse.de>
3910 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3911 return early, inline manually is_gimple_sizepos. Make sure if we
3912 call gimplify_expr we don't end up with a gimple constant.
3913 * tree.c (variably_modified_type_p): Don't return true for
3914 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3915 * gimplify.h (is_gimple_sizepos): Remove.
3917 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3919 PR tree-optimization/83857
3920 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3921 vectorizable_live_operation for pure SLP statements.
3922 (vectorizable_live_operation): Handle PHIs.
3924 2018-01-16 Richard Biener <rguenther@suse.de>
3926 PR tree-optimization/83867
3927 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3928 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3930 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3933 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3934 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3935 If off is not INTEGER_CST, issue a may not be aligned warning
3936 rather than isn't aligned. Use isn%'t rather than isn't.
3937 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3939 <case MULT_EXPR>: Improve the case when bottom and one of the
3940 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3941 operand, in that case check if the other operand is multiple of
3942 bottom divided by the INTEGER_CST operand.
3944 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3947 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3948 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3949 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3950 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3951 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3952 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3953 (pa_function_arg_size): New function.
3955 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3957 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3958 in a separate statement.
3960 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3962 PR tree-optimization/83847
3963 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3964 group gathers and scatters.
3966 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3968 PR rtl-optimization/86620
3969 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3971 PR rtl-optimization/83213
3972 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3973 to last if both are JUMP_INSNs.
3975 PR tree-optimization/83843
3976 * gimple-ssa-store-merging.c
3977 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3978 store_immediate_info for bswap/nop orig_stores.
3980 2018-01-15 Andrew Waterman <andrew@sifive.com>
3982 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3984 <UDIV>: Increase cost if !TARGET_DIV.
3986 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3988 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3989 (define_attr "cr_logical_3op"): New.
3990 (cceq_ior_compare): Adjust.
3991 (cceq_ior_compare_complement): Adjust.
3992 (*cceq_rev_compare): Adjust.
3993 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3994 (is_cracked_insn): Adjust.
3995 (insn_must_be_first_in_group): Adjust.
3996 * config/rs6000/40x.md: Adjust.
3997 * config/rs6000/440.md: Adjust.
3998 * config/rs6000/476.md: Adjust.
3999 * config/rs6000/601.md: Adjust.
4000 * config/rs6000/603.md: Adjust.
4001 * config/rs6000/6xx.md: Adjust.
4002 * config/rs6000/7450.md: Adjust.
4003 * config/rs6000/7xx.md: Adjust.
4004 * config/rs6000/8540.md: Adjust.
4005 * config/rs6000/cell.md: Adjust.
4006 * config/rs6000/e300c2c3.md: Adjust.
4007 * config/rs6000/e500mc.md: Adjust.
4008 * config/rs6000/e500mc64.md: Adjust.
4009 * config/rs6000/e5500.md: Adjust.
4010 * config/rs6000/e6500.md: Adjust.
4011 * config/rs6000/mpc.md: Adjust.
4012 * config/rs6000/power4.md: Adjust.
4013 * config/rs6000/power5.md: Adjust.
4014 * config/rs6000/power6.md: Adjust.
4015 * config/rs6000/power7.md: Adjust.
4016 * config/rs6000/power8.md: Adjust.
4017 * config/rs6000/power9.md: Adjust.
4018 * config/rs6000/rs64.md: Adjust.
4019 * config/rs6000/titan.md: Adjust.
4021 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4023 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4024 ix86_indirect_branch_register logic.
4026 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4028 * config/i386/constraints.md (Bs): Update
4029 ix86_indirect_branch_register check. Don't check
4030 ix86_indirect_branch_register with GOT_memory_operand.
4032 * config/i386/predicates.md (GOT_memory_operand): Don't check
4033 ix86_indirect_branch_register here.
4034 (GOT32_symbol_operand): Likewise.
4036 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4038 * config/i386/predicates.md (constant_call_address_operand):
4039 Rewrite ix86_indirect_branch_register logic.
4040 (sibcall_insn_operand): Likewise.
4042 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4044 * config/i386/constraints.md (Bs): Replace
4045 ix86_indirect_branch_thunk_register with
4046 ix86_indirect_branch_register.
4048 * config/i386/i386.md (indirect_jump): Likewise.
4049 (tablejump): Likewise.
4050 (*sibcall_memory): Likewise.
4051 (*sibcall_value_memory): Likewise.
4052 Peepholes of indirect call and jump via memory: Likewise.
4053 * config/i386/i386.opt: Likewise.
4054 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4055 (GOT_memory_operand): Likewise.
4056 (call_insn_operand): Likewise.
4057 (sibcall_insn_operand): Likewise.
4058 (GOT32_symbol_operand): Likewise.
4060 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4063 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4064 type rather than type addr's type points to.
4065 (expand_omp_atomic_mutex): Likewise.
4066 (expand_omp_atomic): Likewise.
4068 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4071 * config/i386/i386.c (output_indirect_thunk_function): Use
4072 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4073 for __x86_return_thunk.
4075 2018-01-15 Richard Biener <rguenther@suse.de>
4078 * expmed.c (extract_bit_field_1): Fix typo.
4080 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4083 * config/arm/iterators.md (VF): New mode iterator.
4084 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4085 Remove integer-related logic from pattern.
4086 (neon_vabd<mode>_3): Likewise.
4088 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4091 * common.opt (fstrict-overflow): No longer an alias.
4092 (fwrapv-pointer): New option.
4093 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4094 also for pointer types based on flag_wrapv_pointer.
4095 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4096 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4097 opts->x_flag_wrapv got set.
4098 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4099 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4100 POINTER_TYPE_OVERFLOW_UNDEFINED.
4101 * match.pd: Likewise in address comparison pattern.
4102 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4104 2018-01-15 Richard Biener <rguenther@suse.de>
4107 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4108 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4109 Reset type names to their identifier if their TYPE_DECL doesn't
4110 have linkage (and thus is used for ODR and devirt).
4111 (save_debug_info_for_decl): Remove.
4112 (save_debug_info_for_type): Likewise.
4113 (add_tree_to_fld_list): Adjust.
4114 * tree-pretty-print.c (dump_generic_node): Make dumping of
4115 type names more robust.
4117 2018-01-15 Richard Biener <rguenther@suse.de>
4119 * BASE-VER: Bump to 8.0.1.
4121 2018-01-14 Martin Sebor <msebor@redhat.com>
4124 * builtins.c (check_access): Avoid warning when the no-warning bit
4127 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4129 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4130 * ira-color (allocno_hard_regs_compare): Likewise.
4132 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4135 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4136 Use .pushsection/.popsection.
4138 2018-01-14 Martin Sebor <msebor@redhat.com>
4141 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4143 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4145 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4146 entry from extra_headers.
4147 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4148 extra_headers, make the list bitwise identical to the i?86-*-* one.
4150 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4152 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4153 -mcmodel=large with -mindirect-branch=thunk,
4154 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4155 -mfunction-return=thunk-extern.
4156 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4157 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4158 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4160 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4162 * config/i386/i386.c (print_reg): Print the name of the full
4163 integer register without '%'.
4164 (ix86_print_operand): Handle 'V'.
4165 * doc/extend.texi: Document 'V' modifier.
4167 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4169 * config/i386/constraints.md (Bs): Disallow memory operand for
4170 -mindirect-branch-register.
4172 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4173 (GOT_memory_operand): Likewise.
4174 (call_insn_operand): Likewise.
4175 (sibcall_insn_operand): Likewise.
4176 (GOT32_symbol_operand): Likewise.
4177 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4178 for -mindirect-branch-register.
4179 (tablejump): Likewise.
4180 (*sibcall_memory): Likewise.
4181 (*sibcall_value_memory): Likewise.
4182 Disallow peepholes of indirect call and jump via memory for
4183 -mindirect-branch-register.
4184 (*call_pop): Replace m with Bw.
4185 (*call_value_pop): Likewise.
4186 (*sibcall_pop_memory): Replace m with Bs.
4187 * config/i386/i386.opt (mindirect-branch-register): New option.
4188 * doc/invoke.texi: Document -mindirect-branch-register option.
4190 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4192 * config/i386/i386-protos.h (ix86_output_function_return): New.
4193 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4194 set function_return_type.
4195 (indirect_thunk_name): Add ret_p to indicate thunk for function
4197 (output_indirect_thunk_function): Pass false to
4198 indirect_thunk_name.
4199 (ix86_output_indirect_branch_via_reg): Likewise.
4200 (ix86_output_indirect_branch_via_push): Likewise.
4201 (output_indirect_thunk_function): Create alias for function
4202 return thunk if regno < 0.
4203 (ix86_output_function_return): New function.
4204 (ix86_handle_fndecl_attribute): Handle function_return.
4205 (ix86_attribute_table): Add function_return.
4206 * config/i386/i386.h (machine_function): Add
4207 function_return_type.
4208 * config/i386/i386.md (simple_return_internal): Use
4209 ix86_output_function_return.
4210 (simple_return_internal_long): Likewise.
4211 * config/i386/i386.opt (mfunction-return=): New option.
4212 (indirect_branch): Mention -mfunction-return=.
4213 * doc/extend.texi: Document function_return function attribute.
4214 * doc/invoke.texi: Document -mfunction-return= option.
4216 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4218 * config/i386/i386-opts.h (indirect_branch): New.
4219 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4220 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4221 with local indirect jump when converting indirect call and jump.
4222 (ix86_set_indirect_branch_type): New.
4223 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4224 (indirectlabelno): New.
4225 (indirect_thunk_needed): Likewise.
4226 (indirect_thunk_bnd_needed): Likewise.
4227 (indirect_thunks_used): Likewise.
4228 (indirect_thunks_bnd_used): Likewise.
4229 (INDIRECT_LABEL): Likewise.
4230 (indirect_thunk_name): Likewise.
4231 (output_indirect_thunk): Likewise.
4232 (output_indirect_thunk_function): Likewise.
4233 (ix86_output_indirect_branch_via_reg): Likewise.
4234 (ix86_output_indirect_branch_via_push): Likewise.
4235 (ix86_output_indirect_branch): Likewise.
4236 (ix86_output_indirect_jmp): Likewise.
4237 (ix86_code_end): Call output_indirect_thunk_function if needed.
4238 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4240 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4241 (ix86_attribute_table): Add indirect_branch.
4242 * config/i386/i386.h (machine_function): Add indirect_branch_type
4243 and has_local_indirect_jump.
4244 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4246 (tablejump): Likewise.
4247 (*indirect_jump): Use ix86_output_indirect_jmp.
4248 (*tablejump_1): Likewise.
4249 (simple_return_indirect_internal): Likewise.
4250 * config/i386/i386.opt (mindirect-branch=): New option.
4251 (indirect_branch): New.
4254 (thunk-inline): Likewise.
4255 (thunk-extern): Likewise.
4256 * doc/extend.texi: Document indirect_branch function attribute.
4257 * doc/invoke.texi: Document -mindirect-branch= option.
4259 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4262 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4264 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4266 * ipa-inline.c (want_inline_small_function_p): Return false if
4267 inlining has already failed with CIF_FINAL_ERROR.
4268 (update_caller_keys): Call want_inline_small_function_p before
4270 (update_callee_keys): Likewise.
4272 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4274 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4276 (rs6000_quadword_masked_address_p): Likewise.
4277 (quad_aligned_load_p): Likewise.
4278 (quad_aligned_store_p): Likewise.
4279 (const_load_sequence_p): Add comment to describe the outer-most loop.
4280 (mimic_memory_attributes_and_flags): New function.
4281 (rs6000_gen_stvx): Likewise.
4282 (replace_swapped_aligned_store): Likewise.
4283 (rs6000_gen_lvx): Likewise.
4284 (replace_swapped_aligned_load): Likewise.
4285 (replace_swapped_load_constant): Capitalize argument name in
4286 comment describing this function.
4287 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4288 and stores that access quad-word aligned addresses and replace
4289 with stvx or lvx instructions when appropriate.
4290 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4291 New function prototype.
4292 (rs6000_quadword_masked_address_p): Likewise.
4293 (rs6000_gen_lvx): Likewise.
4294 (rs6000_gen_stvx): Likewise.
4295 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4296 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4297 when memory address is aligned.
4298 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4299 this split to select lvx instruction when memory address is aligned.
4300 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4301 instruction when memory address is aligned.
4302 (*vsx_le_perm_load_v16qi): Likewise.
4303 (four unnamed splitters): Modify to select the stvx instruction
4304 when memory is aligned.
4306 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4308 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4309 which appears in the queue multiple times.
4311 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4312 Alan Hayward <alan.hayward@arm.com>
4313 David Sherwood <david.sherwood@arm.com>
4315 * tree-vectorizer.h (vec_lower_bound): New structure.
4316 (_loop_vec_info): Add check_nonzero and lower_bounds.
4317 (LOOP_VINFO_CHECK_NONZERO): New macro.
4318 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4319 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4320 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4321 fields. Make seg_len the distance travelled, not including the
4323 (dr_direction_indicator): Declare.
4324 (dr_zero_step_indicator): Likewise.
4325 (dr_known_forward_stride_p): Likewise.
4326 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4328 (runtime_alias_check_p): Allow runtime alias checks with
4330 (operator ==): Compare access_size and align.
4331 (prune_runtime_alias_test_list): Rework for new distinction between
4332 the access_size and seg_len.
4333 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4335 (get_segment_min_max): New function.
4336 (create_intersect_range_checks): Use it.
4337 (dr_step_indicator): New function.
4338 (dr_direction_indicator): Likewise.
4339 (dr_zero_step_indicator): Likewise.
4340 (dr_known_forward_stride_p): Likewise.
4341 * tree-loop-distribution.c (data_ref_segment_size): Return
4342 DR_STEP * (niters - 1).
4343 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4345 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4346 (vect_preserves_scalar_order_p): New function, split out from...
4347 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4348 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4349 (vect_vfa_access_size): New function.
4350 (vect_vfa_align): Likewise.
4351 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4352 (dump_lower_bound): New function.
4353 (vect_check_lower_bound): Likewise.
4354 (vect_small_gap_p): Likewise.
4355 (vectorizable_with_step_bound_p): Likewise.
4356 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4357 depencies if the vectorization factor is 1. Convert the checks
4358 for nonzero steps into checks on the bounds of DR_STEP. Try using
4359 a bunds check for variable steps if the minimum required step is
4360 relatively small. Update calls to the dr_with_seg_len
4361 constructor and to vect_compile_time_alias.
4362 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4364 (vect_loop_versioning): Call it.
4365 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4367 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4369 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4370 Alan Hayward <alan.hayward@arm.com>
4371 David Sherwood <david.sherwood@arm.com>
4373 * doc/sourcebuild.texi (vect_scatter_store): Document.
4374 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4376 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4378 * genopinit.c (main): Add supports_vec_scatter_store and
4379 supports_vec_scatter_store_cached to target_optabs.
4380 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4381 IFN_MASK_SCATTER_STORE.
4382 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4384 * internal-fn.h (internal_store_fn_p): Declare.
4385 (internal_fn_stored_value_index): Likewise.
4386 * internal-fn.c (scatter_store_direct): New macro.
4387 (expand_scatter_store_optab_fn): New function.
4388 (direct_scatter_store_optab_supported_p): New macro.
4389 (internal_store_fn_p): New function.
4390 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4391 IFN_MASK_SCATTER_STORE.
4392 (internal_fn_mask_index): Likewise.
4393 (internal_fn_stored_value_index): New function.
4394 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4396 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4397 * optabs-query.c (supports_vec_scatter_store_p): New function.
4398 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4399 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4400 true for scatter stores.
4401 (vect_gather_scatter_fn_p): Handle scatter stores too.
4402 (vect_check_gather_scatter): Consider using scatter stores if
4403 supports_vec_scatter_store_p.
4404 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4406 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4407 internal_fn_stored_value_index.
4408 (check_load_store_masking): Handle scatter stores too.
4409 (vect_get_store_rhs): Make public.
4410 (vectorizable_call): Use internal_store_fn_p.
4411 (vectorizable_store): Handle scatter store internal functions.
4412 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4413 when deciding whether the end of the group has been reached.
4414 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4415 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4416 (mask_scatter_store<mode>): New insns.
4418 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4419 Alan Hayward <alan.hayward@arm.com>
4420 David Sherwood <david.sherwood@arm.com>
4422 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4423 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4424 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4426 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4427 Use vect_truncate_gather_scatter_offset if we can't treat the
4428 operation as a normal gather load or scatter store.
4429 (get_group_load_store_type): Take the gather_scatter_info
4430 as argument. Try using a gather load or scatter store for
4431 single-element groups.
4432 (get_load_store_type): Update calls to get_group_load_store_type
4433 and vect_use_strided_gather_scatters_p.
4435 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4436 Alan Hayward <alan.hayward@arm.com>
4437 David Sherwood <david.sherwood@arm.com>
4439 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4440 optional tree argument.
4441 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4443 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4444 but continue to use the current value as a fallback.
4445 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4446 to compare the updates.
4447 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4448 (get_load_store_type): Use it when handling a strided access.
4449 (vect_get_strided_load_store_ops): New function.
4450 (vect_get_data_ptr_increment): Likewise.
