1 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5 INTEGER_CST offset, add it together with bitpos / 8 and
6 sign extend based on POINTER_SIZE.
10 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
13 * config/i386/constraints.md (Yc): New register constraint.
14 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
15 Use Yc constraint for alternative 2 of operand 0. Remove
16 preferred_for_speed attribute.
18 2018-03-14 Richard Biener <rguenther@suse.de>
20 PR tree-optimization/84830
21 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
22 with the old one to avoid oscillations.
24 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
27 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
29 (assign_by_spills): Return a flag of reload assignment failure.
30 Do not process the reload assignment failures. Do not spill other
31 reload pseudos if they has the same reg class. Update n if
33 (lra_assign): Add a return arg. Set up from the result of
34 assign_by_spills call.
35 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
36 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
37 usage_insns if it is not NULL.
38 (spill_hard_reg_in_range): New function.
39 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
40 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
42 (lra_assign): Change prototype.
43 * lra.c (lra): Add code to deal with fails by splitting hard reg
46 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
48 * config/riscv/riscv.opt (mrelax): New option.
49 * config/riscv/riscv.c (riscv_file_start): Emit ".option
50 "norelax" when riscv_mrelax is disabled.
51 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
53 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
56 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
57 reassociation for int modes.
59 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
61 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
62 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
64 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
65 * config/aarch64/aarch64-sve.md
66 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
67 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
68 (*extend<mode><Vwide>2): Rename to...
69 (aarch64_sve_extend<mode><Vwide>2): ...this.
70 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
71 renaming the old pattern to...
72 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
74 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
75 define_expand, renaming the old pattern to...
76 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
77 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
78 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
79 account when deciding which SVE instruction the optab should use.
80 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
82 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
84 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
85 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
86 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
87 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
88 (tlsdesc_small_<mode>): Turn a define_expand and use
89 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
90 (tlsdesc_small_advsimd_<mode>): ...this.
91 (tlsdesc_small_sve_<mode>): New pattern.
93 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
95 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
96 (UNSPEC_UMUL_HIGHPART): New constants.
97 (MUL_HIGHPART): New int iteraor.
98 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
99 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
101 (*<su>mul<mode>3_highpart): New define_insn.
103 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
106 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
109 2018-03-13 Martin Liska <mliska@suse.cz>
112 * (sem_item_optimizer::sem_item_optimizer): Initialize new
114 (sem_item_optimizer::~sem_item_optimizer): Release it.
115 (sem_item_optimizer::merge_classes): Register variable aliases.
116 (sem_item_optimizer::fixup_pt_set): New function.
117 (sem_item_optimizer::fixup_points_to_sets): Likewise.
118 * ipa-icf.h: Declare new variables and functions.
120 2018-03-13 Jakub Jelinek <jakub@redhat.com>
123 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
124 integer_pow2p@2 and test integer_pow2p in condition.
125 (A < 0 ? C : 0): Similarly for @1.
128 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
129 characters starting at p contain '\0' character, don't look beyond
133 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
134 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
137 * reg-stack.c (change_stack): Change update_end var from int to
138 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
139 also call set_block_for_insn on the newly added insns and rescan.
142 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
146 * tree.c (stabilize_reference_1): Return save_expr (e) for
147 STATEMENT_LIST even if it doesn't have side-effects.
149 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
151 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
153 2018-03-12 Renlin Li <renlin.li@arm.com>
155 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
156 aarch64_output_scalar_simd_mov_immediate.
158 2018-03-12 Martin Sebor <msebor@redhat.com>
160 PR tree-optimization/83456
161 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
162 for perfectly overlapping calls to memcpy.
163 (gimple_fold_builtin_memory_chk): Same.
164 (gimple_fold_builtin_strcpy): Handle no-warning.
165 (gimple_fold_builtin_stxcpy_chk): Same.
166 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
168 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
170 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
171 parameter. Use it for SFmode.
172 (rs6000_function_arg_advance_1): Adjust.
173 (rs6000_function_arg): Adjust.
174 (rs6000_gimplify_va_arg): Pass false for that new parameter.
176 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
178 PR rtl-optimization/84169
179 PR rtl-optimization/84780
180 * combine.c (can_combine_p): Check for a 2-insn combination whether
181 the destination register is used between the two insns, too.
183 2018-03-12 Richard Biener <rguenther@suse.de>
185 PR tree-optimization/84803
186 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
187 for refs DR analysis didn't process.
189 2018-03-12 Richard Biener <rguenther@suse.de>
191 PR tree-optimization/84777
192 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
193 force-vectorize loops ignore whether we are optimizing for size.
195 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
197 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
198 (TARGET_MD_ASM_ADJUST): Define.
200 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
201 Kito Cheng <kito.cheng@gmail.com>
202 Chung-Ju Wu <jasonwucj@gmail.com>
204 * config/nds32/nds32.c (nds32_compute_stack_frame,
205 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
206 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
207 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
208 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
209 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
210 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
211 * config/nds32/nds32.md (prologue, epilogue): Use macro
212 NDS32_V3PUSH_AVAILABLE_P to do checking.
214 2018-03-11 Jakub Jelinek <jakub@redhat.com>
217 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
218 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
219 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
220 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
221 addition of most attributes on !orig_type_die or the attribute not
222 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
224 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
225 Chung-Ju Wu <jasonwucj@gmail.com>
227 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
229 * config/nds32/nds32.opt (mvh): New option.
231 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
232 Chung-Ju Wu <jasonwucj@gmail.com>
234 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
236 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
237 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
240 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
241 Chung-Ju Wu <jasonwucj@gmail.com>
243 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
245 * config/nds32/nds32-multiple.md (strlensi): New pattern.
246 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
248 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
249 Kito Cheng <kito.cheng@gmail.com>
250 Chung-Ju Wu <jasonwucj@gmail.com>
252 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
253 UNSPEC_FFMISM and UNSPEC_FLMISM.
254 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
255 for ffb, ffmism and flmism.
256 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
257 (unspec_ffmism): Ditto.
258 (unspec_flmism): Ditto.
259 (nds32_expand_builtin_impl): Check if string extension is available.
260 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
261 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
263 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
266 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
269 * lra-assigns.c (assign_by_spills): Return a flag of reload
270 assignment failure. Do not process the reload assignment
271 failures. Do not spill other reload pseudos if they has the same
273 (lra_assign): Add a return arg. Set up from the result of
274 assign_by_spills call.
275 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
276 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
277 usage_insns if it is not NULL.
278 (spill_hard_reg_in_range): New function.
279 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
280 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
282 (lra_assign): Change prototype.
283 * lra.c (lra): Add code to deal with fails by splitting hard reg
286 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
289 * config/i386/i386.opt: Replace Enforcment with Enforcement.
291 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
294 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
295 (dw_val_node): Add val_symbolic_view.
296 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
297 (symview_upper_bound): New.
298 (new_line_info_table): Initialize symviews_since_reset.
299 (dwarf2out_source_line): Count symviews_since_reset and set
301 (dw_val_equal_p): Handle symview.
302 (add_AT_symview): New.
303 (print_dw_val): Handle symview.
304 (attr_checksum, attr_checksum_ordered): Likewise.
305 (same_dw_val_p, size_of_die): Likewise.
306 (value_format, output_die): Likewise.
307 (add_high_low_attributes): Use add_AT_symview for entry_view.
308 (dwarf2out_finish): Reset symview_upper_bound, clear
311 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
314 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
315 Add strict argument and use it.
316 (rs6000_split_multireg_move): Update for new strict argument.
317 (mem_operand_gpr): Disallow all non-offsettable addresses.
318 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
320 2018-03-09 Jakub Jelinek <jakub@redhat.com>
323 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
324 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
325 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
328 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
329 decl, use remap_type if we want to use the type.
331 2018-03-09 Martin Sebor <msebor@redhat.com>
333 PR tree-optimization/84526
334 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
336 (builtin_access::generic_overlap): Be prepared to handle non-array
339 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
341 PR rtl-optimization/84682
342 * lra-constraints.c (process_address_1): Check is_address flag
343 for address constraints.
344 (process_alt_operands): Likewise.
345 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
346 preprocess_constraints.
347 * recog.h (preprocess_constraints): Add oploc parameter.
349 * recog.c (preprocess_constraints): Test address_operand for
350 CT_ADDRESS constraints.
352 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
355 * lra-assigns.c (assign_by_spills): Return a flag of reload
356 assignment failure. Do not process the reload assignment
357 failures. Do not spill other reload pseudos if they has the same
359 (lra_assign): Add a return arg. Set up from the result of
360 assign_by_spills call.
361 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
362 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
363 usage_insns if it is not NULL.
364 (spill_hard_reg_in_range): New function.
365 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
366 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
368 (lra_assign): Change prototype.
369 * lra.c (lra): Add code to deal with fails by splitting hard reg
372 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
375 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
376 Accept complain bool parameter. Only emit errors if it is true.
377 (arm_parse_cpu_option_name): Likewise.
378 (arm_target_thumb_only): Adjust callers of the above.
379 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
380 prototype to take a default true bool parameter.
381 (arm_parse_arch_option_name): Likewise.
383 2018-03-09 David Malcolm <dmalcolm@redhat.com>
384 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
388 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
389 * configure: Regenerate.
390 * configure.ac ("linker --version-script option"): New.
391 ("linker soname option"): New.
393 2018-03-09 Richard Biener <rguenther@suse.de>
395 PR tree-optimization/84775
396 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
397 immediate uses of predicate stmts and mark them modified.
400 PR tree-optimization/84178
401 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
403 (version_loop_for_if_conversion): Delay update_ssa call.
404 (tree_if_conversion): Delay update_ssa until after predicate
407 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
410 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
411 when the function accesses prior frames.
413 2018-03-08 Jakub Jelinek <jakub@redhat.com>
416 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
417 gen_llsym, otherwise call maybe_gen_llsym.
420 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
421 has ',' character inside of it.
423 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
426 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
427 as clobbering CC_REGNUM.
429 2018-03-08 Richard Biener <rguenther@suse.de>
432 * tree-scalar-evolution.c: Include tree-into-ssa.h.
433 (follow_copies_to_constant): Do not follow SSA names registered
436 2018-03-08 Richard Biener <rguenther@suse.de>
438 PR tree-optimization/84178
439 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
441 (version_loop_for_if_conversion): Delay update_ssa call.
442 (tree_if_conversion): Delay update_ssa until after predicate
445 2018-03-08 David Malcolm <dmalcolm@redhat.com>
447 PR tree-optimization/84178
448 * tree-if-conv.c (release_bb_predicate): Remove the
449 the assertion that the stmts have NULL use_ops.
450 Discard the statements, asserting that they haven't
451 yet been added to a BB.
453 2018-03-08 Richard Biener <rguenther@suse.de>
455 PR tree-optimization/84746
456 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
457 (phi_translate): Pass in destination ANTIC_OUT set.
458 (phi_translate_1): Likewise. For a simplified result lookup
459 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
460 (phi_translate_set): Adjust.
461 (do_pre_regular_insertion): Likewise.
462 (do_pre_partial_partial_insertion): Likewise.
464 2018-03-08 Martin Liska <mliska@suse.cz>
466 PR gcov-profile/84735
467 * doc/gcov.texi: Document usage of profile files.
468 * gcov-io.h: Document changes in the format.
470 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
474 * dwarf2out.c (struct dw_line_info_table): Update comments for
476 (FORCE_RESET_NEXT_VIEW): New.
477 (FORCE_RESETTING_VIEW_P): New.
478 (RESETTING_VIEW_P): Check for -1 too.
479 (ZERO_VIEW_P): Likewise.
480 (new_line_info_table): Force-reset next view.
481 (dwarf2out_begin_function): Likewise.
482 (dwarf2out_source_line): Simplify zero_view_p initialization.
483 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
484 view directly. Omit view when omitting .loc at line 0.
486 2018-03-08 Jakub Jelinek <jakub@redhat.com>
488 PR tree-optimization/84740
489 * tree-switch-conversion.c (process_switch): Call build_constructors
490 only if info.phi_count is non-zero.
492 PR tree-optimization/84739
493 * tree-tailcall.c (find_tail_calls): Check call arguments against
494 DECL_ARGUMENTS (current_function_decl) rather than
495 DECL_ARGUMENTS (func) when checking for tail recursion.
497 2018-03-07 Jakub Jelinek <jakub@redhat.com>
499 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
500 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
501 Volker Reichelt's entry and add entries for people that perform
502 GCC fuzzy testing and report numerous bugs.
504 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
507 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
508 readonly data in sdata, if that is disabled.
509 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
510 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
511 -mreadonly-in-sdata option.
513 2018-03-07 Martin Sebor <msebor@redhat.com>
515 PR tree-optimization/84468
516 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
517 basic block when looking for nul assignment.
519 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
522 * except.h (output_function_exception_table): Adjust prototype.
523 * except.c (output_function_exception_table): Remove FNNAME parameter
524 and add SECTION parameter. Ouput one part of the table at a time.
525 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
526 the first part of the exception table and emit unwind directives.
527 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
528 (i386_pe_seh_cold_init): Likewise.
529 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
530 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
531 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
532 (ix86_output_call_insn): Emit a nop in one more case for SEH.
533 * config/i386/winnt.c: Include except.h.
534 (struct seh_frame_state): Add reg_offset, after_prologue and
535 in_cold_section fields.
536 (i386_pe_seh_end_prologue): Set seh->after_prologue.
537 (i386_pe_seh_cold_init): New function.
538 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
539 to seh->in_cold_section.
540 (seh_emit_push): Record the offset of the push.
541 (seh_emit_save): Record the offet of the save.
542 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
543 Test seh->after_prologue to disregard the epilogue.
544 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
545 (i386_pe_end_cold_function): New function.
547 2018-03-07 Jakub Jelinek <jakub@redhat.com>
550 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
551 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
554 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
556 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
557 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
560 * multiple_target.c: Include tree-inline.h and intl.h.
561 (expand_target_clones): Diagnose and fail if node->definition and
562 !tree_versionable_function_p (node->decl).
564 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
566 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
568 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
569 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
570 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
572 2018-03-06 Jakub Jelinek <jakub@redhat.com>
575 * combine.c (try_combine): Use reg_or_subregno instead of handling
576 just paradoxical SUBREGs and REGs.
578 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
580 * config/arc/arc.c (arc_finalize_pic): Remove function.
581 (arc_must_save_register): We use single base PIC register, remove
582 checks to save/restore the PIC register.
583 (arc_expand_prologue): Likewise.
584 * config/arc/arc-protos.h (arc_set_default_type_attributes):
586 (arc_verify_short): Likewise.
587 (arc_attr_type): Likewise.
588 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
589 (walk_stores): Likewise.
590 (arc_address_cost): Make it static.
591 (arc_verify_short): Likewise.
592 (branch_dest): Likewise.
593 (arc_attr_type): Likewise.
594 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
595 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
596 (arc_final_prescan_insn): Remove inserting the nops due to
597 hardware hazards. It is done in reorg step.
598 (insn_length_variant_t): Remove.
599 (insn_length_parameters_t): Likewise.
600 (arc_insn_length_parameters): Likewise.
601 (arc_get_insn_variants): Likewise.
602 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
604 2018-03-06 Jakub Jelinek <jakub@redhat.com>
607 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
610 PR tree-optimization/84687
611 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
613 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
615 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
617 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
618 Rename to ppc_speculation_barrier.
619 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
620 __builtin_ppc_speculation_barrier.
622 2018-03-05 Jakub Jelinek <jakub@redhat.com>
625 * combine.c (combine_simplify_rtx): Don't try to simplify if
626 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
629 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
631 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
632 to 32 bytes when compiling for POWER9.