4451 (vectorizable_load): Handle strided gather loads. Always pass
4452 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4454 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4455 Alan Hayward <alan.hayward@arm.com>
4456 David Sherwood <david.sherwood@arm.com>
4458 * doc/md.texi (gather_load@var{m}): Document.
4459 (mask_gather_load@var{m}): Likewise.
4460 * genopinit.c (main): Add supports_vec_gather_load and
4461 supports_vec_gather_load_cached to target_optabs.
4462 * optabs-tree.c (init_tree_optimization_optabs): Use
4463 ggc_cleared_alloc to allocate target_optabs.
4464 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4465 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4467 * internal-fn.h (internal_load_fn_p): Declare.
4468 (internal_gather_scatter_fn_p): Likewise.
4469 (internal_fn_mask_index): Likewise.
4470 (internal_gather_scatter_fn_supported_p): Likewise.
4471 * internal-fn.c (gather_load_direct): New macro.
4472 (expand_gather_load_optab_fn): New function.
4473 (direct_gather_load_optab_supported_p): New macro.
4474 (direct_internal_fn_optab): New function.
4475 (internal_load_fn_p): Likewise.
4476 (internal_gather_scatter_fn_p): Likewise.
4477 (internal_fn_mask_index): Likewise.
4478 (internal_gather_scatter_fn_supported_p): Likewise.
4479 * optabs-query.c (supports_at_least_one_mode_p): New function.
4480 (supports_vec_gather_load_p): Likewise.
4481 * optabs-query.h (supports_vec_gather_load_p): Declare.
4482 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4483 and memory_type field.
4484 (NUM_PATTERNS): Bump to 15.
4485 * tree-vect-data-refs.c: Include internal-fn.h.
4486 (vect_gather_scatter_fn_p): New function.
4487 (vect_describe_gather_scatter_call): Likewise.
4488 (vect_check_gather_scatter): Try using internal functions for
4489 gather loads. Recognize existing calls to a gather load function.
4490 (vect_analyze_data_refs): Consider using gather loads if
4491 supports_vec_gather_load_p.
4492 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4493 (vect_get_gather_scatter_offset_type): Likewise.
4494 (vect_convert_mask_for_vectype): Likewise.
4495 (vect_add_conversion_to_patterm): Likewise.
4496 (vect_try_gather_scatter_pattern): Likewise.
4497 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4498 (vect_vect_recog_func_ptrs): Add it.
4499 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4500 internal_fn_mask_index and internal_gather_scatter_fn_p.
4501 (check_load_store_masking): Take the gather_scatter_info as an
4502 argument and handle gather loads.
4503 (vect_get_gather_scatter_ops): New function.
4504 (vectorizable_call): Check internal_load_fn_p.
4505 (vectorizable_load): Likewise. Handle gather load internal
4507 (vectorizable_store): Update call to check_load_store_masking.
4508 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4509 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4510 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4511 (aarch64_gather_scale_operand_d): New predicates.
4512 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4513 (mask_gather_load<mode>): New insns.
4515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4516 Alan Hayward <alan.hayward@arm.com>
4517 David Sherwood <david.sherwood@arm.com>
4519 * optabs.def (fold_left_plus_optab): New optab.
4520 * doc/md.texi (fold_left_plus_@var{m}): Document.
4521 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4522 * internal-fn.c (fold_left_direct): Define.
4523 (expand_fold_left_optab_fn): Likewise.
4524 (direct_fold_left_optab_supported_p): Likewise.
4525 * fold-const-call.c (fold_const_fold_left): New function.
4526 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4527 * tree-parloops.c (valid_reduction_p): New function.
4528 (gather_scalar_reductions): Use it.
4529 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4530 (vect_finish_replace_stmt): Declare.
4531 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4532 (needs_fold_left_reduction_p): New function, split out from...
4533 (vect_is_simple_reduction): ...here. Accept reductions that
4534 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4535 (vect_force_simple_reduction): Also store the reduction type in
4536 the assignment's STMT_VINFO_REDUC_TYPE.
4537 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4538 (merge_with_identity): New function.
4539 (vect_expand_fold_left): Likewise.
4540 (vectorize_fold_left_reduction): Likewise.
4541 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4542 scalar phi in place for it. Check for target support and reject
4543 cases that would reassociate the operation. Defer the transform
4544 phase to vectorize_fold_left_reduction.
4545 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4546 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4547 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4551 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4552 call to ifc_temp_var.
4554 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4555 Alan Hayward <alan.hayward@arm.com>
4556 David Sherwood <david.sherwood@arm.com>
4558 * target.def (legitimize_address_displacement): Take the original
4559 offset as a poly_int.
4560 * targhooks.h (default_legitimize_address_displacement): Update
4562 * targhooks.c (default_legitimize_address_displacement): Likewise.
4563 * doc/tm.texi: Regenerate.
4564 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4565 as an argument, moving assert of ad->disp == ad->disp_term to...
4566 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4567 Try calling targetm.legitimize_address_displacement before expanding
4568 the address rather than afterwards, and adjust for the new interface.
4569 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4570 Match the new hook interface. Handle SVE addresses.
4571 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4574 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4576 * Makefile.in (OBJS): Add early-remat.o.
4577 * target.def (select_early_remat_modes): New hook.
4578 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4579 * doc/tm.texi: Regenerate.
4580 * targhooks.h (default_select_early_remat_modes): Declare.
4581 * targhooks.c (default_select_early_remat_modes): New function.
4582 * timevar.def (TV_EARLY_REMAT): New timevar.
4583 * passes.def (pass_early_remat): New pass.
4584 * tree-pass.h (make_pass_early_remat): Declare.
4585 * early-remat.c: New file.
4586 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4588 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4595 vfm1 with a bound_epilog parameter.
4596 (vect_do_peeling): Update calls accordingly, and move the prologue
4597 call earlier in the function. Treat the base bound_epilog as 0 for
4598 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4599 this base when peeling for gaps.
4600 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4601 with fully-masked loops.
4602 (vect_estimate_min_profitable_iters): Handle the single peeled
4603 iteration in that case.
4605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4606 Alan Hayward <alan.hayward@arm.com>
4607 David Sherwood <david.sherwood@arm.com>
4609 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4610 single-element interleaving even if the size is not a power of 2.
4611 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4612 accesses for single-element interleaving if the group size is
4615 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4616 Alan Hayward <alan.hayward@arm.com>
4617 David Sherwood <david.sherwood@arm.com>
4619 * doc/md.texi (fold_extract_last_@var{m}): Document.
4620 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4621 * optabs.def (fold_extract_last_optab): New optab.
4622 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4623 * internal-fn.c (fold_extract_direct): New macro.
4624 (expand_fold_extract_optab_fn): Likewise.
4625 (direct_fold_extract_optab_supported_p): Likewise.
4626 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4627 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4628 EXTRACT_LAST_REDUCTION.
4629 (get_initial_def_for_reduction): Do not create an initial vector
4630 for EXTRACT_LAST_REDUCTION reductions.
4631 (vectorizable_reduction): Leave the scalar phi in place for
4632 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4633 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4634 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4635 transform phase to vectorizable_condition.
4636 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4638 (vect_finish_stmt_generation): ...here.
4639 (vect_finish_replace_stmt): New function.
4640 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4641 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4643 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4645 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4646 Alan Hayward <alan.hayward@arm.com>
4647 David Sherwood <david.sherwood@arm.com>
4649 * doc/md.texi (extract_last_@var{m}): Document.
4650 * optabs.def (extract_last_optab): New optab.
4651 * internal-fn.def (EXTRACT_LAST): New internal function.
4652 * internal-fn.c (cond_unary_direct): New macro.
4653 (expand_cond_unary_optab_fn): Likewise.
4654 (direct_cond_unary_optab_supported_p): Likewise.
4655 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4656 loops using EXTRACT_LAST.
4657 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4658 (extract_last_<mode>): ...this optab.
4659 (vec_extract<mode><Vel>): Update accordingly.
4661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4662 Alan Hayward <alan.hayward@arm.com>
4663 David Sherwood <david.sherwood@arm.com>
4665 * target.def (empty_mask_is_expensive): New hook.
4666 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4667 * doc/tm.texi: Regenerate.
4668 * targhooks.h (default_empty_mask_is_expensive): Declare.
4669 * targhooks.c (default_empty_mask_is_expensive): New function.
4670 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4671 if the target says that empty masks are expensive.
4672 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4674 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4677 Alan Hayward <alan.hayward@arm.com>
4678 David Sherwood <david.sherwood@arm.com>
4680 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4681 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4682 (vect_use_loop_mask_for_alignment_p): New function.
4683 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4684 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4685 niters_skip argument. Make sure that the first niters_skip elements
4686 of the first iteration are inactive.
4687 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4688 Update call to vect_set_loop_masks_directly.
4689 (get_misalign_in_elems): New function, split out from...
4690 (vect_gen_prolog_loop_niters): ...here.
4691 (vect_update_init_of_dr): Take a code argument that specifies whether
4692 the adjustment should be added or subtracted.
4693 (vect_update_init_of_drs): Likewise.
4694 (vect_prepare_for_masked_peels): New function.
4695 (vect_do_peeling): Skip prologue peeling if we're using a mask
4696 instead. Update call to vect_update_inits_of_drs.
4697 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4699 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4700 alignment. Do not include the number of peeled iterations in
4701 the minimum threshold in that case.
4702 (vectorizable_induction): Adjust the start value down by
4703 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4704 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4705 Take the number of skipped iterations into account when calculating
4707 * tree-vect-stmts.c (vect_gen_while_not): New function.
4709 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4710 Alan Hayward <alan.hayward@arm.com>
4711 David Sherwood <david.sherwood@arm.com>
4713 * doc/sourcebuild.texi (vect_fully_masked): Document.
4714 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4716 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4718 (vect_analyze_loop_2): ...here. Don't check the vectorization
4719 factor against the number of loop iterations if the loop is
4722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4723 Alan Hayward <alan.hayward@arm.com>
4724 David Sherwood <david.sherwood@arm.com>
4726 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4727 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4728 (dump_groups): Update accordingly.
4729 (iv_use::mem_type): New member variable.
4730 (address_p): New function.
4731 (record_use): Add a mem_type argument and initialize the new
4733 (record_group_use): Add a mem_type argument. Use address_p.
4734 Remove obsolete null checks of base_object. Update call to record_use.
4735 (find_interesting_uses_op): Update call to record_group_use.
4736 (find_interesting_uses_cond): Likewise.
4737 (find_interesting_uses_address): Likewise.
4738 (get_mem_type_for_internal_fn): New function.
4739 (find_address_like_use): Likewise.
4740 (find_interesting_uses_stmt): Try find_address_like_use before
4741 calling find_interesting_uses_op.
4742 (addr_offset_valid_p): Use the iv mem_type field as the type
4743 of the addressed memory.
4744 (add_autoinc_candidates): Likewise.
4745 (get_address_cost): Likewise.
4746 (split_small_address_groups_p): Use address_p.
4747 (split_address_groups): Likewise.
4748 (add_iv_candidate_for_use): Likewise.
4749 (autoinc_possible_for_pair): Likewise.
4750 (rewrite_groups): Likewise.
4751 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4752 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4753 (get_alias_ptr_type_for_ptr_address): New function.
4754 (rewrite_use_address): Rewrite address uses in calls that were
4755 identified by find_address_like_use.
4757 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4758 Alan Hayward <alan.hayward@arm.com>
4759 David Sherwood <david.sherwood@arm.com>
4761 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4763 * gimple-expr.h (is_gimple_addressable: Likewise.
4764 * gimple-expr.c (is_gimple_address): Likewise.
4765 * internal-fn.c (expand_call_mem_ref): New function.
4766 (expand_mask_load_optab_fn): Use it.
4767 (expand_mask_store_optab_fn): Likewise.
4769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4770 Alan Hayward <alan.hayward@arm.com>
4771 David Sherwood <david.sherwood@arm.com>
4773 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4774 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4775 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4776 (cond_umax@var{mode}): Document.
4777 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4778 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4779 (cond_umin_optab, cond_umax_optab): New optabs.
4780 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4781 (COND_IOR, COND_XOR): New internal functions.
4782 * internal-fn.h (get_conditional_internal_fn): Declare.
4783 * internal-fn.c (cond_binary_direct): New macro.
4784 (expand_cond_binary_optab_fn): Likewise.
4785 (direct_cond_binary_optab_supported_p): Likewise.
4786 (get_conditional_internal_fn): New function.
4787 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4788 Cope with reduction statements that are vectorized as calls rather
4790 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4791 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4792 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4793 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4794 (UNSPEC_COND_EOR): New unspecs.
4795 (optab): Add mappings for them.
4796 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4797 (sve_int_op, sve_fp_op): New int attributes.
4799 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4800 Alan Hayward <alan.hayward@arm.com>
4801 David Sherwood <david.sherwood@arm.com>
4803 * optabs.def (while_ult_optab): New optab.
4804 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4805 * internal-fn.def (WHILE_ULT): New internal function.
4806 * internal-fn.h (direct_internal_fn_supported_p): New override
4807 that takes two types as argument.
4808 * internal-fn.c (while_direct): New macro.
4809 (expand_while_optab_fn): New function.
4810 (convert_optab_supported_p): Likewise.
4811 (direct_while_optab_supported_p): New macro.
4812 * wide-int.h (wi::udiv_ceil): New function.
4813 * tree-vectorizer.h (rgroup_masks): New structure.
4814 (vec_loop_masks): New typedef.
4815 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4817 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4818 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4819 (vect_max_vf): New function.
4820 (slpeel_make_loop_iterate_ntimes): Delete.
4821 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4822 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4823 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4824 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4825 internal-fn.h, stor-layout.h and optabs-query.h.
4826 (vect_set_loop_mask): New function.
4827 (add_preheader_seq): Likewise.
4828 (add_header_seq): Likewise.
4829 (interleave_supported_p): Likewise.
4830 (vect_maybe_permute_loop_masks): Likewise.
4831 (vect_set_loop_masks_directly): Likewise.
4832 (vect_set_loop_condition_masked): Likewise.
4833 (vect_set_loop_condition_unmasked): New function, split out from
4834 slpeel_make_loop_iterate_ntimes.
4835 (slpeel_make_loop_iterate_ntimes): Rename to..
4836 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4837 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4838 (vect_do_peeling): Update call accordingly.
4839 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4841 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4842 mask_compare_type, can_fully_mask_p and fully_masked_p.
4843 (release_vec_loop_masks): New function.
4844 (_loop_vec_info): Use it to free the loop masks.
4845 (can_produce_all_loop_masks_p): New function.
4846 (vect_get_max_nscalars_per_iter): Likewise.
4847 (vect_verify_full_masking): Likewise.
4848 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4849 retries, and free the mask rgroups before retrying. Check loop-wide
4850 reasons for disallowing fully-masked loops. Make the final decision
4851 about whether use a fully-masked loop or not.
4852 (vect_estimate_min_profitable_iters): Do not assume that peeling
4853 for the number of iterations will be needed for fully-masked loops.
4854 (vectorizable_reduction): Disable fully-masked loops.
4855 (vectorizable_live_operation): Likewise.
4856 (vect_halve_mask_nunits): New function.
4857 (vect_double_mask_nunits): Likewise.
4858 (vect_record_loop_mask): Likewise.
4859 (vect_get_loop_mask): Likewise.
4860 (vect_transform_loop): Handle the case in which the final loop
4861 iteration might handle a partial vector. Call vect_set_loop_condition
4862 instead of slpeel_make_loop_iterate_ntimes.
4863 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4864 (check_load_store_masking): New function.
4865 (prepare_load_store_mask): Likewise.
4866 (vectorizable_store): Handle fully-masked loops.
4867 (vectorizable_load): Likewise.
4868 (supportable_widening_operation): Use vect_halve_mask_nunits for
4870 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4871 (vect_gen_while): New function.
4872 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4873 (aarch64_uqdec<mode>): New insn.
4875 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4876 Alan Hayward <alan.hayward@arm.com>
4877 David Sherwood <david.sherwood@arm.com>
4879 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4880 (reduc_xor_scal_optab): New optabs.
4881 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4882 (reduc_xor_scal_@var{m}): Document.
4883 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4884 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4886 * fold-const-call.c (fold_const_call): Handle them.
4887 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4888 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4889 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4890 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4891 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4892 (UNSPEC_XORV): New unspecs.
4893 (optab): Add entries for them.
4894 (BITWISEV): New int iterator.
4895 (bit_reduc_op): New int attributes.
4897 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4898 Alan Hayward <alan.hayward@arm.com>
4899 David Sherwood <david.sherwood@arm.com>
4901 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4902 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4903 * optabs.def (vec_shl_insert_optab): New optab.
4904 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4905 (duplicate_and_interleave): Likewise.
4906 * tree-vect-loop.c: Include internal-fn.h.
4907 (neutral_op_for_slp_reduction): New function, split out from
4908 get_initial_defs_for_reduction.
4909 (get_initial_def_for_reduction): Handle option 2 for variable-length
4910 vectors by loading the neutral value into a vector and then shifting
4911 the initial value into element 0.
4912 (get_initial_defs_for_reduction): Replace the code argument with
4913 the neutral value calculated by neutral_op_for_slp_reduction.
4914 Use gimple_build_vector for constant-length vectors.
4915 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4916 but the first group_size elements have a neutral value.
4917 Use duplicate_and_interleave otherwise.
4918 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4919 Update call to get_initial_defs_for_reduction. Handle SLP
4920 reductions for variable-length vectors by creating one vector
4921 result for each scalar result, with the elements associated
4922 with other scalar results stubbed out with the neutral value.
4923 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4924 Require IFN_VEC_SHL_INSERT for double reductions on
4925 variable-length vectors, or SLP reductions that have
4926 a neutral value. Require can_duplicate_and_interleave_p
4927 support for variable-length unchained SLP reductions if there
4928 is no neutral value, such as for MIN/MAX reductions. Also require
4929 the number of vector elements to be a multiple of the number of
4930 SLP statements when doing variable-length unchained SLP reductions.
4931 Update call to vect_create_epilog_for_reduction.
4932 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4933 and remove initial values.
4934 (duplicate_and_interleave): Make public.
4935 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4936 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
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 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4943 (can_duplicate_and_interleave_p): New function.
4944 (vect_get_and_check_slp_defs): Take the vector of statements
4945 rather than just the current one. Remove excess parentheses.
4946 Restriction rejectinon of vect_constant_def and vect_external_def
4947 for variable-length vectors to boolean types, or types for which
4948 can_duplicate_and_interleave_p is false.
4949 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4950 (duplicate_and_interleave): New function.
4951 (vect_get_constant_vectors): Use gimple_build_vector for
4952 constant-length vectors and suitable variable-length constant
4953 vectors. Use duplicate_and_interleave for other variable-length
4954 vectors. Don't defer the update when inserting new statements.
4956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4957 Alan Hayward <alan.hayward@arm.com>
4958 David Sherwood <david.sherwood@arm.com>
4960 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4961 min_profitable_iters doesn't go negative.
4963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4964 Alan Hayward <alan.hayward@arm.com>
4965 David Sherwood <david.sherwood@arm.com>
4967 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4968 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4969 * optabs.def (vec_mask_load_lanes_optab): New optab.
4970 (vec_mask_store_lanes_optab): Likewise.
4971 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4972 (MASK_STORE_LANES): Likewise.
4973 * internal-fn.c (mask_load_lanes_direct): New macro.
4974 (mask_store_lanes_direct): Likewise.
4975 (expand_mask_load_optab_fn): Handle masked operations.
4976 (expand_mask_load_lanes_optab_fn): New macro.
4977 (expand_mask_store_optab_fn): Handle masked operations.
4978 (expand_mask_store_lanes_optab_fn): New macro.
4979 (direct_mask_load_lanes_optab_supported_p): Likewise.
4980 (direct_mask_store_lanes_optab_supported_p): Likewise.
4981 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4983 (vect_load_lanes_supported): Likewise.
4984 * tree-vect-data-refs.c (strip_conversion): New function.
4985 (can_group_stmts_p): Likewise.
4986 (vect_analyze_data_ref_accesses): Use it instead of checking
4987 for a pair of assignments.
4988 (vect_store_lanes_supported): Take a masked_p parameter.
4989 (vect_load_lanes_supported): Likewise.
4990 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4991 vect_store_lanes_supported and vect_load_lanes_supported.
4992 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4993 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4994 parameter. Don't allow gaps for masked accesses.
4995 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4996 and vect_load_lanes_supported.
4997 (get_load_store_type): Take a masked_p parameter and update
4998 call to get_group_load_store_type.
4999 (vectorizable_store): Update call to get_load_store_type.
5000 Handle IFN_MASK_STORE_LANES.
5001 (vectorizable_load): Update call to get_load_store_type.
5002 Handle IFN_MASK_LOAD_LANES.
5004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5005 Alan Hayward <alan.hayward@arm.com>
5006 David Sherwood <david.sherwood@arm.com>
5008 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5010 * config/aarch64/aarch64-protos.h
5011 (aarch64_sve_struct_memory_operand_p): Declare.
5012 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5013 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5014 (VPRED, vpred): Handle SVE structure modes.
5015 * config/aarch64/constraints.md (Utx): New constraint.
5016 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5017 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5018 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5019 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5020 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5021 structure modes. Split into pieces after RA.
5022 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5023 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5025 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5026 SVE structure modes.
5027 (aarch64_classify_address): Likewise.
5028 (sizetochar): Move earlier in file.
5029 (aarch64_print_operand): Handle SVE register lists.
5030 (aarch64_array_mode): New function.
5031 (aarch64_sve_struct_memory_operand_p): Likewise.
5032 (TARGET_ARRAY_MODE): Redefine.
5034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5035 Alan Hayward <alan.hayward@arm.com>
5036 David Sherwood <david.sherwood@arm.com>
5038 * target.def (array_mode): New target hook.
5039 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5040 * doc/tm.texi: Regenerate.
5041 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5042 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5043 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5045 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5048 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5049 Alan Hayward <alan.hayward@arm.com>
5050 David Sherwood <david.sherwood@arm.com>
5052 * fold-const.c (fold_binary_loc): Check the argument types
5053 rather than the result type when testing for a vector operation.
5055 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5057 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5058 * doc/tm.texi: Regenerate.
5060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5061 Alan Hayward <alan.hayward@arm.com>
5062 David Sherwood <david.sherwood@arm.com>
5064 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5065 (sve): Document new AArch64 extension.
5066 * doc/md.texi (w): Extend the description of the AArch64
5067 constraint to include SVE vectors.
5068 (Upl, Upa): Document new AArch64 predicate constraints.
5069 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5071 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5072 (msve-vector-bits=): New option.
5073 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5074 SVE when these are disabled.
5075 (sve): New extension.
5076 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5077 modes. Adjust their number of units based on aarch64_sve_vg.
5078 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5079 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5080 aarch64_addr_query_type.
5081 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5082 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5083 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5084 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5085 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5086 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5087 (aarch64_simd_imm_zero_p): Delete.
5088 (aarch64_check_zero_based_sve_index_immediate): Declare.
5089 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5090 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5091 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5092 (aarch64_sve_float_mul_immediate_p): Likewise.
5093 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5095 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5096 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5097 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5098 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5099 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5100 (aarch64_regmode_natural_size): Likewise.
5101 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5102 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5104 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5105 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5106 for VG and the SVE predicate registers.
5107 (V_ALIASES): Add a "z"-prefixed alias.
5108 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5109 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5110 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5111 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5112 (REG_CLASS_NAMES): Add entries for them.
5113 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5114 and the predicate registers.
5115 (aarch64_sve_vg): Declare.
5116 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5117 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5118 (REGMODE_NATURAL_SIZE): Define.
5119 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5121 * config/aarch64/aarch64.c: Include cfgrtl.h.
5122 (simd_immediate_info): Add a constructor for series vectors,
5123 and an associated step field.
5124 (aarch64_sve_vg): New variable.
5125 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5126 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5127 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5128 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5129 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5130 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5131 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5132 (aarch64_get_mask_mode): New functions.
5133 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5134 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5135 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5136 predicate modes and predicate registers. Explicitly restrict
5137 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5138 to store a vector mode if it is recognized by
5139 aarch64_classify_vector_mode.
5140 (aarch64_regmode_natural_size): New function.
5141 (aarch64_hard_regno_caller_save_mode): Return the original mode
5143 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5144 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5145 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5146 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5148 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5149 does not overlap dest if the function is frame-related. Handle
5151 (aarch64_split_add_offset): New function.
5152 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5153 them aarch64_add_offset.
5154 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5155 and update call to aarch64_sub_sp.
5156 (aarch64_add_cfa_expression): New function.
5157 (aarch64_expand_prologue): Pass extra temporary registers to the
5158 functions above. Handle the case in which we need to emit new
5159 DW_CFA_expressions for registers that were originally saved
5160 relative to the stack pointer, but now have to be expressed
5161 relative to the frame pointer.
5162 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5164 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5165 IP0 and IP1 values for SVE frames.
5166 (aarch64_expand_vec_series): New function.
5167 (aarch64_expand_sve_widened_duplicate): Likewise.
5168 (aarch64_expand_sve_const_vector): Likewise.
5169 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5170 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5171 into the register, rather than emitting a SET directly.
5172 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5173 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5174 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5175 (offset_9bit_signed_scaled_p): New functions.
5176 (aarch64_replicate_bitmask_imm): New function.
5177 (aarch64_bitmask_imm): Use it.
5178 (aarch64_cannot_force_const_mem): Reject expressions involving
5179 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5180 (aarch64_classify_index): Handle SVE indices, by requiring
5181 a plain register index with a scale that matches the element size.
5182 (aarch64_classify_address): Handle SVE addresses. Assert that
5183 the mode of the address is VOIDmode or an integer mode.
5184 Update call to aarch64_classify_symbol.
5185 (aarch64_classify_symbolic_expression): Update call to
5186 aarch64_classify_symbol.
5187 (aarch64_const_vec_all_in_range_p): New function.
5188 (aarch64_print_vector_float_operand): Likewise.
5189 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5190 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5191 and the FP immediates 1.0 and 0.5.
5192 (aarch64_print_address_internal): Handle SVE addresses.
5193 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5194 (aarch64_regno_regclass): Handle predicate registers.
5195 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5197 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5198 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5199 (aarch64_convert_sve_vector_bits): New function.
5200 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5201 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5203 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5204 Handle SVE vector and predicate modes. Accept VL-based constants
5205 that need only one temporary register, and VL offsets that require
5206 no temporary registers.
5207 (aarch64_conditional_register_usage): Mark the predicate registers
5208 as fixed if SVE isn't available.
5209 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5210 Return true for SVE vector and predicate modes.
5211 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5212 rather than an unsigned int. Handle SVE modes.
5213 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5215 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5217 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5218 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5219 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5220 (aarch64_sve_float_mul_immediate_p): New functions.
5221 (aarch64_sve_valid_immediate): New function.
5222 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5223 Explicitly reject structure modes. Check for INDEX constants.
5224 Handle PTRUE and PFALSE constants.
5225 (aarch64_check_zero_based_sve_index_immediate): New function.
5226 (aarch64_simd_imm_zero_p): Delete.
5227 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5228 vector modes. Accept constants in the range of CNT[BHWD].
5229 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5230 ask for an Advanced SIMD mode.
5231 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5232 (aarch64_simd_vector_alignment): Handle SVE predicates.
5233 (aarch64_vectorize_preferred_vector_alignment): New function.
5234 (aarch64_simd_vector_alignment_reachable): Use it instead of
5236 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5237 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5239 (MAX_VECT_LEN): Delete.
5240 (expand_vec_perm_d): Add a vec_flags field.
5241 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5242 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5243 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5245 (aarch64_evpc_rev): Rename to...
5246 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5247 (aarch64_evpc_rev_global): New function.
5248 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5249 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5251 (aarch64_evpc_sve_tbl): New function.
5252 (aarch64_expand_vec_perm_const_1): Update after rename of
5253 aarch64_evpc_rev. Handle SVE permutes too, trying
5254 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5255 than aarch64_evpc_tbl.
5256 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5257 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5258 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5259 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5260 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5261 (aarch64_expand_sve_vcond): New functions.
5262 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5263 of aarch64_vector_mode_p.
5264 (aarch64_dwarf_poly_indeterminate_value): New function.
5265 (aarch64_compute_pressure_classes): Likewise.
5266 (aarch64_can_change_mode_class): Likewise.
5267 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5268 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5269 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5270 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5271 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5272 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5273 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5274 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5276 (Dn, Dl, Dr): Accept const as well as const_vector.
5277 (Dz): Likewise. Compare against CONST0_RTX.
5278 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5279 of "vector" where appropriate.
5280 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5281 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5282 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5283 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5284 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5285 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5286 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5287 (v_int_equiv): Extend to SVE modes.
5288 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5290 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5291 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5292 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5293 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5294 (SVE_COND_FP_CMP): New int iterators.
5295 (perm_hilo): Handle the new unpack unspecs.
5296 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5298 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5299 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5300 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5301 (aarch64_equality_operator, aarch64_constant_vector_operand)
5302 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5303 (aarch64_sve_nonimmediate_operand): Likewise.
5304 (aarch64_sve_general_operand): Likewise.
5305 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5306 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5307 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5308 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5309 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5310 (aarch64_sve_float_arith_immediate): Likewise.
5311 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5312 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5313 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5314 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5315 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5316 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5317 (aarch64_sve_float_arith_operand): Likewise.
5318 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5319 (aarch64_sve_float_mul_operand): Likewise.
5320 (aarch64_sve_vec_perm_operand): Likewise.
5321 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5322 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5323 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5324 as well as const_vector.
5325 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5326 in file. Use CONST0_RTX and CONSTM1_RTX.
5327 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5328 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5329 Use aarch64_simd_imm_zero.
5330 * config/aarch64/aarch64-sve.md: New file.
5331 * config/aarch64/aarch64.md: Include it.
5332 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5333 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5334 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5335 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5336 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5337 (sve): New attribute.
5338 (enabled): Disable instructions with the sve attribute unless
5340 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5341 aarch64_expand_mov_immediate.
5342 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5343 CNT[BHSD] immediates.
5344 (movti): Split CONST_POLY_INT moves into two halves.
5345 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5346 Split additions that need a temporary here if the destination
5347 is the stack pointer.
5348 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5349 (*add<mode>3_poly_1): New instruction.
5350 (set_clobber_cc): New expander.
5352 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5354 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5355 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5356 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5357 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5358 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5359 Change innermode from fixed_mode_size to machine_mode.
5360 (simplify_subreg): Update call accordingly. Handle a constant-sized
5361 subreg of a variable-length CONST_VECTOR.
5363 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5364 Alan Hayward <alan.hayward@arm.com>
5365 David Sherwood <david.sherwood@arm.com>
5367 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5368 (add_offset_to_base): New function, split out from...
5369 (create_mem_ref): ...here. When handling a scale other than 1,
5370 check first whether the address is valid without the offset.
5371 Add it into the base if so, leaving the index and scale as-is.
5373 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5376 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5377 fold_for_warn before checking if arg2 is INTEGER_CST.
5379 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5381 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5382 (store_multiple_operation): Delete.
5383 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5384 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5385 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5386 guarded by TARGET_STRING.
5387 (rs6000_output_load_multiple): Delete.
5388 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5389 OPTION_MASK_STRING / TARGET_STRING handling.
5390 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5391 (const rs6000_opt_masks) <"string">: Change mask to 0.
5392 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5393 (MASK_STRING): Delete.
5394 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5396 (load_multiple): Delete.
5403 (store_multiple): Delete.
5410 (movmemsi_8reg): Delete.
5411 (corresponding unnamed define_insn): Delete.
5412 (movmemsi_6reg): Delete.
5413 (corresponding unnamed define_insn): Delete.
5414 (movmemsi_4reg): Delete.
5415 (corresponding unnamed define_insn): Delete.
5416 (movmemsi_2reg): Delete.
5417 (corresponding unnamed define_insn): Delete.
5418 (movmemsi_1reg): Delete.
5419 (corresponding unnamed define_insn): Delete.
5420 * config/rs6000/rs6000.opt (mno-string): New.
5421 (mstring): Replace by deprecation warning stub.
5422 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5424 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5426 * regrename.c (regrename_do_replace): If replacing the same
5427 reg multiple times, try to reuse last created gen_raw_REG.
5430 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5431 main to workaround a bug in GDB.
5433 2018-01-12 Tom de Vries <tom@codesourcery.com>
5436 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5438 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5440 PR rtl-optimization/80481
5441 * ira-color.c (get_cap_member): New function.
5442 (allocnos_conflict_by_live_ranges_p): Use it.
5443 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5444 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5446 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5449 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5450 (*saddl_se_1): Ditto.
5452 (*ssubl_se_1): Ditto.
5454 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5456 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5457 rather than wi::to_widest for DR_INITs.
5458 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5459 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5460 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5462 (vect_analyze_group_access_1): Note that here.
5464 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5466 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5467 polynomial type sizes.
5469 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5471 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5472 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5473 (gimple_add_tmp_var): Likewise.
5475 2018-01-12 Martin Liska <mliska@suse.cz>
5477 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5478 (gimple_alloc_sizes): Likewise.
5479 (dump_gimple_statistics): Use PRIu64 in printf format.
5480 * gimple.h: Change uint64_t to int.
5482 2018-01-12 Martin Liska <mliska@suse.cz>
5484 * tree-core.h: Use uint64_t instead of int.
5485 * tree.c (tree_node_counts): Likewise.
5486 (tree_node_sizes): Likewise.
5487 (dump_tree_statistics): Use PRIu64 in printf format.
5489 2018-01-12 Martin Liska <mliska@suse.cz>
5491 * Makefile.in: As qsort_chk is implemented in vec.c, add
5492 vec.o to linkage of gencfn-macros.
5493 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5494 passing the info to record_node_allocation_statistics.
5495 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5497 * ggc-common.c (struct ggc_usage): Add operator== and use
5498 it in operator< and compare function.
5499 * mem-stats.h (struct mem_usage): Likewise.
5500 * vec.c (struct vec_usage): Remove operator< and compare
5501 function. Can be simply inherited.