634 2018-03-05 Jakub Jelinek <jakub@redhat.com>
637 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
638 regparm >= 3 with no arg reg available also for calls with
639 flag_force_indirect_call. Pass decl to ix86_function_regparm.
642 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
644 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
646 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
649 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
651 2018-03-05 Richard Biener <rguenther@suse.de>
653 PR tree-optimization/84486
654 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
655 When inserting a __builtin_assume_aligned call set the LHS
656 SSA name alignment info accordingly.
658 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
660 PR tree-optimization/84114
661 * config/aarch64/aarch64.c (aarch64_reassociation_width)
662 Avoid reassociation of FLOAT_MODE addition.
664 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
666 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
667 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
668 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
669 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
670 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
671 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
672 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
674 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
675 __builtin_ia32_wbinvd): New builtins.
676 (SPECIAL_ARGS2): New.
677 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
678 (SPECIAL_ARGS2): New.
679 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
680 (ix86_valid_target_attribute_inner_p): Ditto.
681 (ix86_init_mmx_sse_builtins): Add special_args2.
682 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
683 TARGET_WBNOINVD_P): New.
684 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
685 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
686 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
687 * config/i386/immintrin.h (_wbinvd): New intrinsic.
688 * config/i386/pconfigintrin.h: New file.
689 * config/i386/wbnoinvdintrin.h: Ditto.
690 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
691 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
693 2018-03-05 Richard Biener <rguenther@suse.de>
695 PR tree-optimization/84670
696 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
698 (BB_VISITED_WITH_VISITED_SUCCS): New define.
699 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
700 (compute_antic_aux): Only assert the number of values in ANTIC_IN
701 doesn't grow if all successors (recursively) were visited at least
704 2018-03-05 Richard Biener <rguenther@suse.de>
706 PR tree-optimization/84650
707 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
708 if executed in the loop pipeline.
710 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
712 * doc/configfiles.texi (Configuration Files): Move info about
713 conditionalizing $target-protos.h to...
714 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
715 differs from $target-protos.h.
717 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
718 Chung-Ju Wu <jasonwucj@gmail.com>
720 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
721 * config/nds32/nds32-multiple.md (setmemsi): Define.
722 * config/nds32/nds32-memory-manipulation.c
723 (nds32_gen_dup_4_byte_to_word_value): New.
724 (emit_setmem_word_loop): New.
725 (emit_setmem_byte_loop): New.
726 (nds32_expand_setmem_loop): New.
727 (nds32_expand_setmem_loop_v3m): New.
728 (nds32_expand_setmem_unroll): New.
729 (nds32_expand_setmem): New.
731 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
732 Chung-Ju Wu <jasonwucj@gmail.com>
734 * config/nds32/nds32-memory-manipulation.c
735 (nds32_emit_load_store): New.
736 (nds32_emit_post_inc_load_store): New.
737 (nds32_emit_mem_move): New.
738 (nds32_emit_mem_move_block): New.
739 (nds32_expand_movmemsi_loop_unknown_size): New.
740 (nds32_expand_movmemsi_loop_known_size): New.
741 (nds32_expand_movmemsi_loop): New.
742 (nds32_expand_movmemsi_unroll): New.
743 (nds32_expand_movmemqi): Rename ...
744 (nds32_expand_movmemsi): ... to this.
745 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
746 (movmemsi): ... to this.
747 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
748 (nds32_expand_movmemsi): ... to this.
750 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
751 Monk Chiang <sh.chiang04@gmail.com>
752 Chung-Ju Wu <jasonwucj@gmail.com>
754 * config/nds32/nds32-protos.h
755 (nds32_expand_load_multiple): New arguments.
756 (nds32_expand_store_multiple): Ditto.
757 (nds32_valid_multiple_load_store): Rename ...
758 (nds32_valid_multiple_load_store_p): ... to this.
759 * config/nds32/nds32-memory-manipulation.c
760 (nds32_expand_load_multiple): Refine implementation.
761 (nds32_expand_store_multiple): Ditto.
762 * config/nds32/nds32-multiple.md
763 (load_multiple): Update nds32_expand_load_multiple interface.
764 (store_multiple): Update nds32_expand_store_multiple interface.
765 * config/nds32/nds32-predicates.c
766 (nds32_valid_multiple_load_store): Rename ...
767 (nds32_valid_multiple_load_store_p): ... to this and refine
769 * config/nds32/predicates.md
770 (nds32_load_multiple_and_update_address_operation): New predicate.
771 (nds32_store_multiple_and_update_address_operation): New predicate.
773 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
774 Chung-Ju Wu <jasonwucj@gmail.com>
776 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
777 (combo): New attribute.
778 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
780 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
782 * config/nds32/nds32.opt: Change -mcmodel= default value.
784 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
785 Monk Chiang <sh.chiang04@gmail.com>
786 Chung-Ju Wu <jasonwucj@gmail.com>
788 * config/nds32/constants.md (unspec_element): New enum.
789 * config/nds32/constraints.md (Umw): New constraint.
790 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
791 * config/nds32/nds32-intrinsic.md: Likewise.
792 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
793 (nds32_valid_smw_lwm_base_p): New.
794 (nds32_output_smw_single_word): New.
795 (nds32_output_lmw_single_word): New.
796 (nds32_expand_unaligned_load): New.
797 (nds32_expand_unaligned_store): New.
798 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
799 (nds32_output_smw_single_word): Declare.
800 (nds32_output_lmw_single_word): Declare.
801 (nds32_expand_unaligned_load): Declare.
802 (nds32_expand_unaligned_store): Declare.
803 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
804 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
805 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
806 NDS32_BUILTIN_UASTORE_DW.
807 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
810 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
811 Kito Cheng <kito.cheng@gmail.com>
812 Chung-Ju Wu <jasonwucj@gmail.com>
814 * config/nds32/nds32-intrinsic.c
815 (nds32_expand_builtin_null_ftype_reg): Delete.
816 (nds32_expand_builtin_reg_ftype_imm): Ditto.
817 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
818 (nds32_read_argument): New.
819 (nds32_legitimize_target): Ditto.
820 (nds32_legitimize_argument): Ditto.
821 (nds32_check_constant_argument): Ditto.
822 (nds32_expand_unop_builtin): Ditto.
823 (nds32_expand_unopimm_builtin): Ditto.
824 (nds32_expand_binop_builtin): Ditto.
825 (nds32_builtin_decl_impl): Ditto.
826 (builtin_description): Ditto.
827 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
828 (nds32_init_builtins_impl): Ditto.
829 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
830 (nds32_builtin_decl): New.
831 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
832 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
834 2018-03-02 Jeff Law <law@redhat.com>
836 * reorg.c (stop_search_p): Handle DEBUG_INSN.
837 (redundant_insn, fill_simple_delay_slots): Likewise.
838 (fill_slots_from_thread): Likewise.
839 * resource.c (mark_referenced_resources): Likewise.
840 (mark_set_resources, find_dead_or_set_registers): Likewise.
842 2018-03-02 Jakub Jelinek <jakub@redhat.com>
844 * substring-locations.h (format_warning_va): Formatting fix for
846 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
848 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
849 * substring-locations.c: Include intl.h.
850 (format_warning_va): Turned into small wrapper around
851 format_warning_n_va, renamed to ...
852 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
853 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
855 (format_warning_at_substring_n): New function.
856 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
857 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
858 format_warning_at_substring with just a shorter name instead of
859 const function pointer.
860 (fmtwarn_n): New function.
861 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
862 appropriate, get rid of all the fmtstr temporaries, move conditionals
863 with G_() wrapped string literals directly into fmtwarn arguments,
864 cast dir.len to (int), formatting fixes.
866 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
868 * doc/invoke.texi: Remove "Cilk Plus" references.
870 2018-03-02 Jakub Jelinek <jakub@redhat.com>
871 Richard Biener <rguenther@suse.de>
874 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
875 for error or warning attributes if CALL_FROM_THUNK_P is set.
878 2018-03-02 Jakub Jelinek <jakub@redhat.com>
881 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
882 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
885 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
886 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
888 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
889 instead of -1U in last predictors element's probability member.
891 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
894 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
895 arguments if they are comparable.
897 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
899 PR tree-optimization/84634
900 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
901 masks and masked_loop_p with a single loop_masks, making sure it's
902 null for bb vectorization.
904 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
906 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
907 (vect_analyze_data_ref_access): Use loop->safe_len rather than
908 loop->force_vectorize to check whether there is no alias.
910 2018-03-02 Jakub Jelinek <jakub@redhat.com>
913 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
915 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
917 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
918 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
919 instead of a loop around prev_real_insn.
920 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
924 * config/i386/i386.c (ix86_print_operand): Use conditional
925 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
928 2018-03-02 Richard Biener <rguenther@suse.de>
930 PR tree-optimization/84427
931 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
932 (bitmap_set_subtract_values): Rewrite to handle multiple
935 (prune_clobbered_mems): Likewise.
936 (phi_translate): Take edge instead of pred/phiblock.
937 (phi_translate_1): Likewise.
938 (phi_translate_set): Likewise. Insert all translated
939 exprs for a value into the set, keeping possibly multiple
940 expressions per value.
941 (compute_antic_aux): Adjust for phi_translate changes.
942 When intersecting union the expressions and prune those
943 not in the final value set, keeping possibly multiple
944 expressions per value. Do not use value-insertion
945 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
946 all expressions. Add verification that the value-sets
947 only shrink during iteration.
948 (compute_partial_antic_aux): Adjust for the phi_translate changes.
949 (do_pre_regular_insertion): Likewise.
950 (do_pre_partial_partial_insertion): Likewise.
952 2018-03-02 Richard Biener <rguenther@suse.de>
955 * config/darwin.c (saved_debug_info_level): New static global.
956 (darwin_asm_lto_start): Disable debug info generation for LTO out.
957 (darwin_asm_lto_end): Restore debug info generation settings.
959 2018-03-01 Martin Liska <mliska@suse.cz>
962 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
965 2018-03-01 Richard Biener <rguenther@suse.de>
968 * dwarf2out.c (gen_variable_die): Properly handle late VLA
969 type annotation with LTO when debug was disabled at compile-time.
971 2018-03-01 Matthew Fortune <mfortune@gmail.com>
973 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
975 (mips_final_postscan_insn): Likewise.
977 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
979 PR rtl-optimization/84528
980 * alias.c (init_alias_target): Add commentary.
981 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
982 a unique base value if the frame pointer is not eliminated
983 to the stack pointer.
985 2018-03-01 Tom de Vries <tom@codesourcery.com>
987 PR rtl-optimization/83327
988 * lra-int.h (hard_regs_spilled_into): Declare.
989 * lra.c (hard_regs_spilled_into): Define.
990 (init_reg_info): Init hard_regs_spilled_into.
991 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
992 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
993 (process_bb_lives): Handle hard_regs_spilled_into.
994 (lra_create_live_ranges_1): Before doing liveness propagation, clear
995 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
997 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
999 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1000 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1001 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1002 * config/rs6000/aix72.h: New file.
1004 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1006 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1007 instead of warning_at with conditional singular and plural messages
1011 * stor-layout.c (update_alignment_for_field): For
1012 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1013 && !DECL_PACKED (field), do the alignment update, just use
1014 only desired_align instead of MAX (type_align, desired_align)
1016 (place_field): Don't do known_align < desired_align handling
1017 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1018 is non-NULL, instead do it after rli->prev_field handling and
1019 only if not within a bitfield word. For DECL_PACKED (field)
1020 use type_align of BITS_PER_UNIT.
1022 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1024 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1025 superfluous parentheses and trailing spaces.
1027 2018-02-28 Richard Biener <rguenther@suse.de>
1029 PR tree-optimization/84584
1030 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1031 SCoPs with fake exit edge.
1033 2018-02-28 Martin Liska <mliska@suse.cz>
1036 * timevar.c (timer::print): Fix format to properly print 100%
1039 2018-02-28 Richard Biener <rguenther@suse.de>
1042 * genmatch.c (capture_info::walk_match): Do not mark
1043 captured expressions without operands as expr_p given
1044 they act more like predicates and should be subject to
1045 "lost tail" side-effect preserving.
1047 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1049 PR rtl-optimization/81611
1050 * auto-inc-dec.c (attempt_change): Move dead note from
1051 mem_insn if it's the next use of regno
1052 (find_address): Take address use of reg holding
1053 non-incremented value. Add parm to limit search to the named
1055 (merge_in_block): Attempt to use a mem insn that is the next
1056 use of the original regno.
1058 2018-02-27 Martin Sebor <msebor@redhat.com>
1061 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1062 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1064 2018-02-27 Martin Sebor <msebor@redhat.com>
1066 PR translation/84207
1067 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1068 n argument to unsigned HOST_WIDE_INT.
1069 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1070 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1071 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1072 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1074 2018-02-27 Richard Biener <rguenther@suse.de>
1076 PR tree-optimization/84512
1077 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1078 Do not use the estimate returned from record_stmt_cost for
1079 the scalar iteration cost but sum properly using add_stmt_cost.
1081 2018-02-27 Richard Biener <rguenther@suse.de>
1083 PR tree-optimization/84466
1084 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1085 Adjust last change to less strictly validate use operands.
1087 2018-02-27 Martin Liska <mliska@suse.cz>
1089 PR gcov-profile/84548
1090 * gcov.c (process_file): Allow partial overlap and consider it
1091 also as group functions.
1092 (output_lines): Properly calculate range of lines for a group.
1094 2018-02-27 Martin Liska <mliska@suse.cz>
1096 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1097 'ggc' suffixes. Change first column width.
1098 (timer::print): Fix formatting of the column.
1100 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1102 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1103 preserve inline entry blocks for the sake of debug inline
1104 entry point markers alone.
1105 (remove_unused_locals): Suggest in comments a better place to
1106 force the preservation of inline entry blocks that are
1107 otherwise unused, but do not preserve them.
1109 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1111 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1113 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1116 * config/i386/constraints.md (Bs): Replace
1117 ix86_indirect_branch_register with
1118 TARGET_INDIRECT_BRANCH_REGISTER.
1120 * config/i386/i386.md (indirect_jump): Likewise.
1121 (tablejump): Likewise.
1122 (*sibcall_memory): Likewise.
1123 (*sibcall_value_memory): Likewise.
1124 Peepholes of indirect call and jump via memory: Likewise.
1125 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1126 (*sibcall_value_GOT_32): Likewise.
1127 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1128 (GOT_memory_operand): Likewise.
1129 (call_insn_operand): Likewise.
1130 (sibcall_insn_operand): Likewise.
1131 (GOT32_symbol_operand): Likewise.
1132 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1134 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1136 PR rtl-optimization/83496
1137 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1138 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1139 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1140 redundant insn, if any.
1141 (relax_delay_slots): Likewise.
1142 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1144 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1146 PR tree-optimization/83965
1147 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1148 that grouped statements are part of a reduction chain. Return
1149 true if the statement is not marked as a reduction itself but
1151 (vect_recog_dot_prod_pattern): Don't check whether the statement
1152 is part of a group here.
1153 (vect_recog_sad_pattern): Likewise.
1154 (vect_recog_widen_sum_pattern): Likewise.
1156 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1159 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1161 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1164 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1166 (ix86_output_indirect_function_return): New prototype.
1167 (ix86_split_simple_return_pop_internal): Likewise.
1168 * config/i386/i386.c (indirect_return_via_cx): New.
1169 (indirect_return_via_cx_bnd): Likewise.
1170 (indirect_thunk_name): Handle return va CX_REG.