5503 2018-01-12 Martin Jambor <mjambor@suse.cz>
5506 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5507 * tree-ssa-math-opts.c: Include domwalk.h.
5508 (convert_mult_to_fma_1): New function.
5509 (fma_transformation_info): New type.
5510 (fma_deferring_state): Likewise.
5511 (cancel_fma_deferring): New function.
5512 (result_of_phi): Likewise.
5513 (last_fma_candidate_feeds_initial_phi): Likewise.
5514 (convert_mult_to_fma): Added deferring logic, split actual
5515 transformation to convert_mult_to_fma_1.
5516 (math_opts_dom_walker): New type.
5517 (math_opts_dom_walker::after_dom_children): New method, body moved
5518 here from pass_optimize_widening_mul::execute, added deferring logic
5520 (pass_optimize_widening_mul::execute): Moved most of code to
5521 math_opts_dom_walker::after_dom_children.
5522 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5523 * config/i386/i386.c (ix86_option_override_internal): Added
5524 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5526 2018-01-12 Richard Biener <rguenther@suse.de>
5529 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5530 inline instance vars.
5532 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5535 * config/rx/rx.c (rx_is_restricted_memory_address):
5538 2018-01-12 Richard Biener <rguenther@suse.de>
5540 PR tree-optimization/80846
5541 * target.def (split_reduction): New target hook.
5542 * targhooks.c (default_split_reduction): New function.
5543 * targhooks.h (default_split_reduction): Declare.
5544 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5545 target requests first reduce vectors by combining low and high
5547 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5548 (get_vectype_for_scalar_type_and_size): Export.
5549 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5550 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5551 * doc/tm.texi: Regenerate.
5552 * config/i386/i386.c (ix86_split_reduction): Implement
5553 TARGET_VECTORIZE_SPLIT_REDUCTION.
5555 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5558 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5559 in PIC mode except for TARGET_VXWORKS_RTP.
5560 * config/sparc/sparc.c: Include cfgrtl.h.
5561 (TARGET_INIT_PIC_REG): Define.
5562 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5563 (sparc_pic_register_p): New predicate.
5564 (sparc_legitimate_address_p): Use it.
5565 (sparc_legitimize_pic_address): Likewise.
5566 (sparc_delegitimize_address): Likewise.
5567 (sparc_mode_dependent_address_p): Likewise.
5568 (gen_load_pcrel_sym): Remove 4th parameter.
5569 (load_got_register): Adjust call to above. Remove obsolete stuff.
5570 (sparc_expand_prologue): Do not call load_got_register here.
5571 (sparc_flat_expand_prologue): Likewise.
5572 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5573 (sparc_use_pseudo_pic_reg): New function.
5574 (sparc_init_pic_reg): Likewise.
5575 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5576 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5578 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5580 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5581 Add item for branch_cost.
5583 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5585 PR rtl-optimization/83565
5586 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5587 not extend the result to a larger mode for rotate operations.
5588 (num_sign_bit_copies1): Likewise.
5590 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5593 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5595 Use values-Xc.o for -pedantic.
5596 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5598 2018-01-12 Martin Liska <mliska@suse.cz>
5601 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5603 (possible_polymorphic_call_targets): Use it.
5604 (ipa_devirt): Likewise.
5606 2018-01-12 Martin Liska <mliska@suse.cz>
5608 * profile-count.h (enum profile_quality): Use 0 as invalid
5609 enum value of profile_quality.
5611 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5613 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5614 -mext-string options.
5616 2018-01-12 Richard Biener <rguenther@suse.de>
5618 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5619 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5620 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5622 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5624 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5626 * configure.ac (--with-long-double-format): Add support for the
5627 configuration option to change the default long double format on
5629 * config.gcc (powerpc*-linux*-*): Likewise.
5630 * configure: Regenerate.
5631 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5632 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5633 used without modification.
5635 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5637 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5638 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5639 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5640 MISC_BUILTIN_SPEC_BARRIER.
5641 (rs6000_init_builtins): Likewise.
5642 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5644 (speculation_barrier): New define_insn.
5645 * doc/extend.texi: Document __builtin_speculation_barrier.
5647 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5650 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5651 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5652 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5654 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5655 integral modes instead of "ss" and "sd".
5656 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5657 vectors with 32-bit and 64-bit elements.
5658 (vecdupssescalarmodesuffix): New mode attribute.
5659 (vec_dup<mode>): Use it.
5661 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5664 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5665 frame if argument is passed on stack.
5667 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5670 * ree.c (combine_reaching_defs): Optimize also
5671 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5672 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5674 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5677 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5679 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5682 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5683 after they are computed.
5685 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5687 PR tree-optimization/83695
5688 * gimple-loop-linterchange.cc
5689 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5690 reset cached scev information after interchange.
5691 (pass_linterchange::execute): Remove call to scev_reset_htab.
5693 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5695 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5696 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5697 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5698 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5699 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5700 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5701 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5702 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5703 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5704 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5705 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5706 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5707 (V_lane_reg): Likewise.
5708 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5710 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5711 (vfmal_lane_low<mode>_intrinsic,
5712 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5713 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5714 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5715 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5716 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5717 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5719 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5721 * config/arm/arm-cpus.in (fp16fml): New feature.
5722 (ALL_SIMD): Add fp16fml.
5723 (armv8.2-a): Add fp16fml as an option.
5724 (armv8.3-a): Likewise.
5725 (armv8.4-a): Add fp16fml as part of fp16.
5726 * config/arm/arm.h (TARGET_FP16FML): Define.
5727 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5729 * config/arm/arm-modes.def (V2HF): Define.
5730 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5731 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5732 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5733 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5734 vfmsl_low, vfmsl_high): New set of builtins.
5735 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5736 (vfml_op): New code attribute.
5737 (VFMLHALVES): New int iterator.
5738 (VFML, VFMLSEL): New mode attributes.
5739 (V_reg): Define mapping for V2HF.
5740 (V_hi, V_lo): New mode attributes.
5741 (VF_constraint): Likewise.
5742 (vfml_half, vfml_half_selector): New int attributes.
5743 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5745 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5746 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5748 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5749 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5750 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5751 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5753 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5754 Document new effective target and option set.
5756 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5758 * config/arm/arm-cpus.in (armv8_4): New feature.
5759 (ARMv8_4a): New fgroup.
5760 (armv8.4-a): New arch.
5761 * config/arm/arm-tables.opt: Regenerate.
5762 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5763 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5764 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5765 Add matching rules for -march=armv8.4-a and extensions.
5766 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5768 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5771 * config/rx/rx.md (BW): New mode attribute.
5772 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5774 2018-01-11 Richard Biener <rguenther@suse.de>
5776 PR tree-optimization/83435
5777 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5778 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5779 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5781 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5782 Alan Hayward <alan.hayward@arm.com>
5783 David Sherwood <david.sherwood@arm.com>
5785 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5787 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5788 (aarch64_print_address_internal): Use it to check for a zero offset.
5790 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5791 Alan Hayward <alan.hayward@arm.com>
5792 David Sherwood <david.sherwood@arm.com>
5794 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5795 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5796 Return a poly_int64 rather than a HOST_WIDE_INT.
5797 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5798 rather than a HOST_WIDE_INT.
5799 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5800 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5801 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5802 final_offset from HOST_WIDE_INT to poly_int64.
5803 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5804 to_constant when getting the number of units in an Advanced SIMD
5806 (aarch64_builtin_vectorized_function): Check for a constant number
5808 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5810 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5811 attribute instead of GET_MODE_NUNITS.
5812 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5813 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5814 GET_MODE_SIZE for fixed-size registers.
5815 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5816 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5817 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5818 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5819 (aarch64_print_operand, aarch64_print_address_internal)
5820 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5821 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5822 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5823 Handle polynomial GET_MODE_SIZE.
5824 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5825 wider than SImode without modification.
5826 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5827 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5828 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5829 passing and returning SVE modes.
5830 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5831 rather than GEN_INT.
5832 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5833 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5834 (aarch64_allocate_and_probe_stack_space): Likewise.
5835 (aarch64_layout_frame): Cope with polynomial offsets.
5836 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5837 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5839 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5840 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5841 poly_int64 rather than a HOST_WIDE_INT.
5842 (aarch64_get_separate_components, aarch64_process_components)
5843 (aarch64_expand_prologue, aarch64_expand_epilogue)
5844 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5845 (aarch64_anchor_offset): New function, split out from...
5846 (aarch64_legitimize_address): ...here.
5847 (aarch64_builtin_vectorization_cost): Handle polynomial
5848 TYPE_VECTOR_SUBPARTS.
5849 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5851 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5852 number of elements from the PARALLEL rather than the mode.
5853 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5854 rather than GET_MODE_BITSIZE.
5855 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5856 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5857 (aarch64_expand_vec_perm_const_1): Handle polynomial
5858 d->perm.length () and d->perm elements.
5859 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5860 Apply to_constant to d->perm elements.
5861 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5862 polynomial CONST_VECTOR_NUNITS.
5863 (aarch64_move_pointer): Take amount as a poly_int64 rather
5865 (aarch64_progress_pointer): Avoid temporary variable.
5866 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5867 the mode attribute instead of GET_MODE.
5869 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5870 Alan Hayward <alan.hayward@arm.com>
5871 David Sherwood <david.sherwood@arm.com>
5873 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5874 x exists before using it.
5875 (aarch64_add_constant_internal): Rename to...
5876 (aarch64_add_offset_1): ...this. Replace regnum with separate
5877 src and dest rtxes. Handle the case in which they're different,
5878 including when the offset is zero. Replace scratchreg with an rtx.
5879 Use 2 additions if there is no spare register into which we can
5880 move a 16-bit constant.
5881 (aarch64_add_constant): Delete.
5882 (aarch64_add_offset): Replace reg with separate src and dest
5883 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5884 Use aarch64_add_offset_1.
5885 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5886 an rtx rather than an int. Take the delta as a poly_int64
5887 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5888 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5889 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5890 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5891 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5893 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5894 aarch64_add_constant.
5896 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5898 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5899 Use scalar_float_mode.
5901 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5903 * config/aarch64/aarch64-simd.md
5904 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5905 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5906 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5907 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5908 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5909 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5910 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5911 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5912 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5913 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5915 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5918 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5919 targ_options->x_arm_arch_string is non NULL.
5921 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5923 * config/aarch64/aarch64.h
5924 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5926 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5929 * expmed.c (emit_store_flag_force): Swap if const op0
5930 and change VOIDmode to mode of op0.
5932 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5934 PR rtl-optimization/83761
5935 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5936 than bytes to mode_for_size.
5938 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5941 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5942 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5945 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5948 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5949 when in layout mode.
5950 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5951 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5954 2018-01-10 Michael Collison <michael.collison@arm.com>
5956 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5957 * config/aarch64/aarch64-option-extension.def: Add
5958 AARCH64_OPT_EXTENSION of 'fp16fml'.
5959 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5960 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5961 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5962 * config/aarch64/constraints.md (Ui7): New constraint.
5963 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5964 (VFMLA_SEL_W): Ditto.
5967 (VFMLA16_LOW): New int iterator.
5968 (VFMLA16_HIGH): Ditto.
5969 (UNSPEC_FMLAL): New unspec.
5970 (UNSPEC_FMLSL): Ditto.
5971 (UNSPEC_FMLAL2): Ditto.
5972 (UNSPEC_FMLSL2): Ditto.
5973 (f16mac): New code attribute.
5974 * config/aarch64/aarch64-simd-builtins.def
5975 (aarch64_fmlal_lowv2sf): Ditto.
5976 (aarch64_fmlsl_lowv2sf): Ditto.
5977 (aarch64_fmlalq_lowv4sf): Ditto.
5978 (aarch64_fmlslq_lowv4sf): Ditto.
5979 (aarch64_fmlal_highv2sf): Ditto.
5980 (aarch64_fmlsl_highv2sf): Ditto.
5981 (aarch64_fmlalq_highv4sf): Ditto.
5982 (aarch64_fmlslq_highv4sf): Ditto.
5983 (aarch64_fmlal_lane_lowv2sf): Ditto.
5984 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5985 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5986 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5987 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5988 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5989 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5990 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5991 (aarch64_fmlal_lane_highv2sf): Ditto.
5992 (aarch64_fmlsl_lane_highv2sf): Ditto.
5993 (aarch64_fmlal_laneq_highv2sf): Ditto.
5994 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5995 (aarch64_fmlalq_lane_highv4sf): Ditto.
5996 (aarch64_fmlsl_lane_highv4sf): Ditto.
5997 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5998 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5999 * config/aarch64/aarch64-simd.md:
6000 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6001 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6002 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6003 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6004 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6005 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6006 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6007 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6008 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6009 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6010 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6011 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6012 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6013 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6014 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6015 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6016 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6017 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6018 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6019 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6020 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6021 (vfmlsl_low_u32): Ditto.
6022 (vfmlalq_low_u32): Ditto.
6023 (vfmlslq_low_u32): Ditto.
6024 (vfmlal_high_u32): Ditto.
6025 (vfmlsl_high_u32): Ditto.
6026 (vfmlalq_high_u32): Ditto.
6027 (vfmlslq_high_u32): Ditto.
6028 (vfmlal_lane_low_u32): Ditto.
6029 (vfmlsl_lane_low_u32): Ditto.
6030 (vfmlal_laneq_low_u32): Ditto.
6031 (vfmlsl_laneq_low_u32): Ditto.
6032 (vfmlalq_lane_low_u32): Ditto.
6033 (vfmlslq_lane_low_u32): Ditto.
6034 (vfmlalq_laneq_low_u32): Ditto.
6035 (vfmlslq_laneq_low_u32): Ditto.
6036 (vfmlal_lane_high_u32): Ditto.
6037 (vfmlsl_lane_high_u32): Ditto.
6038 (vfmlal_laneq_high_u32): Ditto.
6039 (vfmlsl_laneq_high_u32): Ditto.
6040 (vfmlalq_lane_high_u32): Ditto.
6041 (vfmlslq_lane_high_u32): Ditto.
6042 (vfmlalq_laneq_high_u32): Ditto.
6043 (vfmlslq_laneq_high_u32): Ditto.
6044 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6045 (AARCH64_FL_FOR_ARCH8_4): New.
6046 (AARCH64_ISA_F16FML): New ISA flag.
6047 (TARGET_F16FML): New feature flag for fp16fml.
6048 (doc/invoke.texi): Document new fp16fml option.
6050 2018-01-10 Michael Collison <michael.collison@arm.com>
6052 * config/aarch64/aarch64-builtins.c:
6053 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6054 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6055 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6056 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6057 (AARCH64_ISA_SHA3): New ISA flag.
6058 (TARGET_SHA3): New feature flag for sha3.
6059 * config/aarch64/iterators.md (sha512_op): New int attribute.
6060 (CRYPTO_SHA512): New int iterator.
6061 (UNSPEC_SHA512H): New unspec.
6062 (UNSPEC_SHA512H2): Ditto.
6063 (UNSPEC_SHA512SU0): Ditto.
6064 (UNSPEC_SHA512SU1): Ditto.
6065 * config/aarch64/aarch64-simd-builtins.def
6066 (aarch64_crypto_sha512hqv2di): New builtin.
6067 (aarch64_crypto_sha512h2qv2di): Ditto.
6068 (aarch64_crypto_sha512su0qv2di): Ditto.
6069 (aarch64_crypto_sha512su1qv2di): Ditto.
6070 (aarch64_eor3qv8hi): Ditto.
6071 (aarch64_rax1qv2di): Ditto.
6072 (aarch64_xarqv2di): Ditto.
6073 (aarch64_bcaxqv8hi): Ditto.
6074 * config/aarch64/aarch64-simd.md:
6075 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6076 (aarch64_crypto_sha512su0qv2di): Ditto.
6077 (aarch64_crypto_sha512su1qv2di): Ditto.
6078 (aarch64_eor3qv8hi): Ditto.
6079 (aarch64_rax1qv2di): Ditto.
6080 (aarch64_xarqv2di): Ditto.
6081 (aarch64_bcaxqv8hi): Ditto.
6082 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6083 (vsha512h2q_u64): Ditto.
6084 (vsha512su0q_u64): Ditto.
6085 (vsha512su1q_u64): Ditto.
6086 (veor3q_u16): Ditto.
6087 (vrax1q_u64): Ditto.
6089 (vbcaxq_u16): Ditto.
6090 * config/arm/types.md (crypto_sha512): New type attribute.
6091 (crypto_sha3): Ditto.
6092 (doc/invoke.texi): Document new sha3 option.
6094 2018-01-10 Michael Collison <michael.collison@arm.com>
6096 * config/aarch64/aarch64-builtins.c:
6097 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6098 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6099 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6100 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6101 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6102 (AARCH64_ISA_SM4): New ISA flag.
6103 (TARGET_SM4): New feature flag for sm4.
6104 * config/aarch64/aarch64-simd-builtins.def
6105 (aarch64_sm3ss1qv4si): Ditto.
6106 (aarch64_sm3tt1aq4si): Ditto.
6107 (aarch64_sm3tt1bq4si): Ditto.
6108 (aarch64_sm3tt2aq4si): Ditto.
6109 (aarch64_sm3tt2bq4si): Ditto.