1171 (output_indirect_thunk_function): Create alias for
1172 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1173 (ix86_output_indirect_jmp): Remove the bool argument.
1174 (ix86_output_indirect_function_return): New function.
1175 (ix86_split_simple_return_pop_internal): Likewise.
1176 * config/i386/i386.md (*indirect_jump): Don't pass false
1177 to ix86_output_indirect_jmp.
1178 (*tablejump_1): Likewise.
1179 (simple_return_pop_internal): Change it to define_insn_and_split.
1180 Call ix86_split_simple_return_pop_internal to split it for
1182 (simple_return_indirect_internal): Call
1183 ix86_output_indirect_function_return instead of
1184 ix86_output_indirect_jmp.
1186 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1189 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1190 memset and value initialization afterwards.
1192 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1194 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1196 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1199 * common/config/aarch64/aarch64-common.c
1200 (aarch_option_optimization_table[]): Switch
1201 off fomit-frame-pointer
1203 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1204 Chung-Ju Wu <jasonwucj@gmail.com>
1206 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1208 (store_multiple): Ditto.
1210 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1212 * config.gcc: Add --with-cpu support for nds32 target.
1213 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1214 * config/nds32/nds32.opt: Add -mcpu= option.
1216 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1218 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1219 isel=yes): Warn for these deprecated options.
1221 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1223 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1224 ISA_2_5_MASKS_EMBEDDED.
1226 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1228 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1229 p->max as pointers rather than using iterative_hash_expr.
1231 2018-02-23 Carl Love <cel@us.ibm.com>
1233 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1234 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1236 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1237 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1238 P8V_BUILTIN_VEC_VUNSIGNED2.
1240 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1243 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1244 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1246 * lra-constraints.c (curr_insn_transform): Set up
1247 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1249 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1250 * lra-eliminations.c (spill_pseudos): Ditto.
1251 (process_insn_for_elimination): Ditto.
1252 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1253 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1254 LRA_NON_CLOBBERED_ALT.
1256 2018-02-22 Martin Sebor <msebor@redhat.com>
1258 PR tree-optimization/84480
1259 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1260 to maybe_diag_stxncpy_trunc. Call it.
1261 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1262 from gimple_fold_builtin_strcpy. Print inlining stack.
1263 (handle_builtin_stxncpy): Print inlining stack.
1264 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1266 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1269 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1270 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1271 and -fcheck-pointer-bounds are used together.
1272 (indirect_thunk_prefix): New enum.
1273 (indirect_thunk_need_prefix): New function.
1274 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1275 "_nt" instead of "_bnd" for NOTRACK prefix.
1276 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1277 (output_indirect_thunk_function): Likewise.
1279 (ix86_code_end): Update output_indirect_thunk_function calls.
1280 (ix86_output_indirect_branch_via_reg): Replace
1281 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1282 (ix86_output_indirect_branch_via_push): Likewise.
1283 (ix86_output_function_return): Likewise.
1284 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1285 incompatible with -fcf-protection=branch and
1286 -fcheck-pointer-bounds.
1288 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1291 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1292 Change gcc_assert call to output_operand_lossage.
1294 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1296 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1298 2018-02-22 DJ Delorie <dj@redhat.com>
1299 Sebastian Perta <sebastian.perta@renesas.com>
1300 Oleg Endo <olegendo@gcc.gnu.org>
1302 * config/rx/rx.c (rx_rtx_costs): New function.
1303 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1305 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1307 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1309 2018-02-22 Martin Liska <mliska@suse.cz>
1312 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1313 Add "native" as a possible value.
1315 2018-02-22 Martin Liska <mliska@suse.cz>
1318 * config/i386/i386.c (ix86_option_override_internal):
1319 Add "native" as a possible value for -march and -mtune.
1321 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1324 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1325 to all type variants.
1327 PR tree-optimization/84503
1328 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1329 width as info->bitpos + info->bitsize - start.
1330 (merged_store_group::merge_overlapping): Simplify width computation.
1331 (check_no_overlap): New function.
1332 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1333 start + width and last_order of the group, fail if check_no_overlap
1335 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1336 to group if check_no_overlap fails.
1338 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1340 * config/rs6000/altivec.md: Delete contraint arguments to
1341 define_expand, define_split, and define_peephole2, and in
1342 define_insn_and_split if always unused.
1343 * config/rs6000/darwin.md: Ditto.
1344 * config/rs6000/dfp.md: Ditto.
1345 * config/rs6000/rs6000.md: Ditto.
1346 * config/rs6000/sync.md: Ditto.
1347 * config/rs6000/vector.md: Ditto.
1348 * config/rs6000/vsx.md: Ditto.
1350 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1352 * config/rs6000/altivec.md: Write output control strings as braced
1353 blocks instead of double-quoted strings.
1354 * config/rs6000/darwin.md: Ditto.
1355 * config/rs6000/rs6000.md: Ditto.
1356 * config/rs6000/vector.md: Ditto.
1357 * config/rs6000/vsx.md: Ditto.
1359 2018-02-21 Jason Merrill <jason@redhat.com>
1361 PR c++/84314 - ICE with templates and fastcall attribute.
1362 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1364 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1366 * ipa-cp.c (determine_versionability): Fix comment typos.
1368 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1371 * ipa-cp.c (determine_versionability): Do not version functions caling
1374 2018-02-21 Martin Liska <mliska@suse.cz>
1377 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1378 Add "native" as a possible value.
1379 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1380 the macro when native cpu detection is available.
1382 2018-02-21 Martin Liska <mliska@suse.cz>
1385 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1386 Add "native" as a possible value.
1387 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1388 when native cpu detection is available.
1390 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1391 Martin Sebor <msebor@redhat.com>
1393 PR tree-optimization/84478
1394 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1396 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1397 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1398 support which is conservatively correct, for 2 only stay conservative
1399 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1400 argument to the 2 argument get_range_strlen, adjust 6 arg
1401 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1403 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1404 (gimple_fold_builtin_strlen): Pass true as last argument to
1407 2018-02-20 Martin Sebor <msebor@redhat.com>
1410 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1411 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1412 (builtin_memref::builtin_memref): Factor out parts into
1413 set_base_and_offset and call it.
1415 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1418 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1419 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1420 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1421 search at the associated MODE_INT.
1423 2018-02-20 Jeff Law <law@redhat.com>
1426 PR tree-optimization/81592
1428 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1429 for range data rather than using global data.
1430 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1431 range data rather than using global data.
1432 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1433 pass it to children as needed.
1434 (struct directive::fmtresult): Similarly.
1435 (struct directive::set_width): Similarly.
1436 (struct directive::set_precision): Similarly.
1437 (format_integer, format_directive, parse_directive): Similarly.
1438 (format_none): Accept unnamed vr_values parameter.
1439 (format_percent, format_floating, format_character): Similarly.
1440 (format_string, format_plain): Similarly.
1441 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1442 the EVRP range analyzer for range data rather than using global data.
1443 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1444 gimple-ssa-evrp-analyze.h
1445 (class sprintf_dom_walker): Add after_dom_children member function.
1446 Add evrp_range_analyzer member.
1447 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1448 range analyzer as needed.
1449 (sprintf_dom_walker::after_dom_children): New member function.
1450 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1452 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1453 (evrp_range_analyzer::pop_to_marker): Likewise.
1455 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1457 PR tree-optimization/84419
1458 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1459 with the required type if its current type is compatible but
1462 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1465 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1466 after vectorization.
1468 2018-02-20 Martin Liska <mliska@suse.cz>
1471 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1472 possible values if we don't have a hint.
1474 2018-02-20 Martin Liska <mliska@suse.cz>
1478 * final.c (shorten_branches): Build align_tab array with one
1480 * opts.c (finish_options): Add alignment option limit check.
1481 (MAX_CODE_ALIGN): Likewise.
1482 (MAX_CODE_ALIGN_VALUE): Likewise.
1483 * doc/invoke.texi: Document maximum allowed option value for
1484 all -falign-* options.
1486 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1489 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1490 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1491 * var-tracking.c (emit_note_insn_var_location): Remove all references
1492 to NOTE_INSN_CALL_ARG_LOCATION.
1493 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1494 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1495 Use copy_rtx_if_shared.
1496 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1497 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1498 (dwarf2out_var_location): Remove handling of
1499 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1501 * final.c (final_scan_insn): Remove all references to
1502 NOTE_INSN_CALL_ARG_LOCATION.
1503 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1504 before dumping final insns.
1505 * except.c (emit_note_eh_region_end): Remove all references to
1506 NOTE_INSN_CALL_ARG_LOCATION.
1507 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1508 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1509 * config/arc/arc.c (hwloop_optimize): Likewise.
1510 * config/arm/arm.c (create_fix_barrier): Likewise.
1511 * config/s390/s390.c (s390_chunkify_start): Likewise.
1512 * config/sh/sh.c (find_barrier): Likewise.
1513 * config/i386/i386.c (rest_of_insert_endbranch,
1514 ix86_seh_fixup_eh_fallthru): Likewise.
1515 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1516 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1517 * config/frv/frv.c (frv_function_prologue): Likewise.
1518 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1520 (note_outside_basic_block_p): Remove all references to
1521 NOTE_INSN_CALL_ARG_LOCATION.
1522 * gengtype.c (adjust_field_rtx_def): Likewise.
1523 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1525 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1526 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1529 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1532 PR tree-optimization/84452
1533 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1534 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1537 2018-02-19 Martin Liska <mliska@suse.cz>
1540 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1542 2018-02-19 Martin Liska <mliska@suse.cz>
1543 Richard Sandiford <richard.sandiford@linaro.org>
1545 PR tree-optimization/82491
1546 * gimple-fold.c (get_base_constructor): Make earlier bail out
1549 2018-02-19 Carl Love <cel@us.ibm.com>
1551 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1552 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1554 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1555 P8V_BUILTIN_VEC_NEG.
1557 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1559 * config/rl78/rl78.md (movdf): New define expand.
1561 2018-02-19 Martin Liska <mliska@suse.cz>
1564 * doc/invoke.texi: Fix typo.
1565 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1567 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1569 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1570 handle rs6000_single_float and rs6000_double_float specially for
1573 2018-02-16 Jeff Law <law@redhat.com>
1575 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1578 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1581 * ipa-inline.c (inline_small_functions): Fix a typo.
1583 2018-02-16 Nathan Sidwell <nathan@acm.org>
1585 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1587 2018-02-16 Carl Love <cel@us.ibm.com>
1589 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1590 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1591 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1592 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1593 expansion to P8V_BUILTIN_VEC_FLOAT2.
1595 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1597 PR rtl-optimization/70023
1598 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1599 src_regno into account.
1601 2018-02-16 Carl Love <cel@us.ibm.com>
1603 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1604 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1605 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1606 * config/rs6000/rs6000.c: Remove case statements for
1607 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1608 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1609 and P9V_BUILTIN_VEC_VINSERT4B.
1610 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1611 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1612 * config/rs6000/vsx.md:
1613 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1616 2018-02-16 Carl Love <cel@us.ibm.com>
1618 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1620 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1622 * config/rs6000/rs6000-c.c: Add the definitions for
1623 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1624 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1625 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1626 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1627 definition for insert4b and define insn *insert3b_internal.
1628 * doc/extend.texi: Add documentation for vec_extract4b.
1630 2018-02-16 Nathan Sidwell <nathan@acm.org>
1632 * doc/extend.texi (Backwards Compatibility): Mention friend
1633 injection. Note for-scope is deprecated.
1634 * doc/invoke.texi (-ffriend-injection): Deprecate.
1636 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1638 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1639 that moved to I2, also allow destinations that are a paradoxical
1640 subreg (instead of a normal reg).
1642 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1645 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1648 2018-02-16 Richard Biener <rguenther@suse.de>
1650 PR tree-optimization/84037
1651 PR tree-optimization/84016
1653 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1654 Adjust vec_construct for the fact we need additional higher latency
1655 128bit inserts for AVX256 and AVX512 vector builds.
1656 (ix86_add_stmt_cost): Scale vector construction cost for
1659 2018-02-16 Richard Biener <rguenther@suse.de>
1661 PR tree-optimization/84417
1662 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1663 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1664 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1666 2018-02-16 Martin Liska <mliska@suse.cz>
1669 * internal-fn.def (ASAN_CHECK): Set proper flags.
1670 (ASAN_MARK): Likewise.
1672 2018-02-16 Julia Koval <julia.koval@intel.com>
1674 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1675 from PTA_CANNONLAKE.
1677 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1680 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1681 Use ++iter rather than iter++ for std::list iterators.
1682 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1683 defer deleting them until all nodes in the forest are processed. Do
1684 free even leaf nodes. Change to_process into auto_vec.
1687 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1688 * vec.h (vec_default_construct): Use memset instead of placement new
1689 if BROKEN_VALUE_INITIALIZATION is defined.
1690 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1691 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1694 PR rtl-optimization/83723
1695 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1696 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1697 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1699 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1701 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1703 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1705 PR rtl-optimization/81443
1706 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1707 from inner REGs to paradoxical SUBREGs.
1709 2018-02-16 Richard Biener <rguenther@suse.de>
1711 PR tree-optimization/84399
1712 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1713 For operands we can analyze at their definition make sure we can
1714 analyze them at each use as well.
1716 2018-02-16 Richard Biener <rguenther@suse.de>
1718 PR tree-optimization/84190
1719 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1720 volatile accesses if the decl isn't volatile.
1722 2018-02-15 Jason Merrill <jason@redhat.com>
1724 PR c++/84314 - ICE with templates and fastcall attribute.
1725 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1726 TYPE_CANONICAL on an existing type.
1728 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1730 PR tree-optimization/84383
1731 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1732 dstoff nor call operand_equal_p if dstbase is NULL.
1734 PR tree-optimization/84334
1735 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1736 also a CONSTANT_CLASS_P, punt.
1738 2018-02-14 Jim Wilson <jimw@sifive.com>
1740 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1741 first SMALL_OPERAND check. New local min_second_step. Move assert
1742 to where locals are set. Add TARGET_RVC support.
1743 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1745 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1747 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1749 2018-02-14 Martin Sebor <msebor@redhat.com>
1751 PR tree-optimization/83698
1752 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1753 arrays constrain the offset range to their bounds.
1754 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1755 (builtin_access::overlap): Avoid setting the size of overlap if it's
1757 (maybe_diag_overlap): Also consider arrays when deciding what values
1758 of offsets to include in diagnostics.
1760 2018-02-14 Martin Sebor <msebor@redhat.com>
1763 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1764 that correspond to the kind of a declaration.
1766 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1769 * config/pa/pa.md: Load address of PIC label using the linkage table
1770 if the label is nonlocal.
1772 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1774 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1775 warning message if user requests -maltivec=be.
1776 * doc/invoke.texi: Document deprecation of -maltivec=be.
1778 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1781 * config/rs6000/rs6000-c.c: Update definitions for
1782 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1783 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1785 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1788 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1789 add _get_ssp intrinsics. Remove argument from
1790 __builtin_ia32_rdssp[d|q].
1791 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1792 * config/i386/i386-builtin.def: Remove argument from
1793 __builtin_ia32_rdssp[d|q].
1794 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1795 ix86_expand_special_args_builtin for _rdssp[d|q].
1796 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1797 Clear register before usage.
1798 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1799 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1801 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1803 PR tree-optimization/84357
1804 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1805 operand 1 of an ARRAY_REF too.
1807 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1810 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1811 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1813 (set_of_reg): New struct.
1814 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1815 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1816 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1818 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1819 Split into bitclr, bitset, bitinvert patterns if appropriate.
1820 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1821 use rx_fuse_in_memory_bitop.