6110 (aarch64_sm3partw1qv4si): Ditto.
6111 (aarch64_sm3partw2qv4si): Ditto.
6112 (aarch64_sm4eqv4si): Ditto.
6113 (aarch64_sm4ekeyqv4si): Ditto.
6114 * config/aarch64/aarch64-simd.md:
6115 (aarch64_sm3ss1qv4si): Ditto.
6116 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6117 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6118 (aarch64_sm4eqv4si): Ditto.
6119 (aarch64_sm4ekeyqv4si): Ditto.
6120 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6121 (sm3part_op): Ditto.
6122 (CRYPTO_SM3TT): Ditto.
6123 (CRYPTO_SM3PART): Ditto.
6124 (UNSPEC_SM3SS1): New unspec.
6125 (UNSPEC_SM3TT1A): Ditto.
6126 (UNSPEC_SM3TT1B): Ditto.
6127 (UNSPEC_SM3TT2A): Ditto.
6128 (UNSPEC_SM3TT2B): Ditto.
6129 (UNSPEC_SM3PARTW1): Ditto.
6130 (UNSPEC_SM3PARTW2): Ditto.
6131 (UNSPEC_SM4E): Ditto.
6132 (UNSPEC_SM4EKEY): Ditto.
6133 * config/aarch64/constraints.md (Ui2): New constraint.
6134 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6135 * config/arm/types.md (crypto_sm3): New type attribute.
6136 (crypto_sm4): Ditto.
6137 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6138 (vsm3tt1aq_u32): Ditto.
6139 (vsm3tt1bq_u32): Ditto.
6140 (vsm3tt2aq_u32): Ditto.
6141 (vsm3tt2bq_u32): Ditto.
6142 (vsm3partw1q_u32): Ditto.
6143 (vsm3partw2q_u32): Ditto.
6144 (vsm4eq_u32): Ditto.
6145 (vsm4ekeyq_u32): Ditto.
6146 (doc/invoke.texi): Document new sm4 option.
6148 2018-01-10 Michael Collison <michael.collison@arm.com>
6150 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6151 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6152 (AARCH64_FL_FOR_ARCH8_4): New.
6153 (AARCH64_FL_V8_4): New flag.
6154 (doc/invoke.texi): Document new armv8.4-a option.
6156 2018-01-10 Michael Collison <michael.collison@arm.com>
6158 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6159 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6160 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6161 * config/aarch64/aarch64-option-extension.def: Add
6162 AARCH64_OPT_EXTENSION of 'sha2'.
6163 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6164 (crypto): Disable sha2 and aes if crypto disabled.
6165 (crypto): Enable aes and sha2 if enabled.
6166 (simd): Disable sha2 and aes if simd disabled.
6167 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6169 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6170 (TARGET_SHA2): New feature flag for sha2.
6171 (TARGET_AES): New feature flag for aes.
6172 * config/aarch64/aarch64-simd.md:
6173 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6174 conditional on TARGET_AES.
6175 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6176 (aarch64_crypto_sha1hsi): Make pattern conditional
6178 (aarch64_crypto_sha1hv4si): Ditto.
6179 (aarch64_be_crypto_sha1hv4si): Ditto.
6180 (aarch64_crypto_sha1su1v4si): Ditto.
6181 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6182 (aarch64_crypto_sha1su0v4si): Ditto.
6183 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6184 (aarch64_crypto_sha256su0v4si): Ditto.
6185 (aarch64_crypto_sha256su1v4si): Ditto.
6186 (doc/invoke.texi): Document new aes and sha2 options.
6188 2018-01-10 Martin Sebor <msebor@redhat.com>
6190 PR tree-optimization/83781
6191 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6194 2018-01-11 Martin Sebor <msebor@gmail.com>
6195 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6197 PR tree-optimization/83501
6198 PR tree-optimization/81703
6200 * tree-ssa-strlen.c (get_string_cst): Rename...
6201 (get_string_len): ...to this. Handle global constants.
6202 (handle_char_store): Adjust.
6204 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6205 Jim Wilson <jimw@sifive.com>
6207 * config/riscv/riscv-protos.h (riscv_output_return): New.
6208 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6209 (riscv_attribute_table, riscv_output_return),
6210 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6211 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6212 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6214 (riscv_expand_prologue): Add early return for naked function.
6215 (riscv_expand_epilogue): Likewise.
6216 (riscv_function_ok_for_sibcall): Return false for naked function.
6217 (riscv_set_current_function): New.
6218 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6219 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6220 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6221 * doc/extend.texi (RISC-V Function Attributes): New.
6223 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6225 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6226 check for 128-bit long double before checking TCmode.
6227 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6228 128-bit long doubles before checking TFmode or TCmode.
6229 (FLOAT128_IBM_P): Likewise.
6231 2018-01-10 Martin Sebor <msebor@redhat.com>
6233 PR tree-optimization/83671
6234 * builtins.c (c_strlen): Unconditionally return zero for the empty
6236 Use -Warray-bounds for warnings.
6237 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6238 for non-constant array indices with COMPONENT_REF, arrays of
6239 arrays, and pointers to arrays.
6240 (gimple_fold_builtin_strlen): Determine and set length range for
6241 non-constant character arrays.
6243 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6246 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6249 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6251 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6253 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6256 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6257 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6258 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6259 indexed_or_indirect_operand predicate.
6260 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6261 (*vsx_le_perm_load_v8hi): Likewise.
6262 (*vsx_le_perm_load_v16qi): Likewise.
6263 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6264 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6265 (*vsx_le_perm_store_v8hi): Likewise.
6266 (*vsx_le_perm_store_v16qi): Likewise.
6267 (eight unnamed splitters): Likewise.
6269 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6271 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6272 * config/rs6000/emmintrin.h: Likewise.
6273 * config/rs6000/mmintrin.h: Likewise.
6274 * config/rs6000/xmmintrin.h: Likewise.
6276 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6279 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6281 * tree.c (tree_nop_conversion): Return true for location wrapper
6283 (maybe_wrap_with_location): New function.
6284 (selftest::check_strip_nops): New function.
6285 (selftest::test_location_wrappers): New function.
6286 (selftest::tree_c_tests): Call it.
6287 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6288 (maybe_wrap_with_location): New decl.
6289 (EXPR_LOCATION_WRAPPER_P): New macro.
6290 (location_wrapper_p): New inline function.
6291 (tree_strip_any_location_wrapper): New inline function.
6293 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6296 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6297 stack_realign_offset for the largest alignment of stack slot
6299 (ix86_find_max_used_stack_alignment): New function.
6300 (ix86_finalize_stack_frame_flags): Use it. Set
6301 max_used_stack_alignment if we don't realign stack.
6302 * config/i386/i386.h (machine_function): Add
6303 max_used_stack_alignment.
6305 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6307 * config/arm/arm.opt (-mbranch-cost): New option.
6308 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6311 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6314 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6315 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6317 2018-01-10 Richard Biener <rguenther@suse.de>
6320 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6321 early out so it also covers the case where we have a non-NULL
6324 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6326 PR tree-optimization/83753
6327 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6328 for non-strided grouped accesses if the number of elements is 1.
6330 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6333 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6334 * i386.h (TARGET_USE_GATHER): Define.
6335 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6337 2018-01-10 Martin Liska <mliska@suse.cz>
6340 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6341 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6343 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6344 CLEANUP_NO_PARTITIONING is not set.
6346 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6348 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6349 for vectors, as a partial revert of r254296.
6350 * rtl.h (const_vec_p): Delete.
6351 (const_vec_duplicate_p): Don't test for vector CONSTs.
6352 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6353 * expmed.c (make_tree): Likewise.
6356 * common.md (E, F): Use CONSTANT_P instead of checking for
6358 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6359 checking for CONST_VECTOR.
6361 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6364 * predict.c (force_edge_cold): Handle in more sane way edges
6367 2018-01-09 Carl Love <cel@us.ibm.com>
6369 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6371 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6372 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6373 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6374 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6375 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6376 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6377 * config/rs6000/rs6000-protos.h: Add extern defition for
6378 rs6000_generate_float2_double_code.
6379 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6381 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6382 (float2_v2df): Add define_expand.
6384 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6387 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6388 op_mode in the force_to_mode call.
6390 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6392 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6393 instead of checking each element individually.
6394 (aarch64_evpc_uzp): Likewise.
6395 (aarch64_evpc_zip): Likewise.
6396 (aarch64_evpc_ext): Likewise.
6397 (aarch64_evpc_rev): Likewise.
6398 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6399 instead of checking each element individually. Return true without
6401 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6402 whether all selected elements come from the same input, instead of
6403 checking each element individually. Remove calls to gen_rtx_REG,
6404 start_sequence and end_sequence and instead assert that no rtl is
6407 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6409 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6410 order of HIGH and CONST checks.
6412 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6414 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6415 if the destination isn't an SSA_NAME.
6417 2018-01-09 Richard Biener <rguenther@suse.de>
6419 PR tree-optimization/83668
6420 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6422 (canonicalize_loop_form): ... here, renamed from ...
6423 (canonicalize_loop_closed_ssa_form): ... this and amended to
6424 swap successor edges for loop exit blocks to make us use
6425 the RPO order we need for initial schedule generation.
6427 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6429 PR tree-optimization/64811
6430 * match.pd: When optimizing comparisons with Inf, avoid
6431 introducing or losing exceptions from comparisons with NaN.
6433 2018-01-09 Martin Liska <mliska@suse.cz>
6436 * asan.c (shadow_mem_size): Add gcc_assert.
6438 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6440 Don't save registers in main().
6443 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6444 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6445 * config/avr/avr.c (avr_set_current_function): Don't error if
6446 naked, OS_task or OS_main are specified at the same time.
6447 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6449 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6451 * common/config/avr/avr-common.c (avr_option_optimization_table):
6452 Switch on -mmain-is-OS_task for optimizing compilations.
6454 2018-01-09 Richard Biener <rguenther@suse.de>
6456 PR tree-optimization/83572
6457 * graphite.c: Include cfganal.h.
6458 (graphite_transform_loops): Connect infinite loops to exit
6459 and remove fake edges at the end.
6461 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6463 * ipa-inline.c (edge_badness): Revert accidental checkin.
6465 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6468 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6469 symbols; not inline clones.
6471 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6474 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6475 hard registers. Formatting fixes.
6477 PR preprocessor/83722
6478 * gcc.c (try_generate_repro): Pass
6479 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6480 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6483 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6484 Kito Cheng <kito.cheng@gmail.com>
6486 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6487 (riscv_leaf_function_p): Delete.
6488 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6490 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6492 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6494 (do_ifelse): New function.
6495 (do_isel): New function.
6496 (do_sub3): New function.
6497 (do_add3): New function.
6498 (do_load_mask_compare): New function.
6499 (do_overlap_load_compare): New function.
6500 (expand_compare_loop): New function.
6501 (expand_block_compare): Call expand_compare_loop() when appropriate.
6502 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6504 (-mblock-compare-inline-loop-limit): New option.
6506 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6509 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6510 Reverse order of second and third operands in first alternative.
6511 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6512 of first and second elements in UNSPEC_VPERMR vector.
6513 (altivec_expand_vec_perm_le): Likewise.
6515 2018-01-08 Jeff Law <law@redhat.com>
6517 PR rtl-optimizatin/81308
6518 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6519 (process_switch): If group_case_labels makes a change, then set
6521 (pass_convert_switch::execute): If a switch is converted, then
6522 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6524 PR rtl-optimization/81308
6525 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6528 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6530 PR target/83663 - Revert r255946
6531 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6532 generation for cases where splatting a value is not useful.
6533 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6534 across a vec_duplicate and a paradoxical subreg forming a vector
6535 mode to a vec_concat.
6537 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6539 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6540 -march=armv8.3-a variants.
6541 * config/arm/t-multilib: Likewise.
6542 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6544 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6546 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6548 (cceq_ior_compare_complement): Give it a name so I can use it, and
6549 change boolean_or_operator predicate to boolean_operator so it can
6550 be used to generate a crand.
6551 (eqne): New code iterator.
6552 (bd/bd_neg): New code_attrs.
6553 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6554 a single define_insn.
6555 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6556 decrement (bdnzt/bdnzf/bdzt/bdzf).
6557 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6558 with the new names of the branch decrement patterns, and added the
6559 names of the branch decrement conditional patterns.
6561 2018-01-08 Richard Biener <rguenther@suse.de>
6563 PR tree-optimization/83563
6564 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6567 2018-01-08 Richard Biener <rguenther@suse.de>
6570 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6572 2018-01-08 Richard Biener <rguenther@suse.de>
6574 PR tree-optimization/83685
6575 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6576 references to abnormals.
6578 2018-01-08 Richard Biener <rguenther@suse.de>
6581 * dwarf2out.c (output_indirect_strings): Handle empty
6582 skeleton_debug_str_hash.
6583 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6585 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6587 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6588 (emit_store_direct): Likewise.
6589 (arc_trampoline_adjust_address): Likewise.
6590 (arc_asm_trampoline_template): New function.
6591 (arc_initialize_trampoline): Use asm_trampoline_template.
6592 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6593 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6594 * config/arc/arc.md (flush_icache): Delete pattern.
6596 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6598 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6599 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6602 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6605 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6606 by not USED_FOR_TARGET.
6607 (make_pass_resolve_sw_modes): Likewise.
6609 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6611 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6614 2018-01-08 Richard Biener <rguenther@suse.de>
6617 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6619 2018-01-08 Richard Biener <rguenther@suse.de>
6622 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6624 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6627 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6628 basic blocks with a small number of successors.
6629 (convert_control_dep_chain_into_preds): Improve handling of
6631 (dump_predicates): Split apart into...
6632 (dump_pred_chain): ...here...
6633 (dump_pred_info): ...and here.
6634 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6635 (can_chain_union_be_invalidated_p): Improve check for invalidation
6637 (uninit_uses_cannot_happen): Avoid unnecessary if
6638 convert_control_dep_chain_into_preds yielded nothing.
6640 2018-01-06 Martin Sebor <msebor@redhat.com>
6642 PR tree-optimization/83640
6643 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6644 subtracting negative offset from size.
6645 (builtin_access::overlap): Adjust offset bounds of the access to fall
6646 within the size of the object if possible.
6648 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6650 PR rtl-optimization/83699
6651 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6652 extract_bit_field_as_subreg to cases in which the extracted
6653 value is also a vector.
6655 * lra-constraints.c (process_alt_operands): Test for the equivalence
6656 substitutions when detecting a possible reload cycle.
6658 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6661 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6662 by default if flag_selective_schedling{,2}. Formatting fixes.
6664 PR rtl-optimization/83682
6665 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6666 if it has non-VECTOR_MODE element mode.
6667 (vec_duplicate_p): Likewise.
6670 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6671 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6673 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6676 * config/i386/i386-builtin.def
6677 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6678 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6679 Require also OPTION_MASK_ISA_AVX512F in addition to
6680 OPTION_MASK_ISA_GFNI.
6681 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6682 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6683 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6684 to OPTION_MASK_ISA_GFNI.
6685 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6686 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6687 OPTION_MASK_ISA_AVX512BW.
6688 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6689 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6690 addition to OPTION_MASK_ISA_GFNI.
6691 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6692 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6693 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6694 to OPTION_MASK_ISA_GFNI.
6695 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6696 a requirement for all ISAs rather than any of them with a few
6698 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6700 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6701 bitmasks to be enabled with 3 exceptions, instead of requiring any
6702 enabled ISA with lots of exceptions.
6703 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6704 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6705 Change avx512bw in isa attribute to avx512f.
6706 * config/i386/sgxintrin.h: Add license boilerplate.
6707 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6708 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6709 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6710 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6712 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6713 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6714 temporarily sse2 rather than sse if not enabled already.
6717 * config/i386/sse.md (VI248_VLBW): Rename to ...
6718 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6719 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6720 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6721 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6722 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6723 mode iterator instead of VI248_VLBW.
6725 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6727 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6728 (record_modified): Skip clobbers; add debug output.
6729 (param_change_prob): Use sreal frequencies.
6731 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6733 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6734 punt for user-aligned variables.
6736 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6738 * tree-chrec.c (chrec_contains_symbols): Return true for
6741 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6744 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6745 of (x|y) == x for BICS pattern.
6747 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6749 PR tree-optimization/83605
6750 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6751 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6754 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6756 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6757 * config/epiphany/rtems.h: New file.
6759 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6760 Uros Bizjak <ubizjak@gmail.com>
6763 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6764 QIreg_operand instead of register_operand predicate.
6765 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6766 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6767 comments instead of -fmitigate[-_]rop.
6769 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6772 * cgraphunit.c (symbol_table::compile): Switch to text_section
6773 before calling assembly_start debug hook.
6774 * run-rtl-passes.c (run_rtl_passes): Likewise.
6777 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6779 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6780 range_int_cst_p rather than !symbolic_range_p before calling
6781 extract_range_from_multiplicative_op_1.
6783 2018-01-04 Jeff Law <law@redhat.com>
6785 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6786 redundant test in assertion.
6788 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6790 * doc/rtl.texi: Document machine_mode wrapper classes.
6792 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6794 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6797 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6799 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6800 the VEC_PERM_EXPR fold to fail.