1822 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1823 to named insn, correct maximum insn length.
1825 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1828 * machmode.def: Define a complex mode for PARTIAL_INT.
1829 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1831 * doc/rtl.texi: Document CSPImode.
1832 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1834 (msp430_hard_regno_nregs_with_padding): Likewise.
1836 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1839 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1841 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1843 PR rtl-optimization/84169
1844 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1845 we generated a parallel as new i3 and we split that to new i2 and i3
1846 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1847 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1848 those to i2, not i1. Partially rewrite this scan code.
1850 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1853 * stor-layout.c (place_field): For variable length fields, adjust
1854 offset_align afterwards not just based on the field's alignment,
1855 but also on the size.
1858 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1859 of exps and logs in the use_exp2 case.
1861 2018-02-13 Jeff Law <law@redhat.com>
1863 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1866 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1869 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1873 * common.opt (gas-loc-support, gas-locview-support): New.
1874 (ginline-points, ginternal-reset-location-views): New.
1875 * doc/invoke.texi: Document them. Use @itemx where intended.
1876 (gvariable-location-views): Adjust.
1877 * target.def (reset_location_view): New.
1878 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1879 (TARGET_RESET_LOCATION_VIEW): New.
1880 * doc/tm.texi: Rebuilt.
1881 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1882 (dwarf2out_default_as_locview_support): New.
1883 (output_asm_line_debug_info): Use option variables.
1884 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1885 (output_loc_list): Likewise.
1886 (add_high_low_attributes): Check option variables.
1887 Don't output entry view attribute in strict mode.
1888 (gen_inlined_subroutine_die): Check option variables.
1889 (dwarf2out_inline_entry): Likewise.
1890 (init_sections_and_labels): Likewise.
1891 (dwarf2out_early_finish): Likewise.
1892 (maybe_reset_location_view): New, from...
1893 (dwarf2out_var_location): ... here. Call it.
1894 * debug.h (dwarf2out_default_as_loc_support): Declare.
1895 (dwarf2out_default_as_locview_support): Declare.
1896 * hooks.c (hook_int_rtx_insn_0): New.
1897 * hooks.h (hook_int_rtx_insn_0): Declare.
1898 * toplev.c (process_options): Take -gas-loc-support and
1899 -gas-locview-support from dwarf2out. Enable
1900 -gvariable-location-views by default only with locview
1901 assembler support. Enable -ginternal-reset-location-views by
1902 default only if the target defines the corresponding hook.
1903 Enable -ginline-points by default if location views are
1904 enabled; force it disabled if statement frontiers are
1906 * tree-inline.c (expand_call_inline): Check option variables.
1907 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1909 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1911 PR tree-optimization/84321
1912 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1913 handling. Also check whether the anti-range contains any values
1914 that satisfy the mask; switch to a VR_RANGE if not.
1916 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1919 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1921 2018-02-13 Martin Jambor <mjambor@suse.cz>
1924 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1925 of call statements, also set location of a load to a temporary.
1927 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1929 * config/rl78/rl78.c (add_vector_labels): New function.
1930 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1931 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1932 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1933 which checks that no arguments are passed.
1934 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1935 * doc/extend.texi: Documentation for the new attribute.
1937 2018-02-13 Andreas Schwab <schwab@suse.de>
1939 * config/riscv/linux.h (CPP_SPEC): Define.
1941 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1944 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1945 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1946 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1947 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1948 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1949 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1950 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1951 temporarily for AES and PCLMUL builtins.
1953 PR tree-optimization/84339
1954 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1955 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1959 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1960 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1961 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1963 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1965 * omp-simd-clone.h: New file.
1966 * omp-simd-clone.c: Include omp-simd-clone.h.
1967 (expand_simd_clones): No longer static.
1968 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1969 cgraph.h and omp-simd-clone.h.
1970 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1971 (vect_recog_widen_shift_pattern): Formatting fix.
1972 (vect_pattern_recog_1): Don't check optab for calls.
1975 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1976 operands[2] into a REG before using gen_lowpart on it.
1978 2018-02-12 Jeff Law <law@redhat.com>
1981 * config/sh/sh.c (find_barrier): Consider a sibling call
1984 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1985 successfully back substituting a reg.
1987 2018-02-12 Richard Biener <rguenther@suse.de>
1989 PR tree-optimization/84037
1990 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1991 parameter, move visited init to caller.
1992 (vect_slp_analyze_operations): Separate cost from validity
1993 check, initialize visited once for all instances.
1994 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1996 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1998 (vect_model_promotion_demotion_cost): Likewise.
1999 (vectorizable_bswap): Guard cost modeling with !slp_node
2000 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2002 (vectorizable_call): Likewise.
2003 (vectorizable_conversion): Likewise.
2004 (vectorizable_assignment): Likewise.
2005 (vectorizable_shift): Likewise.
2006 (vectorizable_operation): Likewise.
2007 (vectorizable_store): Likewise.
2008 (vectorizable_load): Likewise.
2009 (vectorizable_condition): Likewise.
2010 (vectorizable_comparison): Likewise.
2012 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2015 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2016 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2017 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2019 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2022 * params.def (inline-min-speedup): Increase from 8 to 15.
2023 (max-inline-insns-auto): Decrease from 40 to 30.
2024 * ipa-split.c (consider_split): Add some buffer for function to
2025 be considered inlining candidate.
2026 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2029 2018-02-12 Richard Biener <rguenther@suse.de>
2031 PR tree-optimization/84037
2032 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2033 matched stmts if we cannot swap the non-matched ones.
2035 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2037 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2038 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2039 _mm_maskz_scalef_round_ss): New intrinsics.
2040 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2041 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2042 __builtin_ia32_scalefss_round): Remove.
2043 (__builtin_ia32_scalefsd_mask_round,
2044 __builtin_ia32_scalefss_mask_round): New intrinsics.
2045 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2046 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2047 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2048 "<round_constraint>")): Changed to ...
2049 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2050 "<round_scalar_constraint>")): ... this.
2051 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2052 %0, %1, %2<round_op3>}"): Changed to ...
2053 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2054 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2055 %2<round_scalar_mask_op3>}"): ... this.
2056 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2058 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2060 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2061 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2062 (_mm_maskz_sqrt_round_ss): New intrinsics.
2063 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2064 (__builtin_ia32_sqrtsd_mask_round)
2065 (__builtin_ia32_sqrtss_mask_round): New builtins.
2066 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2067 (__builtin_ia32_sqrtss_round): Remove.
2068 (__builtin_ia32_sqrtsd_mask_round)
2069 (__builtin_ia32_sqrtss_mask_round): New builtins.
2070 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2071 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2072 ((match_operand:VF_128 1 "vector_operand"
2073 "xBm,<round_constraint>")): Changed to ...
2074 ((match_operand:VF_128 1 "vector_operand"
2075 "xBm,<round_scalar_constraint>")): ... this.
2076 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2077 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2078 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2079 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2080 %<iptr>1<round_scalar_mask_op3>}): ... this.
2081 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2082 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2084 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2087 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2088 Cast vec_cmpeq result to correct type.
2089 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2090 Cast vec_cmpgt result to correct type.
2092 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2094 * final.c (final_scan_insn_1): Renamed from...
2095 (final_scan_insn): ... this. New wrapper, to recover
2096 seen from the outermost call in recursive ones.
2097 * config/sparc/sparc.c (output_return): Drop seen from call.
2098 (output_sibcall): Likewise.
2099 * config/visium/visium.c (output_branch): Likewise.
2101 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2103 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2106 2018-02-10 Alan Modra <amodra@gmail.com>
2109 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2110 Specify LR as an input.
2112 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2115 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2116 remove_member_access_dummy_vars): New functions.
2117 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2118 lower_omp_1, execute_lower_omp): Use them.
2120 PR rtl-optimization/84308
2121 * shrink-wrap.c (spread_components): Release todo vector.
2123 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2125 PR rtl-optimization/57193
2126 * ira-color.c (struct allocno_color_data): Add member
2127 conflict_allocno_hard_prefs.
2128 (update_conflict_allocno_hard_prefs): New.
2129 (bucket_allocno_compare_func): Add a preference based on
2130 conflict_allocno_hard_prefs.
2131 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2132 (color_allocnos): Remove a dead code. Initiate
2133 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2135 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2138 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2139 constraint from =wa to wa. Avoid a subreg on the output operand,
2140 instead use a pseudo and subreg it in a move.
2141 (p9_xxbrd_<mode>): Changed to ...
2142 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2143 (p9_xxbrd_v2df): New expander.
2144 (p9_xxbrw_<mode>): Changed to ...
2145 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2146 (p9_xxbrw_v4sf): New expander.
2148 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2150 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2152 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2155 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2156 multiply in 32-bit mode.
2157 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2158 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2161 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2163 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2164 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2165 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2166 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2168 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2171 * dwarf2out.c (is_trivial_indirect_ref): New function.
2172 (dwarf2out_late_global_decl): Do not generate a location
2173 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2174 and that are not defined in the current unit.
2176 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2178 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2179 instead of a libcall for UNORDERED.
2181 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2184 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2185 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2187 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2190 * config/s390/s390.c (s390_set_current_function): Invoke
2191 s390_indirect_branch_settings also if fndecl didn't change.
2193 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2195 * config/rs6000/rs6000.md (blockage): Set length to zero.
2197 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2199 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2201 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2204 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2205 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2209 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2210 PARALLEL incoming that failed vt_get_decl_and_offset check.
2213 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2214 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2216 (get_variable_section): For decls in named .bss* sections pass true as
2217 second argument to bss_initializer_p.
2219 2018-02-09 Marek Polacek <polacek@redhat.com>
2220 Jakub Jelinek <jakub@redhat.com>
2223 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2224 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2225 Sync some changes from cxx_fold_indirect_ref.
2227 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2229 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2231 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2232 (BLOCK_INLINE_ENTRY_LABEL): New.
2233 (dwarf2out_var_location): Disregard inline entry markers.
2234 (inline_entry_data): New struct.
2235 (inline_entry_data_hasher): New hashtable type.
2236 (inline_entry_data_hasher::hash): New.
2237 (inline_entry_data_hasher::equal): New.
2238 (inline_entry_data_table): New variable.
2239 (add_high_low_attributes): Add DW_AT_entry_pc and
2240 DW_AT_GNU_entry_view attributes if a pending entry is found
2241 in inline_entry_data_table. Add old entry_pc attribute only
2242 if debug nonbinding markers are disabled.
2243 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2244 markers are enabled.
2245 (block_within_block_p, dwarf2out_inline_entry): New.
2246 (dwarf2out_finish): Check that no entries remained in
2247 inline_entry_data_table.
2248 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2249 (final_scan_insn, notice_source_line): Likewise.
2250 (rest_of_clean_state): Skip inline entry markers.
2251 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2253 * gimple.c (gimple_build_debug_inline_entry): New.
2254 * gimple.h (enum gimple_debug_subcode): Add
2255 GIMPLE_DEBUG_INLINE_ENTRY.
2256 (gimple_build_debug_inline_entry): Declare.
2257 (gimple_debug_inline_entry_p): New.
2258 (gimple_debug_nonbind_marker_p): Adjust.
2259 * insn-notes.def (INLINE_ENTRY): New.
2260 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2261 inline entry marker notes.
2262 (print_insn): Likewise.
2263 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2264 (INSN_DEBUG_MARKER_KIND): Likewise.
2265 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2266 * tree-inline.c (expand_call_inline): Build and insert
2267 debug_inline_entry stmt.
2268 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2269 inline entry blocks early, if nonbind markers are enabled.
2270 (dump_scope_block): Dump fragment info.
2271 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2272 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2273 (gimple_build_debug_inline_entry): New.
2274 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2275 Enable/disable inline entry points too.
2276 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2277 (DEBUG_INSN): Describe inline entry markers.
2279 * common.opt (gvariable-location-views): New.
2280 (gvariable-location-views=incompat5): New.
2281 * config.in: Rebuilt.
2282 * configure: Rebuilt.
2283 * configure.ac: Test assembler for view support.
2284 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2285 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2286 * dwarf2out.c (var_loc_view): New typedef.
2287 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2288 (dwarf2out_locviews_in_attribute): New.
2289 (dwarf2out_locviews_in_loclist): New.
2290 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2291 (enum dw_line_info_opcode): Add LI_adv_address.
2292 (struct dw_line_info_table): Add view.
2293 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2294 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2295 (zero_view_p): New variable.
2296 (ZERO_VIEW_P): New macro.
2297 (output_asm_line_debug_info): New.
2298 (struct var_loc_node): Add view.
2299 (add_AT_view_list, AT_loc_list): New.
2300 (add_var_loc_to_decl): Add view param. Test it against last.
2301 (new_loc_list): Add view params. Record them.
2302 (AT_loc_list_ptr): Handle loc and view lists.
2303 (view_list_to_loc_list_val_node): New.
2304 (print_dw_val): Handle dw_val_class_view_list.
2305 (size_of_die): Likewise.
2306 (value_format): Likewise.
2307 (loc_list_has_views): New.
2308 (gen_llsym): Set vl_symbol too.
2309 (maybe_gen_llsym, skip_loc_list_entry): New.
2310 (dwarf2out_maybe_output_loclist_view_pair): New.
2311 (output_loc_list): Output view list or entries too.
2312 (output_view_list_offset): New.
2313 (output_die): Handle dw_val_class_view_list.
2314 (output_dwarf_version): New.
2315 (output_compilation_unit_header): Use it.
2316 (output_skeleton_debug_sections): Likewise.
2317 (output_rnglists, output_line_info): Likewise.
2318 (output_pubnames, output_aranges): Update version comments.
2319 (output_one_line_info_table): Output view numbers in asm comments.
2320 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2321 Call maybe_gen_llsym.
2322 (loc_list_from_tree_1): Adjust.
2323 (add_AT_location_description): Create view list attribute if
2324 needed, check it's absent otherwise.
2325 (convert_cfa_to_fb_loc_list): Adjust.
2326 (maybe_emit_file): Call output_asm_line_debug_info for test.
2327 (dwarf2out_var_location): Reset views as needed. Precompute
2328 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2329 attribute. Set view.
2330 (new_line_info_table): Reset next view.
2331 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2332 (dwarf2out_source_line): Likewise. Output view resets and labels to
2333 the assembler, or select appropriate line info opcodes.
2334 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2335 (optimize_string_length): Catch it. Adjust.
2336 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2337 dw_val_class_view_list, and remove it if no longer needed.
2338 (hash_loc_list): Hash view numbers.
2339 (loc_list_hasher::equal): Compare them.
2340 (optimize_location_lists): Check whether a view list symbol is
2341 needed, and whether the locview attribute is present, and
2342 whether they match. Remove the locview attribute if no longer
2344 (index_location_lists): Call skip_loc_list_entry for test.
2345 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2346 Use output_dwarf_version.
2347 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2348 (struct dw_val_node): Add val_view_list.
2349 * final.c (SEEN_NEXT_VIEW): New.
2350 (set_next_view_needed): New.
2351 (clear_next_view_needed): New.
2352 (maybe_output_next_view): New.
2353 (final_start_function): Rename to...
2354 (final_start_function_1): ... this. Take pointer to FIRST,
2355 add SEEN parameter. Emit param bindings in the initial view.
2356 (final_start_function): Reintroduce SEEN-less interface.
2357 (final): Rename to...
2358 (final_1): ... this. Take SEEN parameter. Output final pending
2359 next view at the end.
2360 (final): Reintroduce seen-less interface.
2361 (final_scan_insn): Output pending next view before switching
2362 sections or ending a block. Mark the next view as needed when
2363 outputting variable locations. Notify debug backend of section
2364 changes, and of location view changes.