6802 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6805 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6806 to switched_sections.
6808 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6811 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6814 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6817 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6818 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6820 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6823 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6824 is BLKmode and bitpos not zero or mode change is needed.
6826 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6829 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6832 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6835 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6836 instead of MULT rtx. Update all corresponding splitters.
6838 (*ssub<modesuffix>): Ditto.
6840 (*cmp_sadd_di): Update split patterns.
6841 (*cmp_sadd_si): Ditto.
6842 (*cmp_sadd_sidi): Ditto.
6843 (*cmp_ssub_di): Ditto.
6844 (*cmp_ssub_si): Ditto.
6845 (*cmp_ssub_sidi): Ditto.
6846 * config/alpha/predicates.md (const23_operand): New predicate.
6847 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6848 Look for ASHIFT, not MULT inner operand.
6849 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6851 2018-01-04 Martin Liska <mliska@suse.cz>
6853 PR gcov-profile/83669
6854 * gcov.c (output_intermediate_file): Add version to intermediate
6856 * doc/gcov.texi: Document new field 'version' in intermediate
6857 file format. Fix location of '-k' option of gcov command.
6859 2018-01-04 Martin Liska <mliska@suse.cz>
6862 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6864 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6866 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6868 2018-01-03 Martin Sebor <msebor@redhat.com>
6870 PR tree-optimization/83655
6871 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6872 checking calls with invalid arguments.
6874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6876 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6877 (vectorizable_mask_load_store): Delete.
6878 (vectorizable_call): Return false for masked loads and stores.
6879 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6880 instead of gimple_assign_rhs1.
6881 (vectorizable_load): Handle IFN_MASK_LOAD.
6882 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6886 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6888 (vectorizable_mask_load_store): ...here.
6889 (vectorizable_load): ...and here.
6891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6893 * tree-vect-stmts.c (vect_build_all_ones_mask)
6894 (vect_build_zero_merge_argument): New functions, split out from...
6895 (vectorizable_load): ...here.
6897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6899 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6901 (vectorizable_mask_load_store): ...here.
6902 (vectorizable_store): ...and here.
6904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6906 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6908 (vectorizable_mask_load_store): ...here.
6910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6912 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6913 (vect_model_store_cost): Take a vec_load_store_type instead of a
6915 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6916 (vect_model_store_cost): Take a vec_load_store_type instead of a
6918 (vectorizable_mask_load_store): Update accordingly.
6919 (vectorizable_store): Likewise.
6920 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6924 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6925 IFN_MASK_LOAD calls here rather than...
6926 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6929 Alan Hayward <alan.hayward@arm.com>
6930 David Sherwood <david.sherwood@arm.com>
6932 * expmed.c (extract_bit_field_1): For vector extracts,
6933 fall back to extract_bit_field_as_subreg if vec_extract
6936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6937 Alan Hayward <alan.hayward@arm.com>
6938 David Sherwood <david.sherwood@arm.com>
6940 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6941 they are variable or constant sized.
6942 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6943 slots for constant-sized data.
6945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6946 Alan Hayward <alan.hayward@arm.com>
6947 David Sherwood <david.sherwood@arm.com>
6949 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6950 handling COND_EXPRs with boolean comparisons, try to find a better
6951 basis for the mask type than the boolean itself.
6953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6955 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6956 is calculated and how it can be overridden.
6957 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6958 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6960 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6964 Alan Hayward <alan.hayward@arm.com>
6965 David Sherwood <david.sherwood@arm.com>
6967 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6968 Remove the mode argument.
6969 (aarch64_simd_valid_immediate): Remove the mode and inverse
6971 * config/aarch64/iterators.md (bitsize): New iterator.
6972 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6973 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6974 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6975 aarch64_simd_valid_immediate.
6976 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6977 (aarch64_reg_or_bic_imm): Likewise.
6978 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6979 with an insn_type enum and msl with a modifier_type enum.
6980 Replace element_width with a scalar_mode. Change the shift
6981 to unsigned int. Add constructors for scalar_float_mode and
6982 scalar_int_mode elements.
6983 (aarch64_vect_float_const_representable_p): Delete.
6984 (aarch64_can_const_movi_rtx_p)
6985 (aarch64_simd_scalar_immediate_valid_for_move)
6986 (aarch64_simd_make_constant): Update call to
6987 aarch64_simd_valid_immediate.
6988 (aarch64_advsimd_valid_immediate_hs): New function.
6989 (aarch64_advsimd_valid_immediate): Likewise.
6990 (aarch64_simd_valid_immediate): Remove mode and inverse
6991 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6992 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6993 and aarch64_float_const_representable_p on the result.
6994 (aarch64_output_simd_mov_immediate): Remove mode argument.
6995 Update call to aarch64_simd_valid_immediate and use of
6996 simd_immediate_info.
6997 (aarch64_output_scalar_simd_mov_immediate): Update call
7000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7001 Alan Hayward <alan.hayward@arm.com>
7002 David Sherwood <david.sherwood@arm.com>
7004 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7005 (mode_nunits): Likewise CONST_MODE_NUNITS.
7006 * machmode.def (ADJUST_NUNITS): Document.
7007 * genmodes.c (mode_data::need_nunits_adj): New field.
7008 (blank_mode): Update accordingly.
7009 (adj_nunits): New variable.
7010 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7012 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7013 listed in adj_nunits.
7014 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7015 listed in adj_nunits. Don't emit case statements for such modes.
7016 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7017 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7018 nothing if adj_nunits is nonnull.
7019 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7020 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7021 (emit_mode_fbit): Update use of print_maybe_const_decl.
7022 (emit_move_size): Likewise. Treat the array as non-const
7024 (emit_mode_adjustments): Handle adj_nunits.
7026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7028 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7029 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7030 (VECTOR_MODES): Use it.
7031 (make_vector_modes): Take the prefix as an argument.
7033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7034 Alan Hayward <alan.hayward@arm.com>
7035 David Sherwood <david.sherwood@arm.com>
7037 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7038 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7039 for MODE_VECTOR_BOOL.
7040 * machmode.def (VECTOR_BOOL_MODE): Document.
7041 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7042 (make_vector_bool_mode): New function.
7043 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7045 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7046 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7048 * stor-layout.c (int_mode_for_mode): Likewise.
7049 * tree.c (build_vector_type_for_mode): Likewise.
7050 * varasm.c (output_constant_pool_2): Likewise.
7051 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7052 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7053 for MODE_VECTOR_BOOL.
7054 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7055 of mode class checks.
7056 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7057 instead of a list of mode class checks.
7058 (expand_vector_scalar_condition): Likewise.
7059 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7062 Alan Hayward <alan.hayward@arm.com>
7063 David Sherwood <david.sherwood@arm.com>
7065 * machmode.h (mode_size): Change from unsigned short to
7067 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7068 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7069 or if measurement_type is not polynomial.
7070 (fixed_size_mode::includes_p): Check for constant-sized modes.
7071 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7072 return a poly_uint16 rather than an unsigned short.
7073 (emit_mode_size): Change the type of mode_size from unsigned short
7074 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7075 (emit_mode_adjustments): Cope with polynomial vector sizes.
7076 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7078 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7080 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7081 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7082 * caller-save.c (setup_save_areas): Likewise.
7083 (replace_reg_with_saved_mem): Likewise.
7084 * calls.c (emit_library_call_value_1): Likewise.
7085 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7086 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7087 (gen_lowpart_for_combine): Likewise.
7088 * convert.c (convert_to_integer_1): Likewise.
7089 * cse.c (equiv_constant, cse_insn): Likewise.
7090 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7091 (cselib_subst_to_values): Likewise.
7092 * dce.c (word_dce_process_block): Likewise.
7093 * df-problems.c (df_word_lr_mark_ref): Likewise.
7094 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7095 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7096 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7097 (rtl_for_decl_location): Likewise.
7098 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7099 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7100 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7101 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7102 (expand_expr_real_1): Likewise.
7103 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7104 (pad_below): Likewise.
7105 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7106 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7107 * ira.c (get_subreg_tracking_sizes): Likewise.
7108 * ira-build.c (ira_create_allocno_objects): Likewise.
7109 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7110 (ira_sort_regnos_for_alter_reg): Likewise.
7111 * ira-costs.c (record_operand_costs): Likewise.
7112 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7113 (resolve_simple_move): Likewise.
7114 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7115 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7116 (lra_constraints): Likewise.
7117 (CONST_POOL_OK_P): Reject variable-sized modes.
7118 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7119 (add_pseudo_to_slot, lra_spill): Likewise.
7120 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7121 * optabs-query.c (get_best_extraction_insn): Likewise.
7122 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7123 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7124 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7125 * recog.c (offsettable_address_addr_space_p): Likewise.
7126 * regcprop.c (maybe_mode_change): Likewise.
7127 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7128 * regrename.c (build_def_use): Likewise.
7129 * regstat.c (dump_reg_info): Likewise.
7130 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7131 (find_reloads, find_reloads_subreg_address): Likewise.
7132 * reload1.c (eliminate_regs_1): Likewise.
7133 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7134 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7135 (simplify_binary_operation_1, simplify_subreg): Likewise.
7136 * targhooks.c (default_function_arg_padding): Likewise.
7137 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7138 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7139 (verify_gimple_assign_ternary): Likewise.
7140 * tree-inline.c (estimate_move_cost): Likewise.
7141 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7142 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7143 (get_address_cost_ainc): Likewise.
7144 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7145 (vect_supportable_dr_alignment): Likewise.
7146 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7147 (vectorizable_reduction): Likewise.
7148 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7149 (vectorizable_operation, vectorizable_load): Likewise.
7150 * tree.c (build_same_sized_truth_vector_type): Likewise.
7151 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7152 * var-tracking.c (emit_note_insn_var_location): Likewise.
7153 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7154 (ADDR_VEC_ALIGN): Likewise.
7156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7157 Alan Hayward <alan.hayward@arm.com>
7158 David Sherwood <david.sherwood@arm.com>
7160 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7162 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7163 or if measurement_type is polynomial.
7164 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7165 * combine.c (make_extraction): Likewise.
7166 * dse.c (find_shift_sequence): Likewise.
7167 * dwarf2out.c (mem_loc_descriptor): Likewise.
7168 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7169 (extract_bit_field, extract_low_bits): Likewise.
7170 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7171 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7172 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7173 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7174 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7175 * reload.c (find_reloads): Likewise.
7176 * reload1.c (alter_reg): Likewise.
7177 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7178 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7179 * tree-if-conv.c (predicate_mem_writes): Likewise.
7180 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7181 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7182 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7183 * valtrack.c (dead_debug_insert_temp): Likewise.
7184 * varasm.c (mergeable_constant_section): Likewise.
7185 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7188 Alan Hayward <alan.hayward@arm.com>
7189 David Sherwood <david.sherwood@arm.com>
7191 * expr.c (expand_assignment): Cope with polynomial mode sizes
7192 when assigning to a CONCAT.
7194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7195 Alan Hayward <alan.hayward@arm.com>
7196 David Sherwood <david.sherwood@arm.com>
7198 * machmode.h (mode_precision): Change from unsigned short to
7200 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7202 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7203 or if measurement_type is not polynomial.
7204 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7205 in which the mode is already known to be a scalar_int_mode.
7206 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7207 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7209 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7210 for GET_MODE_PRECISION.
7211 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7212 for GET_MODE_PRECISION.
7213 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7215 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7216 (expand_field_assignment, make_extraction): Likewise.
7217 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7218 (get_last_value): Likewise.
7219 * convert.c (convert_to_integer_1): Likewise.
7220 * cse.c (cse_insn): Likewise.
7221 * expr.c (expand_expr_real_1): Likewise.
7222 * lra-constraints.c (simplify_operand_subreg): Likewise.
7223 * optabs-query.c (can_atomic_load_p): Likewise.
7224 * optabs.c (expand_atomic_load): Likewise.
7225 (expand_atomic_store): Likewise.
7226 * ree.c (combine_reaching_defs): Likewise.
7227 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7228 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7229 * tree.h (type_has_mode_precision_p): Likewise.
7230 * ubsan.c (instrument_si_overflow): Likewise.
7232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7233 Alan Hayward <alan.hayward@arm.com>
7234 David Sherwood <david.sherwood@arm.com>
7236 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7237 polynomial numbers of units.
7238 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7239 (valid_vector_subparts_p): New function.
7240 (build_vector_type): Remove temporary shim and take the number
7241 of units as a poly_uint64 rather than an int.
7242 (build_opaque_vector_type): Take the number of units as a
7243 poly_uint64 rather than an int.
7244 * tree.c (build_vector_from_ctor): Handle polynomial
7245 TYPE_VECTOR_SUBPARTS.
7246 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7247 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7248 (build_vector_from_val): If the number of units is variable,
7249 use build_vec_duplicate_cst for constant operands and
7250 VEC_DUPLICATE_EXPR otherwise.
7251 (make_vector_type): Remove temporary is_constant ().
7252 (build_vector_type, build_opaque_vector_type): Take the number of
7253 units as a poly_uint64 rather than an int.
7254 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7256 * cfgexpand.c (expand_debug_expr): Likewise.
7257 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7258 (store_constructor, expand_expr_real_1): Likewise.
7259 (const_scalar_mask_from_tree): Likewise.
7260 * fold-const-call.c (fold_const_reduction): Likewise.
7261 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7262 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7263 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7264 (fold_relational_const): Likewise.
7265 (native_interpret_vector): Likewise. Change the size from an
7266 int to an unsigned int.
7267 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7268 TYPE_VECTOR_SUBPARTS.
7269 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7270 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7271 duplicating a non-constant operand into a variable-length vector.
7272 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7273 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7274 * ipa-icf.c (sem_variable::equals): Likewise.
7275 * match.pd: Likewise.
7276 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7277 * print-tree.c (print_node): Likewise.
7278 * stor-layout.c (layout_type): Likewise.
7279 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7280 * tree-cfg.c (verify_gimple_comparison): Likewise.
7281 (verify_gimple_assign_binary): Likewise.
7282 (verify_gimple_assign_ternary): Likewise.
7283 (verify_gimple_assign_single): Likewise.
7284 * tree-pretty-print.c (dump_generic_node): Likewise.
7285 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7286 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7287 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7288 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7289 (vect_shift_permute_load_chain): Likewise.
7290 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7291 (expand_vector_condition, optimize_vector_constructor): Likewise.
7292 (lower_vec_perm, get_compute_type): Likewise.
7293 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7294 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7295 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7296 (vect_recog_mask_conversion_pattern): Likewise.
7297 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7298 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7299 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7300 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7301 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7302 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7303 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7304 (supportable_widening_operation): Likewise.
7305 (supportable_narrowing_operation): Likewise.
7306 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7308 * varasm.c (output_constant): Likewise.
7310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7311 Alan Hayward <alan.hayward@arm.com>
7312 David Sherwood <david.sherwood@arm.com>
7314 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7315 so that both the length == 3 and length != 3 cases set up their
7316 own permute vectors. Add comments explaining why we know the
7317 number of elements is constant.
7318 (vect_permute_load_chain): Likewise.
7320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7321 Alan Hayward <alan.hayward@arm.com>
7322 David Sherwood <david.sherwood@arm.com>
7324 * machmode.h (mode_nunits): Change from unsigned char to
7326 (ONLY_FIXED_SIZE_MODES): New macro.
7327 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7328 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7329 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7331 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7332 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7333 or if measurement_type is not polynomial.
7334 * genmodes.c (ZERO_COEFFS): New macro.
7335 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7337 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7338 Use ZERO_COEFFS when emitting initializers.
7339 * data-streamer.h (bp_pack_poly_value): New function.
7340 (bp_unpack_poly_value): Likewise.
7341 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7342 for GET_MODE_NUNITS.
7343 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7344 for GET_MODE_NUNITS.
7345 * tree.c (make_vector_type): Remove temporary shim and make
7346 the real function take the number of units as a poly_uint64
7348 (build_vector_type_for_mode): Handle polynomial nunits.
7349 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7350 * emit-rtl.c (const_vec_series_p_1): Likewise.
7351 (gen_rtx_CONST_VECTOR): Likewise.
7352 * fold-const.c (test_vec_duplicate_folding): Likewise.
7353 * genrecog.c (validate_pattern): Likewise.
7354 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7355 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7356 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7357 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7358 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7359 * rtlanal.c (subreg_get_info): Likewise.
7360 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7361 (vect_grouped_load_supported): Likewise.
7362 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7363 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7364 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7365 (simplify_const_unary_operation, simplify_binary_operation_1)
7366 (simplify_const_binary_operation, simplify_ternary_operation)
7367 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7368 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7369 instead of CONST_VECTOR_NUNITS.
7370 * varasm.c (output_constant_pool_2): Likewise.
7371 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7372 explicit-encoded elements in the XVEC for variable-length vectors.
7374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7376 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7379 Alan Hayward <alan.hayward@arm.com>
7380 David Sherwood <david.sherwood@arm.com>
7382 * coretypes.h (fixed_size_mode): Declare.
7383 (fixed_size_mode_pod): New typedef.
7384 * builtins.h (target_builtins::x_apply_args_mode)
7385 (target_builtins::x_apply_result_mode): Change type to
7386 fixed_size_mode_pod.