2365 (rest_of_handle_final): Adjust.
2366 * toplev.c (process_options): Autodetect value for debug variable
2367 location views option. Warn on incompat5 without -gdwarf-5.
2368 * doc/invoke.texi (gvariable-location-views): New.
2369 (gvariable-location-views=incompat5): New.
2370 (gno-variable-location-views): New.
2372 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2374 PR tree-optimization/84136
2375 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2376 that the result of find_edge is non-NULL.
2378 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2381 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2382 storing integer register in SImode. Fix cost of 256 and 512
2383 byte aligned SSE register store.
2385 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2387 * config/i386/i386.c (ix86_multiplication_cost): Fix
2388 multiplication cost for TARGET_AVX512DQ.
2390 2018-02-08 Marek Polacek <polacek@redhat.com>
2392 PR tree-optimization/84238
2393 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2396 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2398 PR tree-optimization/84265
2399 * tree-vect-stmts.c (vectorizable_store): Don't treat
2400 VMAT_CONTIGUOUS accesses as grouped.
2401 (vectorizable_load): Likewise.
2403 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2405 PR tree-optimization/81635
2406 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2407 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2408 (test_round_for_mask): New functions.
2409 (wide_int_cc_tests): Call test_round_for_mask.
2410 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2411 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2412 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2413 range returned by get_range_info.
2415 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2418 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2419 * symtab.c: Include builtins.h
2420 (symtab_node::output_to_lto_symbol_table_p): Move here
2421 from lto-streamer-out.c:output_symbol_p.
2422 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2423 (output_symbol_p): Move all logic to symtab.c
2424 (produce_symtab): Update.
2426 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2428 * config/s390/s390-opts.h (enum indirect_branch): Define.
2429 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2430 (s390_indirect_branch_via_thunk)
2431 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2432 (enum s390_indirect_branch_type): Define.
2433 * config/s390/s390.c (struct s390_frame_layout, struct
2434 machine_function): Remove.
2435 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2436 (indirect_branch_table_label_no, indirect_branch_table_name):
2438 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2439 (enum s390_indirect_branch_option): Define.
2440 (s390_return_addr_from_memory): New function.
2441 (s390_handle_string_attribute): New function.
2442 (s390_attribute_table): Add new attribute handler.
2443 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2444 (s390_indirect_branch_via_thunk): New function.
2445 (s390_indirect_branch_via_inline_thunk): New function.
2446 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2447 sibling call optimization for non z10 compiles.
2448 (s390_emit_call): Force indirect branch target to be a single
2449 register. Add r1 clobber for non-z10 compiles.
2450 (s390_emit_epilogue): Emit return jump via return_use expander.
2451 (s390_reorg): Handle JUMP_INSNs as execute targets.
2452 (s390_option_override_internal): Perform validity checks for the
2453 new command line options.
2454 (s390_indirect_branch_attrvalue): New function.
2455 (s390_indirect_branch_settings): New function.
2456 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2457 (s390_output_indirect_thunk_function): New function.
2458 (s390_code_end): Implement target hook.
2459 (s390_case_values_threshold): Implement target hook.
2460 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2462 * config/s390/s390.h (struct s390_frame_layout)
2463 (struct machine_function): Move here from s390.c.
2464 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2465 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2466 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2467 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2468 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2469 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2470 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2471 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2472 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2473 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2474 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2475 (mnemonic attribute): Add values which aren't recognized
2477 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2478 pattern for branch conversion. Fix mnemonic attribute.
2479 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2480 indirect branch via thunk if requested.
2481 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2482 ("*indirect_jump"): Disable for branch conversion using out of
2484 ("indirect_jump_via_thunk<mode>_z10")
2485 ("indirect_jump_via_thunk<mode>")
2486 ("indirect_jump_via_inlinethunk<mode>_z10")
2487 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2488 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2489 ("casesi_jump_via_inlinethunk<mode>_z10")
2490 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2491 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2492 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2493 ("*indirect2_jump"): Disable for branch conversion.
2494 ("casesi_jump"): Turn into expander and expand patterns for branch
2496 ("return_use"): New expander.
2497 ("*return"): Emit return via thunk and rename it to ...
2498 ("*return<mode>"): ... this one.
2499 * config/s390/s390.opt: Add new options and and enum for the
2502 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2504 * lra-constraints.c (match_reload): Unconditionally use
2505 gen_lowpart_SUBREG, rather than selecting between that
2506 and equivalent gen_rtx_SUBREG code.
2508 2018-02-08 Richard Biener <rguenther@suse.de>
2510 PR tree-optimization/84233
2511 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2512 changed flag instead of boguously re-using phi_inserted.
2514 2018-02-08 Martin Jambor <mjambor@suse.cz>
2516 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2517 static local variables.
2519 2018-02-08 Richard Biener <rguenther@suse.de>
2521 PR tree-optimization/84278
2522 * tree-vect-stmts.c (vectorizable_store): When looking for
2523 smaller vector types to perform grouped strided loads/stores
2524 make sure the mode is supported by the target.
2525 (vectorizable_load): Likewise.
2527 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2529 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2530 Increase LDP/STP opportunities by adding adjacent callee-saves.
2532 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2534 PR rtl-optimization/84068
2535 PR rtl-optimization/83459
2536 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2538 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2540 PR tree-optimization/84224
2541 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2542 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2545 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2548 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2549 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2550 count, remove one USE.
2552 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2554 * doc/install.texi (Configuration): Document the
2555 --with-long-double-format={ibm,ieee} PowerPC configuration
2559 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2560 Convert from define_expand to be define_insn_and_split. Rework
2561 float/double/_Float128 conversions to QI/HI/SImode to work with
2562 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2563 conversions to QI/HImode types did a store and then a load to
2564 truncate the value. For conversions to VSX registers, don't split
2565 the insn, instead emit the code directly. Use the code iterator
2566 any_fix to combine signed and unsigned conversions.
2567 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2568 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2569 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2570 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2571 (fix_<mode>di2_hw): Likewise.
2572 (fixuns_<mode>di2_hw): Likewise.
2573 (fix_<mode>si2_hw): Likewise.
2574 (fixuns_<mode>si2_hw): Likewise.
2575 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2576 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2577 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2578 fix<uns>_trunc<SFDF:mode>si2_p8.
2579 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2581 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2582 (fix<uns>_<mode>_mem): Likewise.
2583 (fctiw<u>z_<mode>_mem): Likewise.
2584 (fix<uns>_<mode>_mem): Likewise.
2585 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2586 the register allocator from doing a direct move to the GPRs to do
2587 a store, and instead use the ISA 3.0 store byte/half-word from
2588 vector register instruction. For IEEE 128-bit floating point,
2589 also optimize stores of 32-bit ints.
2590 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2592 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2594 * genextract.c (push_pathstr_operand): New function to support
2596 (walk_rtx): Call push_pathstr_operand.
2597 (print_path): Support [a-zA-Z].
2599 2018-02-07 Richard Biener <rguenther@suse.de>
2601 PR tree-optimization/84037
2602 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2603 (cse_and_gimplify_to_preheader): Declare.
2604 (vect_get_place_in_interleaving_chain): Likewise.
2605 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2607 (_loop_vec_info::~_loop_vec_info): Delete it.
2608 (cse_and_gimplify_to_preheader): New function.
2609 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2610 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2611 (vectorizable_load): Likewise. For grouped stores always base
2612 the IV on the first element.
2613 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2614 condition before gimplifying.
2616 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2618 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2619 *DIV_EXPR and *MOD_EXPR.
2621 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2624 * config/i386/i386.c (ix86_option_override_internal): Mask out
2625 the CF_SET bit when checking -fcf-protection.
2627 2018-02-07 Tom de Vries <tom@codesourcery.com>
2630 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2633 2018-02-07 Richard Biener <rguenther@suse.de>
2635 PR tree-optimization/84204
2636 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2639 PR tree-optimization/84205
2640 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2641 special-case isl_ast_op_zdiv_r.
2643 PR tree-optimization/84223
2644 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2645 Only add conditions from within the region.
2646 (gather_bbs::after_dom_children): Adjust.
2648 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2651 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2652 * config/avr/avr.md: Only post-reload split REG-REG moves if
2653 either register is GENERAL_REG_P.
2655 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2657 PR tree-optimization/84235
2658 * tree-ssa-scopedtables.c
2659 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2660 if the subtraction is performed in floating point type where NaNs are
2661 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2662 build 1. Formatting fix.
2664 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2667 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2668 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2669 and skip it regardless of bb boundaries. Use CALL_P macro,
2670 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2671 unnecessarily, formatting fix.
2673 2018-02-06 Michael Collison <michael.collison@arm.com>
2675 * config/arm/thumb2.md:
2676 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2677 (*thumb_mov_notscc): Ditto.
2679 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2682 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2683 unsigned_fix, not "s".
2685 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2687 * configure.ac (gcc_fn_eh_frame_ro): New function.
2688 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2689 correct .eh_frame permissions.
2690 * configure: Regenerate.
2692 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2694 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2697 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2699 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2700 Display warning message for -mno-speculate-indirect-jumps.
2702 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2704 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2706 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2708 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2710 PR tree-optimization/84225
2711 * tree-eh.c (find_trapping_overflow): Only call
2712 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2714 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2717 * config/i386/i386.c: Reimplement the check of possible options
2718 -mibt/-mshstk conbination. Change error messages.
2719 * doc/invoke.texi: Fix a typo: remove extra '='.
2721 2018-02-06 Marek Polacek <polacek@redhat.com>
2723 PR tree-optimization/84228
2724 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2726 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2729 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2730 emitted arch directives.
2731 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2732 __ARM_FEATURE_COPROC before changing architectures.
2734 2018-02-06 Richard Biener <rguenther@suse.de>
2736 * config/i386/i386.c (print_reg): Fix typo.
2737 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2739 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2741 * configure: Regenerate.
2743 2018-02-05 Martin Sebor <msebor@redhat.com>
2745 PR tree-optimization/83369
2746 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2749 2018-02-05 Martin Liska <mliska@suse.cz>
2751 * doc/invoke.texi: Cherry-pick upstream r323995.
2753 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2755 * ira.c (ira_init_register_move_cost): Adjust comment.
2757 2018-02-05 Martin Liska <mliska@suse.cz>
2759 PR gcov-profile/84137
2760 * doc/gcov.texi: Fix typo in documentation.
2762 2018-02-05 Martin Liska <mliska@suse.cz>
2764 PR gcov-profile/83879
2765 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2766 using dlopen functionality.
2768 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2770 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2771 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2772 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2773 _mm_maskz_range_round_ss): New intrinsics.
2774 (__builtin_ia32_rangesd128_round)
2775 (__builtin_ia32_rangess128_round): Remove.
2776 (__builtin_ia32_rangesd128_mask_round,
2777 __builtin_ia32_rangess128_mask_round): New builtins.
2778 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2779 __builtin_ia32_rangess128_round): Remove.
2780 (__builtin_ia32_rangesd128_mask_round,
2781 __builtin_ia32_rangess128_mask_round): New builtins.
2782 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2783 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2784 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2785 "<round_saeonly_constraint>")): Changed to ...
2786 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2787 "<round_saeonly_scalar_constraint>")): ... this.
2788 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2789 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2790 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2791 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2792 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2794 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2796 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2798 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2799 Remove all values except native, 8540 and 8548.
2801 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2803 * config/i386/i386.c (ix86_output_function_return): Pass
2804 INVALID_REGNUM, instead of -1, as invalid register number to
2805 indirect_thunk_name and output_indirect_thunk.
2807 2018-02-02 Julia Koval <julia.koval@intel.com>
2809 * config.gcc: Add -march=icelake.
2810 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2811 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2812 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2813 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2814 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2815 (processor_target_table): Add icelake.
2816 (ix86_option_override_internal): Handle new PTAs.
2817 (get_builtin_code_for_version): Handle icelake.
2818 (M_INTEL_COREI7_ICELAKE): New.
2819 (fold_builtin_cpu): Handle icelake.
2820 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2821 * doc/invoke.texi: Add -march=icelake.
2823 2018-02-02 Julia Koval <julia.koval@intel.com>
2825 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2826 to wide_int_bitmask.
2827 * wide-int-bitmask.h: New.
2829 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2832 * config/i386/i386.md: Replace Pmode with word_mode in
2833 builtin_setjmp_setup and builtin_longjmp to support x32.
2835 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2839 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2841 (rs6000_supported_cpu_names): New static variable.
2842 (linux_cpu_translation_table): Likewise.
2843 (elf_platform) <cpu>: Define new static variable and use it.
2844 Translate kernel AT_PLATFORM name to canonical name if needed.
2845 Error if platform name is unknown.
2847 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2850 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2852 2018-02-01 Jeff Law <law@redhat.com>
2855 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2856 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2857 the scratch if RELEASE_VIA_POP is false.
2858 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2859 If we have to save a temporary register, decrement SIZE appropriately.
2860 Pass new arguments to release_scratch_register_on_entry.
2861 (ix86_adjust_stack_and_probe): Likewise.
2862 (ix86_emit_probe_stack_range): Pass new arguments to
2863 release_scratch_register_on_entry.
2865 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2867 PR rtl-optimization/84157
2868 * combine.c (change_zero_ext): Use REG_P predicate in
2869 front of HARD_REGISTER_P predicate.
2871 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2873 * config/avr/avr.c (avr_option_override): Move disabling of
2874 -fdelete-null-pointer-checks to...
2875 * common/config/avr/avr-common.c (avr_option_optimization_table):
2878 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2880 PR tree-optimization/81635
2881 * tree-data-ref.c (split_constant_offset_1): For types that
2882 wrap on overflow, try to use range info to prove that wrapping
2885 2018-02-01 Renlin Li <renlin.li@arm.com>
2888 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2890 (aarch64_register_move_cost): Likewise.
2891 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2893 (REG_CLASS_NAMES): Likewise.
2894 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2895 TAILCALL_ADDR_REGS. Remove IP registers.
2896 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2898 2018-02-01 Richard Biener <rguenther@suse.de>
2900 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2901 for specifying RPO order and allow NULL for that.
2902 * domwalk.c (dom_walker::dom_walker): Likewise.
2903 (dom_walker::walk): Handle NULL RPO order.
2904 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2906 (rewrite_update_dom_walker): Likewise.
2907 (mark_def_dom_walker): Likewise.
2909 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2911 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2912 (aarch64_maybe_expand_sve_subreg_move): Declare.
2913 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2914 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2916 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2917 that are semantically a reverse operation.
2918 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2919 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2920 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2922 (aarch64_can_change_mode_class): For big-endian, forbid changes
2923 between two SVE modes if they have different element sizes.
2925 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2927 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2928 the TImode handling for big-endian targets.
2930 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2932 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2933 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2935 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2936 Remove BSWAP handing for big-endian targets and use the form of
2937 LD1RQ appropariate for the mode.
2939 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2941 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2942 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2945 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2948 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2949 check for operands that need to go through aarch64_sve_reload_be.
2951 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2953 PR tree-optimization/81661
2954 PR tree-optimization/84117
2955 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2956 * tree-eh.c: Include gimplify.h.
2957 (find_trapping_overflow, replace_trapping_overflow,
2958 rewrite_to_non_trapping_overflow): New functions.
2959 * tree-vect-loop.c: Include tree-eh.h.
2960 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2961 * tree-data-ref.c: Include tree-eh.h.
2962 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2964 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2966 PR rtl-optimization/84123
2967 * combine.c (change_zero_ext): Check if hard register satisfies
2968 can_change_dest_mode before calling gen_lowpart_SUBREG.
2970 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2973 * ira.c (ira_init_register_move_cost): Remove assert.
2975 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2977 PR rtl-optimization/84071
2978 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2979 * doc/tm.texi: Regenerate.
2981 2018-01-31 Richard Biener <rguenther@suse.de>
2983 PR tree-optimization/84132
2984 * tree-data-ref.c (analyze_miv_subscript): Properly
2985 check whether evolution_function_is_affine_multivariate_p
2986 before calling gcd_of_steps_may_divide_p.
2988 2018-01-31 Julia Koval <julia.koval@intel.com>
2991 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2992 * config/i386/i386.md (rdpid_rex64) New.
2993 (rdpid): Make 32bit only.
2995 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2998 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2999 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3001 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3004 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3006 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3008 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3010 PR rtl-optimization/84071
3011 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3012 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3014 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3016 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3017 (arc_attribute_table): Add 'aux' attribute.
3018 (arc_in_small_data_p): Consider aux like variables.
3019 (arc_is_aux_reg_p): New function.
3020 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3021 (arc_get_aux_arg): New function.
3022 (prepare_move_operands): Handle aux-register access.
3023 (arc_handle_aux_attribute): New function.
3024 * doc/extend.texi (ARC Variable attributes): Add subsection.
3026 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3028 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3029 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3030 (arc_attribute_table): Add 'uncached' attribute.
3031 (arc_print_operand): Print '.di' flag for uncached memory
3033 (arc_in_small_data_p): Do not consider for small data the uncached
3035 (arc_is_uncached_mem_p): New function.
3036 * config/arc/predicates.md (compact_store_memory_operand): Check
3037 for uncached memory accesses.
3038 (nonvol_nonimm_operand): Likewise.
3039 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3041 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3044 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3045 falign-loops=): Add Optimization flag.
3047 2018-01-30 Jeff Law <law@redhat.com>
3050 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3051 INT_REGISTERS_SAVED. Check it prior to calling
3052 get_scratch_register_on_entry.
3053 (ix86_adjust_stack_and_probe): Similarly.
3054 (ix86_emit_probe_stack_range): Similarly.
3055 (ix86_expand_prologue): Corresponding changes.
3057 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3060 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3061 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3063 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3066 * lra-constraints.c (curr_insn_transform): Process AND in the
3069 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3071 PR rtl-optimization/83986
3072 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3073 dependence against last_pending_memory_flush in addition to
3076 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3078 PR tree-optimization/81611
3079 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3082 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3085 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3088 2018-01-30 Richard Biener <rguenther@suse.de>
3089 Jakub Jelinek <jakub@redhat.com>
3091 PR tree-optimization/84111
3092 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3093 inner loops added during recursion, as they don't have up-to-date
3096 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3099 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3100 (can_inline_edge_by_limits_p): ... here.
3101 (can_early_inline_edge_p, check_callers,
3102 update_caller_keys, update_callee_keys, recursive_inlining,
3103 add_new_edges_to_heap, speculation_useful_p,
3104 inline_small_functions,
3105 inline_small_functions, flatten_function,
3106 inline_to_all_callers_1): Update.
3108 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3110 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3113 2018-01-30 Richard Biener <rguenther@suse.de>
3115 PR tree-optimization/83008
3116 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3117 invariant and constant vector uses in stmts when they need
3120 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3123 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3124 * configure: Regenerate.
3126 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3128 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3130 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3131 Use gen_rtx_REG rather than gen_lowpart.
3133 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3135 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3136 rather than 0 when creating partial subregs.
3138 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3140 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3143 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3146 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3147 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3148 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3149 flags. This restores the settings used before the 2017-07-24.
3150 Turning off pre increment/decrement/modify allows IVOPTS to
3151 optimize DF/SF loops where the index is an int.
3153 2018-01-29 Richard Biener <rguenther@suse.de>
3154 Kelvin Nilsen <kelvin@gcc.gnu.org>
3157 * tree-vect-stmts.c (vectorizable_call): Don't call
3158 targetm.vectorize_builtin_md_vectorized_function if callee is
3161 2018-01-22 Carl Love <cel@us.ibm.com>
3163 * doc/extend.tex: Fix typo in second arg in
3164 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3166 2018-01-29 Richard Biener <rguenther@suse.de>
3168 PR tree-optimization/84086
3169 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3170 (flush_ssaname_freelist): When SSA names were released reset
3171 the SCEV hash table.
3173 2018-01-29 Richard Biener <rguenther@suse.de>
3175 PR tree-optimization/84057
3176 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3177 removed paths when removing edges.
3179 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3181 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3182 -mfunction-return=@var{choice}.
3184 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3187 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3189 (layout::print_source_line): Likewise.
3190 (test_get_line_width_without_trailing_whitespace): Add test cases.
3192 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3195 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3198 2018-01-26 Jim Wilson <jimw@sifive.com>
3200 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3202 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3205 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3207 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3208 and CMP + SUB-immediate -> SUBS.
3210 2018-01-26 Martin Sebor <msebor@redhat.com>
3212 PR tree-optimization/83896
3213 * tree-ssa-strlen.c (get_string_len): Rename...
3214 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3215 Avoid assuming length is constant.
3216 (handle_char_store): Use HOST_WIDE_INT for string length.
3218 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3221 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3222 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3224 2018-01-26 Richard Biener <rguenther@suse.de>
3226 PR rtl-optimization/84003
3227 * dse.c (record_store): Only record redundant stores when
3228 the earlier store aliases at least all accesses the later one does.
3230 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3232 PR rtl-optimization/83985
3233 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3234 REG_CFA_RESTORE insns.
3235 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3236 REG_CFA_RESTORE insns here.
3239 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3240 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3242 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3244 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3245 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3246 (arc_init): Likewise.
3247 (arc_override_options): Likewise.
3248 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3250 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3252 * config/arc/arc.h (TARGET_DBNZ): Define.
3253 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3254 properly set the tune attribute.
3255 (dbnz): Use TARGET_DBNZ guard.
3256 * config/arc/arc.opt (mtune): Add core3 option.
3258 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3260 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3261 recognize new pic like addresses.
3262 (arc_delegitimize_address): Clean up.
3264 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3266 * config/arc/arc-arches.def: Option mrf16 valid for all
3268 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3269 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3270 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3271 * config/arc/arc-tables.opt: Regenerate.
3272 * config/arc/arc.c (arc_conditional_register_usage): Handle
3273 reduced register file case.
3274 (arc_file_start): Set must have build attributes.
3275 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3277 * config/arc/arc.opt (mrf16): Add new option.
3278 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3279 * config/arc/genmultilib.awk: Handle new mrf16 option.
3280 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3281 * config/arc/t-multilib: Regenerate.
3282 * doc/invoke.texi (ARC Options): Document mrf16 option.
3284 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3286 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3287 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3288 (arc_attribute_table): Add 'secure_call' attribute.
3289 (arc_print_operand): Print secure call operand.
3290 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3292 (arc_is_secure_call_p): New function. * config/arc/arc.md
3293 (call_i): Add support for sjli instruction.
3294 (call_value_i): Likewise.
3295 * config/arc/constraints.md (Csc): New constraint.
3297 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3298 John Eric Martin <John.Martin@emmicro-us.com>
3300 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3301 * config/arc/arc.c (_arc_jli_section): New struct.
3302 (arc_jli_section): New type.
3303 (rc_jli_sections): New static variable.
3304 (arc_handle_jli_attribute): New function.
3305 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3306 (arc_file_end): New function.
3307 (TARGET_ASM_FILE_END): Define.
3308 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3309 (arc_add_jli_section): New function.
3310 (jli_call_scan): Likewise.
3311 (arc_reorg): Call jli_call_scan.
3312 (arc_output_addsi): Remove 'S' from printing asm operand.
3313 (arc_is_jli_call_p): New function.
3314 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3316 (movhi_insn): Likewise.
3317 (movsi_insn): Likewise.
3318 (movsi_set_cc_insn): Likewise.
3319 (loadqi_update): Likewise.
3320 (load_zeroextendqisi_update): Likewise.
3321 (load_signextendqisi_update): Likewise.
3322 (loadhi_update): Likewise.
3323 (load_zeroextendhisi_update): Likewise.
3324 (load_signextendhisi_update): Likewise.
3325 (loadsi_update): Likewise.
3326 (loadsf_update): Likewise.
3327 (movsicc_insn): Likewise.
3328 (bset_insn): Likewise.
3329 (bxor_insn): Likewise.
3330 (bclr_insn): Likewise.
3331 (bmsk_insn): Likewise.
3332 (bicsi3_insn): Likewise.
3333 (cmpsi_cc_c_insn): Likewise.
3334 (movsi_ne): Likewise.
3335 (movsi_cond_exec): Likewise.
3336 (clrsbsi2): Likewise.
3347 (call_i): Remove 'S' asm letter, add jli instruction.
3348 (call_value_i): Likewise.
3349 * config/arc/arc.op (mjli-always): New option.
3350 * config/arc/constraints.md (Cji): New constraint.
3351 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3353 (subsf3_fpx): Likewise.
3354 (mulsf3_fpx): Likewise.
3355 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3357 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3359 * doc/invoke.texi (ARC): Document mjli-always option.
3361 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3363 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3364 avoid addition with 0 and use incw and decw where possible.
3366 2018-01-26 Richard Biener <rguenther@suse.de>
3368 PR tree-optimization/81082
3369 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3370 association if it requires casting to unsigned.
3371 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3372 from fold_plusminus_mult_expr to catch important cases late when
3373 range info is available.
3375 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3377 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3378 * configure.ac (hidden_linkonce): New test.
3379 * configure: Regenerate.
3380 * config.in: Regenerate.
3382 2018-01-26 Julia Koval <julia.koval@intel.com>
3384 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3385 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3386 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3387 _mm_mask_bitshuffle_epi64_mask): Fix type.
3388 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3389 USI_FTYPE_V4DI_V4DI_USI): Remove.
3390 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3391 __builtin_ia32_vpshufbitqmb256_mask,
3392 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3393 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3394 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3396 2018-01-26 Alan Modra <amodra@gmail.com>
3399 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3400 UNSPEC_VBPERMQ. Sort other unspecs.
3402 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3404 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3406 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3409 * predict.c (drop_profile): Do not push/pop cfun; update also
3411 (handle_missing_profiles): Fix logic looking for zero profiles.
3413 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3416 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3417 on functions with #pragma omp declare simd or functions with simd
3419 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3420 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3421 Remove trailing \n from warning_at calls.
3423 2018-01-25 Tom de Vries <tom@codesourcery.com>
3426 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3427 for neutered workers.
3429 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3432 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3433 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3435 2018-01-24 Jeff Law <law@redhat.com>
3438 * i386.c (get_probe_interval): Move to earlier point.
3439 (ix86_compute_frame_layout): If -fstack-clash-protection and
3440 the frame is larger than the probe interval, then use pushes
3441 to save registers rather than reg->mem moves.
3442 (ix86_expand_prologue): Remove conditional for int_registers_saved
3445 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3448 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3449 min/max for never referenced object.
3451 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3454 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3456 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3457 attributes from DECL_ATTRIBUTES (decl) without affecting
3458 DECL_ATTRIBUTES (current_function_decl).
3459 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3460 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3462 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3464 PR tree-optimization/83979
3465 * fold-const.c (fold_comparison): Use constant_boolean_node
3466 instead of boolean_{true,false}_node.
3468 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3470 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3473 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3475 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3476 Simplify the clause that sets the length attribute.
3477 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3478 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3479 clause that sets the length attribute.
3480 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3482 2018-01-24 Tom de Vries <tom@codesourcery.com>
3485 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3486 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3487 Add strict parameter.
3488 (prevent_branch_around_nothing): Insert dummy insn between branch to
3489 label and label with no ptx insn inbetween.
3490 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3492 2018-01-24 Tom de Vries <tom@codesourcery.com>
3495 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3496 for neutered threads in warp.
3497 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3499 2018-01-24 Richard Biener <rguenther@suse.de>
3501 PR tree-optimization/83176
3502 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3505 2018-01-24 Richard Biener <rguenther@suse.de>
3507 PR tree-optimization/82819
3508 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3509 code generating pluses that are no-ops in the target precision.
3511 2018-01-24 Richard Biener <rguenther@suse.de>
3514 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3516 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3518 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3519 to merge probabilities.
3520 * predict.c (probably_never_executed): Also mark as cold functions
3521 with global 0 profile and guessed local profile.
3522 * profile-count.c (profile_probability::combine_with_count): New
3524 * profile-count.h (profile_probability::operator*,
3525 profile_probability::operator*=, profile_probability::operator/,
3526 profile_probability::operator/=): Reduce precision to adjusted
3527 and set value to guessed on contradictory divisions.
3528 (profile_probability::combine_with_freq): Remove.
3529 (profile_probability::combine_wiht_count): Declare.
3530 (profile_count::force_nonzero):: Set to adjusted.
3531 (profile_count::probability_in):: Set quality to adjusted.
3532 * tree-ssa-tail-merge.c (replace_block_by): Use
3535 2018-01-23 Andrew Waterman <andrew@sifive.com>
3536 Jim Wilson <jimw@sifive.com>
3538 * config/riscv/riscv.c (riscv_stack_boundary): New.
3539 (riscv_option_override): Set riscv_stack_boundary. Handle
3540 riscv_preferred_stack_boundary_arg.
3541 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3542 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3543 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3544 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3545 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3546 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3548 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3551 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3552 of struct ix86_frame.
3553 (ix86_expand_epilogue): Likewise. Add a local variable for
3554 the reg_save_offset field in struct ix86_frame.
3556 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3558 PR tree-optimization/82604
3559 * tree-loop-distribution.c (enum partition_kind): New enum item
3560 PKIND_PARTIAL_MEMSET.
3561 (partition_builtin_p): Support above new enum item.
3562 (generate_code_for_partition): Ditto.
3563 (compute_access_range): Differentiate cases that equality can be
3564 proven at all loops, the innermost loops or no loops.
3565 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3566 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3567 (finalize_partitions, distribute_loop): Don't fuse partition of
3568 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3569 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3572 2018-01-23 Martin Liska <mliska@suse.cz>
3574 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3575 order to ignore the predictor.
3576 (PRED_POLYMORPHIC_CALL): Likewise.
3577 (PRED_RECURSIVE_CALL): Likewise.
3579 2018-01-23 Martin Liska <mliska@suse.cz>
3581 * tree-profile.c (tree_profiling): Print function header to
3582 aware reader which function we are working on.
3583 * value-prof.c (gimple_find_values_to_profile): Do not print
3584 not interesting value histograms.
3586 2018-01-23 Martin Liska <mliska@suse.cz>
3588 * profile-count.h (enum profile_quality): Add
3589 profile_uninitialized as the first value. Do not number values
3590 as they are zero based.
3591 (profile_count::verify): Update sanity check.
3592 (profile_probability::verify): Likewise.
3594 2018-01-23 Nathan Sidwell <nathan@acm.org>
3596 * doc/invoke.texi (ffor-scope): Deprecate.
3598 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3600 PR tree-optimization/83510
3601 * domwalk.c (set_all_edges_as_executable): New function.
3602 (dom_walker::dom_walker): Convert bool param
3603 "skip_unreachable_blocks" to enum reachability. Move setup of
3604 edge flags to set_all_edges_as_executable and only do it when
3605 reachability is REACHABLE_BLOCKS.
3606 * domwalk.h (enum dom_walker::reachability): New enum.
3607 (dom_walker::dom_walker): Convert bool param
3608 "skip_unreachable_blocks" to enum reachability.