7387 * builtins.c (apply_args_size, apply_result_size, result_vector)
7388 (expand_builtin_apply_args_1, expand_builtin_apply)
7389 (expand_builtin_return): Update accordingly.
7391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7393 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7394 * cselib.c (cselib_hash_rtx): Likewise.
7395 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7396 CONST_VECTOR encoding.
7398 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7399 Jeff Law <law@redhat.com>
7402 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7403 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7404 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7405 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7408 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7409 explicitly probe *sp in a noreturn function if there were any callee
7410 register saves or frame pointer is needed.
7412 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7415 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7416 BLKmode for ternary, binary or unary expressions.
7419 * var-tracking.c (delete_vta_debug_insn): New inline function.
7420 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7421 insns from get_insns () to NULL instead of each bb separately.
7422 Use delete_vta_debug_insn. No longer static.
7423 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7424 delete_vta_debug_insns callers.
7425 * rtl.h (delete_vta_debug_insns): Declare.
7426 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7427 instead of variable_tracking_main.
7429 2018-01-03 Martin Sebor <msebor@redhat.com>
7431 PR tree-optimization/83603
7432 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7433 arguments past the endof the argument list in functions declared
7434 without a prototype.
7435 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7436 Avoid checking when arguments are null.
7438 2018-01-03 Martin Sebor <msebor@redhat.com>
7441 * doc/extend.texi (attribute const): Fix a typo.
7442 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7443 issuing -Wsuggest-attribute for void functions.
7445 2018-01-03 Martin Sebor <msebor@redhat.com>
7447 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7448 offset_int::from instead of wide_int::to_shwi.
7449 (maybe_diag_overlap): Remove assertion.
7450 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7451 * gimple-ssa-sprintf.c (format_directive): Same.
7452 (parse_directive): Same.
7453 (sprintf_dom_walker::compute_format_length): Same.
7454 (try_substitute_return_value): Same.
7456 2018-01-03 Jeff Law <law@redhat.com>
7459 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7460 non-constant residual for zero at runtime and avoid probing in
7461 that case. Reorganize code for trailing problem to mirror handling
7464 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7466 PR tree-optimization/83501
7467 * tree-ssa-strlen.c (get_string_cst): New.
7468 (handle_char_store): Call get_string_cst.
7470 2018-01-03 Martin Liska <mliska@suse.cz>
7472 PR tree-optimization/83593
7473 * tree-ssa-strlen.c: Include tree-cfg.h.
7474 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7475 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7476 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7478 (strlen_dom_walker::before_dom_children): Call
7479 gimple_purge_dead_eh_edges. Dump tranformation with details
7481 (strlen_dom_walker::before_dom_children): Update call by adding
7482 new argument cleanup_eh.
7483 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7485 2018-01-03 Martin Liska <mliska@suse.cz>
7488 * cif-code.def (VARIADIC_THUNK): New enum value.
7489 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7492 2018-01-03 Jan Beulich <jbeulich@suse.com>
7494 * sse.md (mov<mode>_internal): Tighten condition for when to use
7495 vmovdqu<ssescalarsize> for TI and OI modes.
7497 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7499 Update copyright years.
7501 2018-01-03 Martin Liska <mliska@suse.cz>
7504 * ipa-visibility.c (function_and_variable_visibility): Skip
7505 functions with noipa attribure.
7507 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7509 * gcc.c (process_command): Update copyright notice dates.
7510 * gcov-dump.c (print_version): Ditto.
7511 * gcov.c (print_version): Ditto.
7512 * gcov-tool.c (print_version): Ditto.
7513 * gengtype.c (create_file): Ditto.
7514 * doc/cpp.texi: Bump @copying's copyright year.
7515 * doc/cppinternals.texi: Ditto.
7516 * doc/gcc.texi: Ditto.
7517 * doc/gccint.texi: Ditto.
7518 * doc/gcov.texi: Ditto.
7519 * doc/install.texi: Ditto.
7520 * doc/invoke.texi: Ditto.
7522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7524 * vector-builder.h (vector_builder::m_full_nelts): Change from
7525 unsigned int to poly_uint64.
7526 (vector_builder::full_nelts): Update prototype accordingly.
7527 (vector_builder::new_vector): Likewise.
7528 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7529 (vector_builder::operator ==): Likewise.
7530 (vector_builder::finalize): Likewise.
7531 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7532 Take the number of elements as a poly_uint64 rather than an
7534 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7535 from unsigned int to poly_uint64.
7536 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7537 (vec_perm_indices::new_vector): Likewise.
7538 (vec_perm_indices::length): Likewise.
7539 (vec_perm_indices::nelts_per_input): Likewise.
7540 (vec_perm_indices::input_nelts): Likewise.
7541 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7542 number of elements per input as a poly_uint64 rather than an
7543 unsigned int. Use the original encoding for variable-length
7544 vectors, rather than clamping each individual element.
7545 For the second and subsequent elements in each pattern,
7546 clamp the step and base before clamping their sum.
7547 (vec_perm_indices::series_p): Handle polynomial element counts.
7548 (vec_perm_indices::all_in_range_p): Likewise.
7549 (vec_perm_indices_to_tree): Likewise.
7550 (vec_perm_indices_to_rtx): Likewise.
7551 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7552 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7553 (tree_vector_builder::new_binary_operation): Handle polynomial
7554 element counts. Return false if we need to know the number
7555 of elements at compile time.
7556 * fold-const.c (fold_vec_perm): Punt if the number of elements
7557 isn't known at compile time.
7559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7561 * vec-perm-indices.h (vec_perm_builder): Change element type
7562 from HOST_WIDE_INT to poly_int64.
7563 (vec_perm_indices::element_type): Update accordingly.
7564 (vec_perm_indices::clamp): Handle polynomial element_types.
7565 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7566 (vec_perm_indices::all_in_range_p): Likewise.
7567 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7569 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7570 polynomial vec_perm_indices element types.
7571 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7572 * fold-const.c (fold_vec_perm): Likewise.
7573 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7574 * tree-vect-generic.c (lower_vec_perm): Likewise.
7575 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7576 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7577 element type to HOST_WIDE_INT.
7579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7580 Alan Hayward <alan.hayward@arm.com>
7581 David Sherwood <david.sherwood@arm.com>
7583 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7584 rather than an int. Use plus_constant.
7585 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7586 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7589 Alan Hayward <alan.hayward@arm.com>
7590 David Sherwood <david.sherwood@arm.com>
7592 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7593 a HOST_WIDE_INT to a poly_int64.
7595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7596 Alan Hayward <alan.hayward@arm.com>
7597 David Sherwood <david.sherwood@arm.com>
7599 * calls.c (load_register_parameters): Cope with polynomial
7600 mode sizes. Require a constant size for BLKmode parameters
7601 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7602 forces a parameter to be padded at the lsb end in order to
7603 fill a complete number of words, require the parameter size
7604 to be ordered wrt UNITS_PER_WORD.
7606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7607 Alan Hayward <alan.hayward@arm.com>
7608 David Sherwood <david.sherwood@arm.com>
7610 * reload1.c (spill_stack_slot_width): Change element type
7611 from unsigned int to poly_uint64_pod.
7612 (alter_reg): Treat mode sizes as polynomial.
7614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7615 Alan Hayward <alan.hayward@arm.com>
7616 David Sherwood <david.sherwood@arm.com>
7618 * reload.c (complex_word_subreg_p): New function.
7619 (reload_inner_reg_of_subreg, push_reload): Use it.
7621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7622 Alan Hayward <alan.hayward@arm.com>
7623 David Sherwood <david.sherwood@arm.com>
7625 * lra-constraints.c (process_alt_operands): Reject matched
7626 operands whose sizes aren't ordered.
7627 (match_reload): Refer to this check here.
7629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7630 Alan Hayward <alan.hayward@arm.com>
7631 David Sherwood <david.sherwood@arm.com>
7633 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7634 that the mode size is in the set {1, 2, 4, 8, 16}.
7636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7637 Alan Hayward <alan.hayward@arm.com>
7638 David Sherwood <david.sherwood@arm.com>
7640 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7641 Use plus_constant instead of gen_rtx_PLUS.
7643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7644 Alan Hayward <alan.hayward@arm.com>
7645 David Sherwood <david.sherwood@arm.com>
7647 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7648 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7649 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7650 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7651 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7652 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7653 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7654 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7655 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7656 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7658 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7659 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7660 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7661 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7662 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7663 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7664 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7665 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7666 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7667 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7669 * expr.c (emit_move_resolve_push): Treat the input and result
7670 of PUSH_ROUNDING as a poly_int64.
7671 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7672 (emit_push_insn): Likewise.
7673 * lra-eliminations.c (mark_not_eliminable): Likewise.
7674 * recog.c (push_operand): Likewise.
7675 * reload1.c (elimination_effects): Likewise.
7676 * rtlanal.c (nonzero_bits1): Likewise.
7677 * calls.c (store_one_arg): Likewise. Require the padding to be
7678 known at compile time.
7680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7681 Alan Hayward <alan.hayward@arm.com>
7682 David Sherwood <david.sherwood@arm.com>
7684 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7685 Use plus_constant instead of gen_rtx_PLUS.
7687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7688 Alan Hayward <alan.hayward@arm.com>
7689 David Sherwood <david.sherwood@arm.com>
7691 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7695 Alan Hayward <alan.hayward@arm.com>
7696 David Sherwood <david.sherwood@arm.com>
7698 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7699 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7700 via stack temporaries. Treat the mode size as polynomial too.
7702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7703 Alan Hayward <alan.hayward@arm.com>
7704 David Sherwood <david.sherwood@arm.com>
7706 * expr.c (expand_expr_real_2): When handling conversions involving
7707 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7708 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7709 as a poly_uint64 too.
7711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7712 Alan Hayward <alan.hayward@arm.com>
7713 David Sherwood <david.sherwood@arm.com>
7715 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7718 Alan Hayward <alan.hayward@arm.com>
7719 David Sherwood <david.sherwood@arm.com>
7721 * combine.c (can_change_dest_mode): Handle polynomial
7722 REGMODE_NATURAL_SIZE.
7723 * expmed.c (store_bit_field_1): Likewise.
7724 * expr.c (store_constructor): Likewise.
7725 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7726 and polynomial REGMODE_NATURAL_SIZE.
7727 (gen_lowpart_common): Likewise.
7728 * reginfo.c (record_subregs_of_mode): Likewise.
7729 * rtlanal.c (read_modify_subreg_p): Likewise.
7731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7732 Alan Hayward <alan.hayward@arm.com>
7733 David Sherwood <david.sherwood@arm.com>
7735 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7736 numbers of elements.
7738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7739 Alan Hayward <alan.hayward@arm.com>
7740 David Sherwood <david.sherwood@arm.com>
7742 * match.pd: Cope with polynomial numbers of vector elements.
7744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7745 Alan Hayward <alan.hayward@arm.com>
7746 David Sherwood <david.sherwood@arm.com>
7748 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7749 in a POINTER_PLUS_EXPR.
7751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7752 Alan Hayward <alan.hayward@arm.com>
7753 David Sherwood <david.sherwood@arm.com>
7755 * omp-simd-clone.c (simd_clone_subparts): New function.
7756 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7757 (ipa_simd_modify_function_body): Likewise.
7759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7760 Alan Hayward <alan.hayward@arm.com>
7761 David Sherwood <david.sherwood@arm.com>
7763 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7764 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7765 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7766 (expand_vector_condition, vector_element): Likewise.
7767 (subparts_gt): New function.
7768 (get_compute_type): Use subparts_gt.
7769 (count_type_subparts): Delete.
7770 (expand_vector_operations_1): Use subparts_gt instead of
7771 count_type_subparts.
7773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7774 Alan Hayward <alan.hayward@arm.com>
7775 David Sherwood <david.sherwood@arm.com>
7777 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7778 (vect_compile_time_alias): ...this new function. Do the calculation
7779 on poly_ints rather than trees.
7780 (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
7788 (vect_schedule_slp_instance): Likewise.
7790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7791 Alan Hayward <alan.hayward@arm.com>
7792 David Sherwood <david.sherwood@arm.com>
7794 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7795 constant and extern definitions for variable-length vectors.
7796 (vect_get_constant_vectors): Note that the number of units
7797 is known to be constant.
7799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7800 Alan Hayward <alan.hayward@arm.com>
7801 David Sherwood <david.sherwood@arm.com>
7803 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7804 of units as polynomial. Choose between WIDE and NARROW based
7807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7808 Alan Hayward <alan.hayward@arm.com>
7809 David Sherwood <david.sherwood@arm.com>
7811 * tree-vect-stmts.c (simd_clone_subparts): New function.
7812 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7815 Alan Hayward <alan.hayward@arm.com>
7816 David Sherwood <david.sherwood@arm.com>
7818 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7819 vectors as polynomial. Use build_index_vector for
7822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7823 Alan Hayward <alan.hayward@arm.com>
7824 David Sherwood <david.sherwood@arm.com>
7826 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7827 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7828 for variable-length vectors.
7829 (vectorizable_mask_load_store): Treat the number of units as
7830 polynomial, asserting that it is constant if the condition has
7831 already been enforced.
7832 (vectorizable_store, vectorizable_load): Likewise.
7834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7835 Alan Hayward <alan.hayward@arm.com>
7836 David Sherwood <david.sherwood@arm.com>
7838 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7839 of units as polynomial. Punt if we can't tell at compile time
7840 which vector contains the final result.
7842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7843 Alan Hayward <alan.hayward@arm.com>
7844 David Sherwood <david.sherwood@arm.com>
7846 * tree-vect-loop.c (vectorizable_induction): Treat the number
7847 of units as polynomial. Punt on SLP inductions. Use an integer
7848 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7849 cast of such a series for variable-length floating-point
7852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7853 Alan Hayward <alan.hayward@arm.com>
7854 David Sherwood <david.sherwood@arm.com>
7856 * tree.h (build_index_vector): Declare.
7857 * tree.c (build_index_vector): New function.
7858 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7859 of units as polynomial, forcibly converting it to a constant if
7860 vectorizable_reduction has already enforced the condition.
7861 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7862 to create a {1,2,3,...} vector.
7863 (vectorizable_reduction): Treat the number of units as polynomial.
7864 Choose vectype_in based on the largest scalar element size rather
7865 than the smallest number of units. Enforce the restrictions
7868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7869 Alan Hayward <alan.hayward@arm.com>
7870 David Sherwood <david.sherwood@arm.com>
7872 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7873 number of units as polynomial.
7875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7876 Alan Hayward <alan.hayward@arm.com>
7877 David Sherwood <david.sherwood@arm.com>
7879 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7880 * target.def (autovectorize_vector_sizes): Return the vector sizes
7881 by pointer, using vector_sizes rather than a bitmask.
7882 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7883 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7884 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7886 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7887 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7888 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7889 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7890 * omp-general.c (omp_max_vf): Likewise.
7891 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7892 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7893 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7894 * tree-vect-slp.c (vect_slp_bb): Likewise.
7895 * doc/tm.texi: Regenerate.
7896 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7898 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7899 the vector size as a poly_uint64 rather than an unsigned int.
7900 (current_vector_size): Change from an unsigned int to a poly_uint64.
7901 (get_vectype_for_scalar_type): Update accordingly.
7902 * tree.h (build_truth_vector_type): Take the size and number of
7903 units as a poly_uint64 rather than an unsigned int.
7904 (build_vector_type): Add a temporary overload that takes
7905 the number of units as a poly_uint64 rather than an unsigned int.
7906 * tree.c (make_vector_type): Likewise.
7907 (build_truth_vector_type): Take the number of units as a poly_uint64
7908 rather than an unsigned int.
7910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7911 Alan Hayward <alan.hayward@arm.com>
7912 David Sherwood <david.sherwood@arm.com>
7914 * target.def (get_mask_mode): Take the number of units and length
7915 as poly_uint64s rather than unsigned ints.
7916 * targhooks.h (default_get_mask_mode): Update accordingly.
7917 * targhooks.c (default_get_mask_mode): Likewise.
7918 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7919 * doc/tm.texi: Regenerate.
7921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7922 Alan Hayward <alan.hayward@arm.com>
7923 David Sherwood <david.sherwood@arm.com>
7925 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7926 * omp-general.c (omp_max_vf): Likewise.
7927 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7928 (expand_omp_simd): Handle polynomial safelen.
7929 * omp-low.c (omplow_simd_context): Add a default constructor.
7930 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7931 (lower_rec_simd_input_clauses): Update accordingly.
7932 (lower_rec_input_clauses): Likewise.
7934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7935 Alan Hayward <alan.hayward@arm.com>
7936 David Sherwood <david.sherwood@arm.com>
7938 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7939 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7940 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7941 (vect_analyze_slp_cost): Likewise.
7942 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7943 (vect_model_load_cost): Likewise.
7945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7946 Alan Hayward <alan.hayward@arm.com>
7947 David Sherwood <david.sherwood@arm.com>
7949 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7950 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7951 from an unsigned int * to a poly_uint64_pod *.
7952 (calculate_unrolling_factor): New function.
7953 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7956 Alan Hayward <alan.hayward@arm.com>
7957 David Sherwood <david.sherwood@arm.com>
7959 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7960 from an unsigned int to a poly_uint64.