3609 (set_all_edges_as_executable): New decl.
3610 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3611 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3613 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3614 but converting true to REACHABLE_BLOCKS.
3615 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3617 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3618 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3619 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3621 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3622 if check_all_array_refs will be called.
3624 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3626 * tree.c (selftest::test_location_wrappers): Add more test
3629 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3631 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3632 (selftest::test_bit_in_range): Likewise.
3634 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3637 * doc/sourcebuild.texi (vect_float): Say that the selector
3638 only describes the situation when -funsafe-math-optimizations is on.
3639 (vect_float_strict): Document.
3641 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3643 PR tree-optimization/83965
3644 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3645 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3646 instead of checking only for a reduction.
3647 (vect_recog_widen_sum_pattern): Likewise.
3649 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3651 * predict.c (probably_never_executed): Only use precise profile info.
3652 (compute_function_frequency): Skip after inlining hack since we now
3653 have quality checking.
3655 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3657 * profile-count.h (profile_probability::very_unlikely,
3658 profile_probability::unlikely, profile_probability::even): Set
3659 precision to guessed.
3661 2018-01-23 Richard Biener <rguenther@suse.de>
3663 PR tree-optimization/83963
3664 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3665 Properly terminate dominator walk when crossing the exit edge not
3666 when visiting its source block.
3668 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3671 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3672 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3674 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3676 PR tree-optimization/83957
3677 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3678 semicolon after for body surrounded by braces.
3680 PR tree-optimization/83081
3681 * profile-count.h (profile_probability::split): New method.
3682 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3683 Use profile_probability::split.
3684 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3685 when splitting a single conditional jump into 2.
3687 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3689 PR tree-optimization/69452
3690 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3693 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3695 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3696 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3697 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3699 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3701 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3702 * config/rl78/rl78.md (movdi): New define_expand.
3703 * config/rl78/rl78.c (rl78_split_movdi): New function.
3705 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3708 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3710 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3711 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3712 128-bit to produce an UNSPEC move to get the double word with the
3713 signbit and then a shift directly to do signbit.
3714 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3715 implementation with a new version that just does either a direct
3716 move or a regular move. Move memory interface to separate insns.
3717 Move insns so they are next to the expander.
3718 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3719 with signbit move. Split big and little endian case.
3720 (signbit<mode>2_dm_mem_le): Likewise.
3721 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3722 (signbit<mode>2_dm2): Likewise.
3724 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3726 * config/rl78/rl78.md (anddi3): New define_expand.
3728 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3730 * config/rl78/rl78.md (umindi3): New define_expand.
3732 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3734 * config/rl78/rl78.md (smindi3): New define_expand.
3736 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3738 * config/rl78/rl78.md (smaxdi3): New define_expand.
3740 2018-01-22 Carl Love <cel@us.ibm.com>
3742 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3743 LVX_V1TI): Add macro expansion.
3744 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3745 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3746 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3747 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3748 Change check to determine if the instruction is a byte reversing
3749 entry. Fix typo in comment.
3750 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3751 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3752 Add def_builtin calls for new builtins.
3753 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3754 Add define_insn expansion.
3756 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3758 * config/rl78/rl78.md (umaxdi3): New define_expand.
3760 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3762 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3763 for non-QImode registers.
3765 2018-01-22 Richard Biener <rguenther@suse.de>
3767 PR tree-optimization/83963
3768 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3769 including the loop exit block.
3770 (scop_detection::merge_sese): Likewise.
3771 (scop_detection::add_scop): Do it here instead.
3773 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3775 * doc/sourcebuild.texi (arm_softfloat): Document.
3777 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3780 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3781 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3782 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3784 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3785 David Edelsohn <dje.gcc@gmail.com>
3788 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3789 Change "crset eq" to "crset 2".
3790 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3791 (*call_indirect_aix<mode>_nospec): Likewise.
3792 (*call_value_indirect_aix<mode>_nospec): Likewise.
3793 (*call_indirect_elfv2<mode>_nospec): Likewise.
3794 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3795 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3796 change assembly output from . to $.
3797 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3798 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3799 (*tablejump<mode>_internal1_nospec): Likewise.
3801 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3804 * config/sh/sh_optimize_sett_clrt.cc:
3805 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3807 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3809 PR tree-optimization/83940
3810 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3811 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3812 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3813 use it to pass back the definition type.
3814 (vect_check_store_rhs): Likewise rhs_dt_out.
3815 (vect_build_gather_load_calls): Add a mask_dt argument and use
3816 it instead of a call to vect_is_simple_use.
3817 (vectorizable_store): Update calls to vect_check_load_store_mask
3818 and vect_check_store_rhs. Use the dt returned by the latter instead
3819 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3820 instead of calls to vect_is_simple_use. Pass the scalar rather
3821 than the vector operand to vect_is_simple_use when handling
3822 second and subsequent copies of an rhs value.
3823 (vectorizable_load): Update calls to vect_check_load_store_mask
3824 and vect_build_gather_load_calls. Use the cached mask_dt and
3825 gs_info.offset_dt instead of calls to vect_is_simple_use.
3827 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3830 * tree-emutls.c: Include gimplify.h.
3831 (lower_emutls_2): New function.
3832 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3833 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3834 it before further processing.
3837 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3838 UINTVAL (trueop1) instead of INTVAL (op1).
3840 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3844 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3845 INCOMING_FRAME_SP_OFFSET if not defined.
3846 (scan_trace): Add ENTRY argument. If true and
3847 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3848 emit a note to adjust the CFA offset.
3849 (create_cfi_notes): Adjust scan_trace callers.
3850 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3851 INCOMING_FRAME_SP_OFFSET in the CIE.
3852 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3853 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3855 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3856 * doc/tm.texi: Regenerated.
3858 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3860 PR rtl-optimization/83147
3861 * lra-constraints.c (remove_inheritance_pseudos): Use
3862 lra_substitute_pseudo_within_insn.
3864 2018-01-19 Tom de Vries <tom@codesourcery.com>
3865 Cesar Philippidis <cesar@codesourcery.com>
3868 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3870 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3873 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3874 spaces for function labels.
3876 2018-01-19 Martin Liska <mliska@suse.cz>
3878 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3879 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3880 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3881 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3882 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3883 (PRED_CONST_RETURN): Change from 69 to 65.
3884 (PRED_NULL_RETURN): Change from 91 to 71.
3885 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3886 (PRED_LOOP_GUARD): Change from 66 to 73.
3888 2018-01-19 Martin Liska <mliska@suse.cz>
3890 * predict.c (predict_insn_def): Add new assert.
3891 (struct branch_predictor): Change type to signed integer.
3892 (test_prediction_value_range): Amend test to cover
3894 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3895 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3896 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3897 (PRED_LOOP_IV_COMPARE): Likewise.
3898 * predict.h (PROB_UNINITIALIZED): Define new constant.
3900 2018-01-19 Martin Liska <mliska@suse.cz>
3902 * predict.c (dump_prediction): Add new format for
3903 analyze_brprob.py script which is enabled with -details
3905 * profile-count.h (precise_p): New function.
3907 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3909 PR tree-optimization/83922
3910 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3911 there are no statements that need masking.
3912 (vect_active_double_reduction_p): New function.
3913 (vect_analyze_loop_operations): Use it when handling phis that
3914 are not in the loop header.
3916 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3918 PR tree-optimization/83914
3919 * tree-vect-loop.c (vectorizable_induction): Don't convert
3920 init_expr or apply the peeling adjustment for inductions
3921 that are nested within the vectorized loop.
3923 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3925 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3928 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3932 * function.h (gimplify_parameters): Add gimple_seq * argument.
3933 * function.c: Include gimple.h and options.h.
3934 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3935 for the added local temporaries if needed.
3936 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3937 if there are any parameter cleanups, wrap whole body into a
3938 try/finally with the cleanups.
3940 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3943 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3944 Use GET_MODE_CLASS for scalar floating point.
3946 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3950 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3951 Fix call of call_cgraph_insertion_hooks.
3953 2018-01-18 Martin Sebor <msebor@redhat.com>
3955 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3957 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3960 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3963 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3966 * common.opt: (-ffile-prefix-map): New option.
3967 * opts.c (common_handle_option): Defer it.
3968 * opts-global.c (handle_common_deferred_options): Handle it.
3969 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3970 * file-prefix-map.h: New file.
3971 (remap_debug_filename, add_debug_prefix_map): ...here.
3972 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3973 * final.c (debug_prefix_map, add_debug_prefix_map
3974 remap_debug_filename): Move to...
3975 * file-prefix-map.c: New file.
3976 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3977 generalize, get rid of alloca(), use strrchr() instead of strchr().
3978 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3979 Implement in terms of add_prefix_map().
3980 (remap_macro_filename, remap_debug_filename): Implement in term of
3982 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3983 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3984 * dbxout.c: Include file-prefix-map.h.
3985 * varasm.c: Likewise.
3986 * vmsdbgout.c: Likewise.
3987 * xcoffout.c: Likewise.
3988 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3989 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3990 * doc/invoke.texi (-ffile-prefix-map): Document.
3991 (-fdebug-prefix-map): Update description.
3993 2018-01-18 Martin Liska <mliska@suse.cz>
3995 * config/i386/i386.c (indirect_thunk_name): Document that also
3997 (output_indirect_thunk): Document why both instructions
3998 (pause and lfence) are generated.
4000 2018-01-18 Richard Biener <rguenther@suse.de>
4002 PR tree-optimization/83887
4003 * graphite-scop-detection.c
4004 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4005 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4006 (scop_detection::merge_sese): Re-implement with a flood-fill
4007 algorithm that properly finds a SESE region if it exists.
4009 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4012 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4013 pointer_diff optimizations use view_convert instead of convert.
4015 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4017 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4018 Generate different code for -mno-speculate-indirect-jumps.
4019 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4020 (*call_indirect_aix<mode>): Disable for
4021 -mno-speculate-indirect-jumps.
4022 (*call_indirect_aix<mode>_nospec): New define_insn.
4023 (*call_value_indirect_aix<mode>): Disable for
4024 -mno-speculate-indirect-jumps.
4025 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4026 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4027 -mno-speculate-indirect-jumps.
4028 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4030 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4032 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4033 long double type, set the flags for noting the default long double
4034 type, even if we don't pass or return a long double type.
4036 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4039 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4042 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4044 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4045 support for merge[hl].
4046 (fold_mergehl_helper): New helper function.
4047 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4048 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4049 (altivec_vmrglw_direct): Add xxmrglw insn.
4051 2018-01-17 Andrew Waterman <andrew@sifive.com>
4053 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4054 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4056 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4059 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4060 call the lto_location_cache before reading the
4061 DECL_SOURCE_LOCATION of the types.
4063 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4064 Richard Sandiford <richard.sandiford@linaro.org>
4066 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4067 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4068 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4069 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4070 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4072 * config/aarch64/constraints.md (aarch64_movti_operand):
4074 * config/aarch64/predicates.md (Uti): Add new constraint.
4076 2018-01-17 Carl Love <cel@us.ibm.com>
4078 * config/rs6000/vsx.md (define_expand xl_len_r,
4079 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4080 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4082 (define_expand, define_insn): Move the shift left from the
4083 define_insn to the define_expand for lxvl and stxvl instructions.
4084 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4085 and XL_LEN_R definitions to PURE.
4087 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4089 * config/i386/i386.c (indirect_thunk_name): Declare regno
4090 as unsigned int. Compare regno with INVALID_REGNUM.
4091 (output_indirect_thunk): Ditto.
4092 (output_indirect_thunk_function): Ditto.
4093 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4094 in the call to output_indirect_thunk_function.
4096 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4099 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4100 rather than the size of inner_type to determine the stack slot size
4101 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4103 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4106 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4109 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4111 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4112 endian Linux systems to optionally enable multilibs for selecting
4113 the long double type if the user configured an explicit type.
4114 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4115 have no long double multilibs if not defined.
4116 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4117 warn if the user used -mabi={ieee,ibm}longdouble and we built
4118 multilibs for long double.
4119 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4120 appropriate multilib option.
4121 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4123 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4124 for building long double multilibs.
4125 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4127 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4129 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4132 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4134 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4137 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4140 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4143 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4145 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4146 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4148 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4150 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4151 different rtl trees depending on TARGET_64BIT.
4152 (rs6000_gen_lvx): Likewise.
4154 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4156 * config/visium/visium.md (nop): Tweak comment.
4157 (hazard_nop): Likewise.
4159 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4161 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4162 -mspeculate-indirect-jumps.
4163 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4164 for -mno-speculate-indirect-jumps.
4165 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4166 (*call_value_indirect_elfv2<mode>): Disable for
4167 -mno-speculate-indirect-jumps.
4168 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4169 (indirect_jump): Emit different RTL for
4170 -mno-speculate-indirect-jumps.
4171 (*indirect_jump<mode>): Disable for
4172 -mno-speculate-indirect-jumps.
4173 (*indirect_jump<mode>_nospec): New define_insn.
4174 (tablejump): Emit different RTL for
4175 -mno-speculate-indirect-jumps.
4176 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4177 (tablejumpsi_nospec): New define_expand.
4178 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4179 (tablejumpdi_nospec): New define_expand.
4180 (*tablejump<mode>_internal1): Disable for
4181 -mno-speculate-indirect-jumps.
4182 (*tablejump<mode>_internal1_nospec): New define_insn.
4183 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4186 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4188 * caller-save.c (insert_save): Drop unnecessary parameter. All
4191 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4192 Richard Biener <rguenth@suse.de>
4195 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4196 return early, inline manually is_gimple_sizepos. Make sure if we
4197 call gimplify_expr we don't end up with a gimple constant.
4198 * tree.c (variably_modified_type_p): Don't return true for
4199 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4200 * gimplify.h (is_gimple_sizepos): Remove.
4202 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4204 PR tree-optimization/83857
4205 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4206 vectorizable_live_operation for pure SLP statements.
4207 (vectorizable_live_operation): Handle PHIs.
4209 2018-01-16 Richard Biener <rguenther@suse.de>
4211 PR tree-optimization/83867
4212 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4213 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4215 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4218 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4219 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4220 If off is not INTEGER_CST, issue a may not be aligned warning
4221 rather than isn't aligned. Use isn%'t rather than isn't.
4222 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4224 <case MULT_EXPR>: Improve the case when bottom and one of the
4225 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4226 operand, in that case check if the other operand is multiple of
4227 bottom divided by the INTEGER_CST operand.
4229 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4232 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4233 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4234 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4235 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4236 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4237 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4238 (pa_function_arg_size): New function.
4240 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4242 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4243 in a separate statement.
4245 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4247 PR tree-optimization/83847
4248 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4249 group gathers and scatters.
4251 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4253 PR rtl-optimization/86620
4254 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4256 PR rtl-optimization/83213
4257 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4258 to last if both are JUMP_INSNs.
4260 PR tree-optimization/83843
4261 * gimple-ssa-store-merging.c
4262 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4263 store_immediate_info for bswap/nop orig_stores.
4265 2018-01-15 Andrew Waterman <andrew@sifive.com>
4267 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4269 <UDIV>: Increase cost if !TARGET_DIV.
4271 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4273 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4274 (define_attr "cr_logical_3op"): New.
4275 (cceq_ior_compare): Adjust.
4276 (cceq_ior_compare_complement): Adjust.
4277 (*cceq_rev_compare): Adjust.
4278 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4279 (is_cracked_insn): Adjust.
4280 (insn_must_be_first_in_group): Adjust.
4281 * config/rs6000/40x.md: Adjust.
4282 * config/rs6000/440.md: Adjust.
4283 * config/rs6000/476.md: Adjust.