7961 (_loop_vec_info::slp_unrolling_factor): Likewise.
7962 (_loop_vec_info::vectorization_factor): Change from an int
7964 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7965 (vect_get_num_vectors): New function.
7966 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7967 (vect_get_num_copies): Use vect_get_num_vectors.
7968 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7969 to an unsigned int *.
7970 (vect_analyze_data_refs): Change min_vf from an int * to a
7972 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7973 than an unsigned HOST_WIDE_INT.
7974 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7975 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7976 to an unsigned int *.
7977 (vect_analyze_data_ref_dependences): Likewise.
7978 (vect_compute_data_ref_alignment): Handle polynomial vf.
7979 (vect_enhance_data_refs_alignment): Likewise.
7980 (vect_prune_runtime_alias_test_list): Likewise.
7981 (vect_shift_permute_load_chain): Likewise.
7982 (vect_supportable_dr_alignment): Likewise.
7983 (dependence_distance_ge_vf): Take the vectorization factor as a
7984 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7985 (vect_analyze_data_refs): Change min_vf from an int * to a
7987 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7988 vfm1 as a poly_uint64 rather than an int. Make the same change
7989 for the returned bound_scalar.
7990 (vect_gen_vector_loop_niters): Handle polynomial vf.
7991 (vect_do_peeling): Likewise. Update call to
7992 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7993 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7995 * tree-vect-loop.c (vect_determine_vectorization_factor)
7996 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7997 (vect_get_known_peeling_cost): Likewise.
7998 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7999 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8000 (vect_transform_loop): Likewise. Use the lowest possible VF when
8001 updating the upper bounds of the loop.
8002 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8004 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8005 polynomial unroll factors.
8006 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8007 (vect_make_slp_decision): Likewise.
8008 (vect_supported_load_permutation_p): Likewise, and polynomial
8010 (vect_analyze_slp_cost): Handle polynomial vf.
8011 (vect_slp_analyze_node_operations): Likewise.
8012 (vect_slp_analyze_bb_1): Likewise.
8013 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8014 than an unsigned HOST_WIDE_INT.
8015 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8016 (vectorizable_load): Handle polynomial vf.
8017 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8019 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8022 Alan Hayward <alan.hayward@arm.com>
8023 David Sherwood <david.sherwood@arm.com>
8025 * match.pd: Handle bit operations involving three constants
8026 and try to fold one pair.
8028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8030 * tree-vect-loop-manip.c: Include gimple-fold.h.
8031 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8032 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8033 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8034 Add a path that uses a step of VF instead of 1, but disable it
8036 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8037 and niters_no_overflow parameters. Update calls to
8038 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8039 Create a new SSA name if the latter choses to use a ste other
8040 than zero, and return it via niters_vector_mult_vf_var.
8041 * tree-vect-loop.c (vect_transform_loop): Update calls to
8042 vect_do_peeling, vect_gen_vector_loop_niters and
8043 slpeel_make_loop_iterate_ntimes.
8044 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8045 (vect_gen_vector_loop_niters): Update declarations after above changes.
8047 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8049 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8050 128-bit round to integer instructions.
8051 (ceil<mode>2): Likewise.
8052 (btrunc<mode>2): Likewise.
8053 (round<mode>2): Likewise.
8055 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8057 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8058 unaligned VSX load/store on P8/P9.
8059 (expand_block_clear): Allow the use of unaligned VSX
8060 load/store on P8/P9.
8062 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8064 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8066 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8067 swap associated with both a load and a store.
8069 2018-01-02 Andrew Waterman <andrew@sifive.com>
8071 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8072 * config/riscv/riscv.md (clear_cache): Use it.
8074 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8076 * web.c: Remove out-of-date comment.
8078 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8080 * expr.c (fixup_args_size_notes): Check that any existing
8081 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8082 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8083 (emit_single_push_insn): ...here.
8085 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8087 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8088 (const_vector_encoded_nelts): New function.
8089 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8090 (const_vector_int_elt, const_vector_elt): Declare.
8091 * emit-rtl.c (const_vector_int_elt_1): New function.
8092 (const_vector_elt): Likewise.
8093 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8094 of CONST_VECTOR_ELT.
8096 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8098 * expr.c: Include rtx-vector-builder.h.
8099 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8100 directly on the tree encoding.
8101 (const_vector_from_tree): Likewise.
8102 * optabs.c: Include rtx-vector-builder.h.
8103 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8104 sequence of "u" values.
8105 * vec-perm-indices.c: Include rtx-vector-builder.h.
8106 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8107 directly on the vec_perm_indices encoding.
8109 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8111 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8112 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8113 * rtx-vector-builder.h: New file.
8114 * rtx-vector-builder.c: Likewise.
8115 * rtl.h (rtx_def::u2): Add a const_vector field.
8116 (CONST_VECTOR_NPATTERNS): New macro.
8117 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8118 (CONST_VECTOR_DUPLICATE_P): Likewise.
8119 (CONST_VECTOR_STEPPED_P): Likewise.
8120 (CONST_VECTOR_ENCODED_ELT): Likewise.
8121 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8122 (unwrap_const_vec_duplicate): Likewise.
8123 (const_vec_series_p): Check for a non-duplicated vector encoding.
8124 Say that the function only returns true for integer vectors.
8125 * emit-rtl.c: Include rtx-vector-builder.h.
8126 (gen_const_vec_duplicate_1): Delete.
8127 (gen_const_vector): Call gen_const_vec_duplicate instead of
8128 gen_const_vec_duplicate_1.
8129 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8130 (gen_const_vec_duplicate): Use rtx_vector_builder.
8131 (gen_const_vec_series): Likewise.
8132 (gen_rtx_CONST_VECTOR): Likewise.
8133 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8134 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8135 Build a new vector rather than modifying a CONST_VECTOR in-place.
8136 (handle_special_swappables): Update call accordingly.
8137 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8138 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8139 Build a new vector rather than modifying a CONST_VECTOR in-place.
8140 (handle_special_swappables): Update call accordingly.
8142 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8144 * simplify-rtx.c (simplify_const_binary_operation): Use
8145 CONST_VECTOR_ELT instead of XVECEXP.
8147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8149 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8150 the selector elements to be different from the data elements
8151 if the selector is a VECTOR_CST.
8152 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8153 ssizetype for the selector.
8155 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8157 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8158 before testing each element individually.
8159 * tree-vect-generic.c (lower_vec_perm): Likewise.
8161 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8163 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8164 * selftest-run-tests.c (selftest::run_tests): Call it.
8165 * vector-builder.h (vector_builder::operator ==): New function.
8166 (vector_builder::operator !=): Likewise.
8167 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8168 (vec_perm_indices::all_from_input_p): New function.
8169 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8170 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8171 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8172 instead of reading the VECTOR_CST directly. Detect whether both
8173 vector inputs are the same before constructing the vec_perm_indices,
8174 and update the number of inputs argument accordingly. Use the
8175 utility functions added above. Only construct sel2 if we need to.
8177 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8179 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8180 the broadcast of the low byte.
8181 (expand_mult_highpart): Use an explicit encoding for the permutes.
8182 * optabs-query.c (can_mult_highpart_p): Likewise.
8183 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8184 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8185 (vectorizable_bswap): Likewise.
8186 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8187 explicit encoding for the power-of-2 permutes.
8188 (vect_permute_store_chain): Likewise.
8189 (vect_grouped_load_supported): Likewise.
8190 (vect_permute_load_chain): Likewise.
8192 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8194 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8195 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8196 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8197 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8198 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8199 (vect_gen_perm_mask_any): Likewise.
8201 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8203 * int-vector-builder.h: New file.
8204 * vec-perm-indices.h: Include int-vector-builder.h.
8205 (vec_perm_indices): Redefine as an int_vector_builder.
8206 (auto_vec_perm_indices): Delete.
8207 (vec_perm_builder): Redefine as a stand-alone class.
8208 (vec_perm_indices::vec_perm_indices): New function.
8209 (vec_perm_indices::clamp): Likewise.
8210 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8211 (vec_perm_indices::new_vector): New function.
8212 (vec_perm_indices::new_expanded_vector): Update for new
8213 vec_perm_indices class.
8214 (vec_perm_indices::rotate_inputs): New function.
8215 (vec_perm_indices::all_in_range_p): Operate directly on the
8216 encoded form, without computing elided elements.
8217 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8218 encoding. Update for new vec_perm_indices class.
8219 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8220 the given vec_perm_builder.
8221 (expand_vec_perm_var): Update vec_perm_builder constructor.
8222 (expand_mult_highpart): Use vec_perm_builder instead of
8223 auto_vec_perm_indices.
8224 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8225 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8226 or double series encoding as appropriate.
8227 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8228 vec_perm_indices instead of auto_vec_perm_indices.
8229 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8230 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8231 (vect_permute_store_chain): Likewise.
8232 (vect_grouped_load_supported): Likewise.
8233 (vect_permute_load_chain): Likewise.
8234 (vect_shift_permute_load_chain): Likewise.
8235 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8236 (vect_transform_slp_perm_load): Likewise.
8237 (vect_schedule_slp_instance): Likewise.
8238 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8239 (vectorizable_mask_load_store): Likewise.
8240 (vectorizable_bswap): Likewise.
8241 (vectorizable_store): Likewise.
8242 (vectorizable_load): Likewise.
8243 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8244 vec_perm_indices instead of auto_vec_perm_indices. Use
8245 tree_to_vec_perm_builder to read the vector from a tree.
8246 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8247 vec_perm_builder instead of a vec_perm_indices.
8248 (have_whole_vector_shift): Use vec_perm_builder and
8249 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8250 truncation to calc_vec_perm_mask_for_shift.
8251 (vect_create_epilog_for_reduction): Likewise.
8252 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8253 from auto_vec_perm_indices to vec_perm_indices.
8254 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8255 instead of changing individual elements.
8256 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8257 the vector in d.perm.
8258 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8259 from auto_vec_perm_indices to vec_perm_indices.
8260 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8261 instead of changing individual elements.
8262 (arm_vectorize_vec_perm_const): Use new_vector to install
8263 the vector in d.perm.
8264 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8265 Update vec_perm_builder constructor.
8266 (rs6000_expand_interleave): Likewise.
8267 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8268 (rs6000_expand_interleave): Likewise.
8270 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8272 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8273 to qimode could truncate the indices.
8274 * optabs.c (expand_vec_perm_var): Likewise.
8276 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8278 * Makefile.in (OBJS): Add vec-perm-indices.o.
8279 * vec-perm-indices.h: New file.
8280 * vec-perm-indices.c: Likewise.
8281 * target.h (vec_perm_indices): Replace with a forward class
8283 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8284 * optabs.h: Include vec-perm-indices.h.
8285 (expand_vec_perm): Delete.
8286 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8287 (expand_vec_perm_const): Declare.
8288 * target.def (vec_perm_const_ok): Replace with...
8289 (vec_perm_const): ...this new hook.
8290 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8291 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8292 * doc/tm.texi: Regenerate.
8293 * optabs.def (vec_perm_const): Delete.
8294 * doc/md.texi (vec_perm_const): Likewise.
8295 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8296 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8297 expand_vec_perm for constant permutation vectors. Assert that
8298 the mode of variable permutation vectors is the integer equivalent
8299 of the mode that is being permuted.
8300 * optabs-query.h (selector_fits_mode_p): Declare.
8301 * optabs-query.c: Include vec-perm-indices.h.
8302 (selector_fits_mode_p): New function.
8303 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8304 is defined, instead of checking whether the vec_perm_const_optab
8305 exists. Use targetm.vectorize.vec_perm_const instead of
8306 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8307 fit in the vector mode before using a variable permute.
8308 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8309 vec_perm_indices instead of an rtx.
8310 (expand_vec_perm): Replace with...
8311 (expand_vec_perm_const): ...this new function. Take the selector
8312 as a vec_perm_indices rather than an rtx. Also take the mode of
8313 the selector. Update call to shift_amt_for_vec_perm_mask.
8314 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8315 Use vec_perm_indices::new_expanded_vector to expand the original
8316 selector into bytes. Check whether the indices fit in the vector
8317 mode before using a variable permute.
8318 (expand_vec_perm_var): Make global.
8319 (expand_mult_highpart): Use expand_vec_perm_const.
8320 * fold-const.c: Includes vec-perm-indices.h.
8321 * tree-ssa-forwprop.c: Likewise.
8322 * tree-vect-data-refs.c: Likewise.
8323 * tree-vect-generic.c: Likewise.
8324 * tree-vect-loop.c: Likewise.
8325 * tree-vect-slp.c: Likewise.
8326 * tree-vect-stmts.c: Likewise.
8327 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8329 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8330 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8331 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8332 (aarch64_vectorize_vec_perm_const): ...this new function.
8333 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8334 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8335 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8336 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8337 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8338 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8339 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8341 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8342 check for NEON modes.
8343 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8344 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8345 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8346 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8348 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8349 the old VEC_PERM_CONST conditions.
8350 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8351 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8352 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8353 (ia64_vectorize_vec_perm_const_ok): Merge into...
8354 (ia64_vectorize_vec_perm_const): ...this new function.
8355 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8356 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8357 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8358 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8359 * config/mips/mips.c (mips_expand_vec_perm_const)
8360 (mips_vectorize_vec_perm_const_ok): Merge into...
8361 (mips_vectorize_vec_perm_const): ...this new function.
8362 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8363 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8364 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8365 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8366 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8367 (rs6000_expand_vec_perm_const): Delete.
8368 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8370 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8371 (altivec_expand_vec_perm_const_le): Take each operand individually.
8372 Operate on constant selectors rather than rtxes.
8373 (altivec_expand_vec_perm_const): Likewise. Update call to
8374 altivec_expand_vec_perm_const_le.
8375 (rs6000_expand_vec_perm_const): Delete.
8376 (rs6000_vectorize_vec_perm_const_ok): Delete.
8377 (rs6000_vectorize_vec_perm_const): New function.
8378 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8379 an element count and rtx array.
8380 (rs6000_expand_extract_even): Update call accordingly.
8381 (rs6000_expand_interleave): Likewise.
8382 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8383 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8384 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8385 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8386 (rs6000_expand_vec_perm_const): Delete.
8387 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8388 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8389 (altivec_expand_vec_perm_const_le): Take each operand individually.
8390 Operate on constant selectors rather than rtxes.
8391 (altivec_expand_vec_perm_const): Likewise. Update call to
8392 altivec_expand_vec_perm_const_le.
8393 (rs6000_expand_vec_perm_const): Delete.
8394 (rs6000_vectorize_vec_perm_const_ok): Delete.
8395 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8396 reference to the SPE evmerge intructions.
8397 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8398 an element count and rtx array.
8399 (rs6000_expand_extract_even): Update call accordingly.
8400 (rs6000_expand_interleave): Likewise.
8401 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8402 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8404 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8406 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8408 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8409 vector mode and that that mode matches the mode of the data
8411 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8412 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8413 directly using expand_vec_perm_1 when forcing selectors into
8415 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8417 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8419 * optabs-query.h (can_vec_perm_p): Delete.
8420 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8421 * optabs-query.c (can_vec_perm_p): Split into...
8422 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8423 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8424 particular selector is valid.
8425 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8426 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8427 (vect_grouped_load_supported): Likewise.
8428 (vect_shift_permute_load_chain): Likewise.
8429 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8430 (vect_transform_slp_perm_load): Likewise.
8431 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8432 (vectorizable_bswap): Likewise.
8433 (vect_gen_perm_mask_checked): Likewise.
8434 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8435 implementations of variable permutation vectors into account
8436 when deciding which selector to use.
8437 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8438 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8439 with a false third argument.
8440 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8441 to test whether the constant selector is valid and can_vec_perm_var_p
8442 to test whether a variable selector is valid.
8444 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8446 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8447 * optabs-query.c (can_vec_perm_p): Likewise.
8448 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8449 instead of vec_perm_indices.
8450 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8451 (vect_gen_perm_mask_checked): Likewise,
8452 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8453 (vect_gen_perm_mask_checked): Likewise,
8455 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8457 * optabs-query.h (qimode_for_vec_perm): Declare.
8458 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8459 (qimode_for_vec_perm): ...this new function.
8460 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8462 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8464 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8465 does not have a conditional at the top.
8467 2018-01-02 Richard Biener <rguenther@suse.de>
8469 * ipa-inline.c (big_speedup_p): Fix expression.
8471 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8474 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8477 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8481 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8482 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8483 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8484 cond_taken_branch_cost 3->4.
8486 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8488 PR tree-optimization/83581
8489 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8490 TODO_cleanup_cfg if any changes have been made.
8493 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8494 convert_modes if target mode has the right side, but different mode
8498 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8499 last argument when extracting from CONCAT. If either from_real or
8500 from_imag is NULL, use expansion through memory. If result is not
8501 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8502 the parts directly to inner mode, if even that fails, use expansion
8506 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8507 check for bswap in mode rather than HImode and use that in expand_unop
8510 Copyright (C) 2018 Free Software Foundation, Inc.
8512 Copying and distribution of this file, with or without modification,
8513 are permitted in any medium without royalty provided the copyright
8514 notice and this notice are preserved.