4284 * config/rs6000/601.md: Adjust.
4285 * config/rs6000/603.md: Adjust.
4286 * config/rs6000/6xx.md: Adjust.
4287 * config/rs6000/7450.md: Adjust.
4288 * config/rs6000/7xx.md: Adjust.
4289 * config/rs6000/8540.md: Adjust.
4290 * config/rs6000/cell.md: Adjust.
4291 * config/rs6000/e300c2c3.md: Adjust.
4292 * config/rs6000/e500mc.md: Adjust.
4293 * config/rs6000/e500mc64.md: Adjust.
4294 * config/rs6000/e5500.md: Adjust.
4295 * config/rs6000/e6500.md: Adjust.
4296 * config/rs6000/mpc.md: Adjust.
4297 * config/rs6000/power4.md: Adjust.
4298 * config/rs6000/power5.md: Adjust.
4299 * config/rs6000/power6.md: Adjust.
4300 * config/rs6000/power7.md: Adjust.
4301 * config/rs6000/power8.md: Adjust.
4302 * config/rs6000/power9.md: Adjust.
4303 * config/rs6000/rs64.md: Adjust.
4304 * config/rs6000/titan.md: Adjust.
4306 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4308 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4309 ix86_indirect_branch_register logic.
4311 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4313 * config/i386/constraints.md (Bs): Update
4314 ix86_indirect_branch_register check. Don't check
4315 ix86_indirect_branch_register with GOT_memory_operand.
4317 * config/i386/predicates.md (GOT_memory_operand): Don't check
4318 ix86_indirect_branch_register here.
4319 (GOT32_symbol_operand): Likewise.
4321 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4323 * config/i386/predicates.md (constant_call_address_operand):
4324 Rewrite ix86_indirect_branch_register logic.
4325 (sibcall_insn_operand): Likewise.
4327 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4329 * config/i386/constraints.md (Bs): Replace
4330 ix86_indirect_branch_thunk_register with
4331 ix86_indirect_branch_register.
4333 * config/i386/i386.md (indirect_jump): Likewise.
4334 (tablejump): Likewise.
4335 (*sibcall_memory): Likewise.
4336 (*sibcall_value_memory): Likewise.
4337 Peepholes of indirect call and jump via memory: Likewise.
4338 * config/i386/i386.opt: Likewise.
4339 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4340 (GOT_memory_operand): Likewise.
4341 (call_insn_operand): Likewise.
4342 (sibcall_insn_operand): Likewise.
4343 (GOT32_symbol_operand): Likewise.
4345 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4348 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4349 type rather than type addr's type points to.
4350 (expand_omp_atomic_mutex): Likewise.
4351 (expand_omp_atomic): Likewise.
4353 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4356 * config/i386/i386.c (output_indirect_thunk_function): Use
4357 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4358 for __x86_return_thunk.
4360 2018-01-15 Richard Biener <rguenther@suse.de>
4363 * expmed.c (extract_bit_field_1): Fix typo.
4365 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4368 * config/arm/iterators.md (VF): New mode iterator.
4369 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4370 Remove integer-related logic from pattern.
4371 (neon_vabd<mode>_3): Likewise.
4373 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4376 * common.opt (fstrict-overflow): No longer an alias.
4377 (fwrapv-pointer): New option.
4378 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4379 also for pointer types based on flag_wrapv_pointer.
4380 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4381 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4382 opts->x_flag_wrapv got set.
4383 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4384 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4385 POINTER_TYPE_OVERFLOW_UNDEFINED.
4386 * match.pd: Likewise in address comparison pattern.
4387 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4389 2018-01-15 Richard Biener <rguenther@suse.de>
4392 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4393 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4394 Reset type names to their identifier if their TYPE_DECL doesn't
4395 have linkage (and thus is used for ODR and devirt).
4396 (save_debug_info_for_decl): Remove.
4397 (save_debug_info_for_type): Likewise.
4398 (add_tree_to_fld_list): Adjust.
4399 * tree-pretty-print.c (dump_generic_node): Make dumping of
4400 type names more robust.
4402 2018-01-15 Richard Biener <rguenther@suse.de>
4404 * BASE-VER: Bump to 8.0.1.
4406 2018-01-14 Martin Sebor <msebor@redhat.com>
4409 * builtins.c (check_access): Avoid warning when the no-warning bit
4412 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4414 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4415 * ira-color (allocno_hard_regs_compare): Likewise.
4417 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4420 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4421 Use .pushsection/.popsection.
4423 2018-01-14 Martin Sebor <msebor@redhat.com>
4426 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4428 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4430 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4431 entry from extra_headers.
4432 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4433 extra_headers, make the list bitwise identical to the i?86-*-* one.
4435 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4437 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4438 -mcmodel=large with -mindirect-branch=thunk,
4439 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4440 -mfunction-return=thunk-extern.
4441 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4442 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4443 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4445 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4447 * config/i386/i386.c (print_reg): Print the name of the full
4448 integer register without '%'.
4449 (ix86_print_operand): Handle 'V'.
4450 * doc/extend.texi: Document 'V' modifier.
4452 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4454 * config/i386/constraints.md (Bs): Disallow memory operand for
4455 -mindirect-branch-register.
4457 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4458 (GOT_memory_operand): Likewise.
4459 (call_insn_operand): Likewise.
4460 (sibcall_insn_operand): Likewise.
4461 (GOT32_symbol_operand): Likewise.
4462 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4463 for -mindirect-branch-register.
4464 (tablejump): Likewise.
4465 (*sibcall_memory): Likewise.
4466 (*sibcall_value_memory): Likewise.
4467 Disallow peepholes of indirect call and jump via memory for
4468 -mindirect-branch-register.
4469 (*call_pop): Replace m with Bw.
4470 (*call_value_pop): Likewise.
4471 (*sibcall_pop_memory): Replace m with Bs.
4472 * config/i386/i386.opt (mindirect-branch-register): New option.
4473 * doc/invoke.texi: Document -mindirect-branch-register option.
4475 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4477 * config/i386/i386-protos.h (ix86_output_function_return): New.
4478 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4479 set function_return_type.
4480 (indirect_thunk_name): Add ret_p to indicate thunk for function
4482 (output_indirect_thunk_function): Pass false to
4483 indirect_thunk_name.
4484 (ix86_output_indirect_branch_via_reg): Likewise.
4485 (ix86_output_indirect_branch_via_push): Likewise.
4486 (output_indirect_thunk_function): Create alias for function
4487 return thunk if regno < 0.
4488 (ix86_output_function_return): New function.
4489 (ix86_handle_fndecl_attribute): Handle function_return.
4490 (ix86_attribute_table): Add function_return.
4491 * config/i386/i386.h (machine_function): Add
4492 function_return_type.
4493 * config/i386/i386.md (simple_return_internal): Use
4494 ix86_output_function_return.
4495 (simple_return_internal_long): Likewise.
4496 * config/i386/i386.opt (mfunction-return=): New option.
4497 (indirect_branch): Mention -mfunction-return=.
4498 * doc/extend.texi: Document function_return function attribute.
4499 * doc/invoke.texi: Document -mfunction-return= option.
4501 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4503 * config/i386/i386-opts.h (indirect_branch): New.
4504 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4505 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4506 with local indirect jump when converting indirect call and jump.
4507 (ix86_set_indirect_branch_type): New.
4508 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4509 (indirectlabelno): New.
4510 (indirect_thunk_needed): Likewise.
4511 (indirect_thunk_bnd_needed): Likewise.
4512 (indirect_thunks_used): Likewise.
4513 (indirect_thunks_bnd_used): Likewise.
4514 (INDIRECT_LABEL): Likewise.
4515 (indirect_thunk_name): Likewise.
4516 (output_indirect_thunk): Likewise.
4517 (output_indirect_thunk_function): Likewise.
4518 (ix86_output_indirect_branch_via_reg): Likewise.
4519 (ix86_output_indirect_branch_via_push): Likewise.
4520 (ix86_output_indirect_branch): Likewise.
4521 (ix86_output_indirect_jmp): Likewise.
4522 (ix86_code_end): Call output_indirect_thunk_function if needed.
4523 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4525 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4526 (ix86_attribute_table): Add indirect_branch.
4527 * config/i386/i386.h (machine_function): Add indirect_branch_type
4528 and has_local_indirect_jump.
4529 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4531 (tablejump): Likewise.
4532 (*indirect_jump): Use ix86_output_indirect_jmp.
4533 (*tablejump_1): Likewise.
4534 (simple_return_indirect_internal): Likewise.
4535 * config/i386/i386.opt (mindirect-branch=): New option.
4536 (indirect_branch): New.
4539 (thunk-inline): Likewise.
4540 (thunk-extern): Likewise.
4541 * doc/extend.texi: Document indirect_branch function attribute.
4542 * doc/invoke.texi: Document -mindirect-branch= option.
4544 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4547 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4549 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4551 * ipa-inline.c (want_inline_small_function_p): Return false if
4552 inlining has already failed with CIF_FINAL_ERROR.
4553 (update_caller_keys): Call want_inline_small_function_p before
4555 (update_callee_keys): Likewise.
4557 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4559 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4561 (rs6000_quadword_masked_address_p): Likewise.
4562 (quad_aligned_load_p): Likewise.
4563 (quad_aligned_store_p): Likewise.
4564 (const_load_sequence_p): Add comment to describe the outer-most loop.
4565 (mimic_memory_attributes_and_flags): New function.
4566 (rs6000_gen_stvx): Likewise.
4567 (replace_swapped_aligned_store): Likewise.
4568 (rs6000_gen_lvx): Likewise.
4569 (replace_swapped_aligned_load): Likewise.
4570 (replace_swapped_load_constant): Capitalize argument name in
4571 comment describing this function.
4572 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4573 and stores that access quad-word aligned addresses and replace
4574 with stvx or lvx instructions when appropriate.
4575 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4576 New function prototype.
4577 (rs6000_quadword_masked_address_p): Likewise.
4578 (rs6000_gen_lvx): Likewise.
4579 (rs6000_gen_stvx): Likewise.
4580 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4581 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4582 when memory address is aligned.
4583 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4584 this split to select lvx instruction when memory address is aligned.
4585 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4586 instruction when memory address is aligned.
4587 (*vsx_le_perm_load_v16qi): Likewise.
4588 (four unnamed splitters): Modify to select the stvx instruction
4589 when memory is aligned.
4591 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4593 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4594 which appears in the queue multiple times.
4596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4597 Alan Hayward <alan.hayward@arm.com>
4598 David Sherwood <david.sherwood@arm.com>
4600 * tree-vectorizer.h (vec_lower_bound): New structure.
4601 (_loop_vec_info): Add check_nonzero and lower_bounds.
4602 (LOOP_VINFO_CHECK_NONZERO): New macro.
4603 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4604 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4605 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4606 fields. Make seg_len the distance travelled, not including the
4608 (dr_direction_indicator): Declare.
4609 (dr_zero_step_indicator): Likewise.
4610 (dr_known_forward_stride_p): Likewise.
4611 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4613 (runtime_alias_check_p): Allow runtime alias checks with
4615 (operator ==): Compare access_size and align.
4616 (prune_runtime_alias_test_list): Rework for new distinction between
4617 the access_size and seg_len.
4618 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4620 (get_segment_min_max): New function.
4621 (create_intersect_range_checks): Use it.
4622 (dr_step_indicator): New function.
4623 (dr_direction_indicator): Likewise.
4624 (dr_zero_step_indicator): Likewise.
4625 (dr_known_forward_stride_p): Likewise.
4626 * tree-loop-distribution.c (data_ref_segment_size): Return
4627 DR_STEP * (niters - 1).
4628 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4630 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4631 (vect_preserves_scalar_order_p): New function, split out from...
4632 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4633 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4634 (vect_vfa_access_size): New function.
4635 (vect_vfa_align): Likewise.
4636 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4637 (dump_lower_bound): New function.
4638 (vect_check_lower_bound): Likewise.
4639 (vect_small_gap_p): Likewise.
4640 (vectorizable_with_step_bound_p): Likewise.
4641 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4642 depencies if the vectorization factor is 1. Convert the checks
4643 for nonzero steps into checks on the bounds of DR_STEP. Try using
4644 a bunds check for variable steps if the minimum required step is
4645 relatively small. Update calls to the dr_with_seg_len
4646 constructor and to vect_compile_time_alias.
4647 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4649 (vect_loop_versioning): Call it.
4650 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4652 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4655 Alan Hayward <alan.hayward@arm.com>
4656 David Sherwood <david.sherwood@arm.com>
4658 * doc/sourcebuild.texi (vect_scatter_store): Document.
4659 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4661 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4663 * genopinit.c (main): Add supports_vec_scatter_store and
4664 supports_vec_scatter_store_cached to target_optabs.
4665 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4666 IFN_MASK_SCATTER_STORE.
4667 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4669 * internal-fn.h (internal_store_fn_p): Declare.
4670 (internal_fn_stored_value_index): Likewise.
4671 * internal-fn.c (scatter_store_direct): New macro.
4672 (expand_scatter_store_optab_fn): New function.
4673 (direct_scatter_store_optab_supported_p): New macro.
4674 (internal_store_fn_p): New function.
4675 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4676 IFN_MASK_SCATTER_STORE.
4677 (internal_fn_mask_index): Likewise.
4678 (internal_fn_stored_value_index): New function.
4679 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4681 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4682 * optabs-query.c (supports_vec_scatter_store_p): New function.
4683 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4684 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4685 true for scatter stores.
4686 (vect_gather_scatter_fn_p): Handle scatter stores too.
4687 (vect_check_gather_scatter): Consider using scatter stores if
4688 supports_vec_scatter_store_p.
4689 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4691 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4692 internal_fn_stored_value_index.
4693 (check_load_store_masking): Handle scatter stores too.
4694 (vect_get_store_rhs): Make public.
4695 (vectorizable_call): Use internal_store_fn_p.
4696 (vectorizable_store): Handle scatter store internal functions.
4697 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4698 when deciding whether the end of the group has been reached.
4699 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4700 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4701 (mask_scatter_store<mode>): New insns.
4703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4704 Alan Hayward <alan.hayward@arm.com>
4705 David Sherwood <david.sherwood@arm.com>
4707 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4708 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4709 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4711 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4712 Use vect_truncate_gather_scatter_offset if we can't treat the
4713 operation as a normal gather load or scatter store.
4714 (get_group_load_store_type): Take the gather_scatter_info
4715 as argument. Try using a gather load or scatter store for
4716 single-element groups.
4717 (get_load_store_type): Update calls to get_group_load_store_type
4718 and vect_use_strided_gather_scatters_p.
4720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4721 Alan Hayward <alan.hayward@arm.com>
4722 David Sherwood <david.sherwood@arm.com>
4724 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4725 optional tree argument.
4726 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4728 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4729 but continue to use the current value as a fallback.
4730 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4731 to compare the updates.
4732 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4733 (get_load_store_type): Use it when handling a strided access.
4734 (vect_get_strided_load_store_ops): New function.
4735 (vect_get_data_ptr_increment): Likewise.
4736 (vectorizable_load): Handle strided gather loads. Always pass
4737 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4739 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4740 Alan Hayward <alan.hayward@arm.com>
4741 David Sherwood <david.sherwood@arm.com>
4743 * doc/md.texi (gather_load@var{m}): Document.
4744 (mask_gather_load@var{m}): Likewise.
4745 * genopinit.c (main): Add supports_vec_gather_load and
4746 supports_vec_gather_load_cached to target_optabs.
4747 * optabs-tree.c (init_tree_optimization_optabs): Use
4748 ggc_cleared_alloc to allocate target_optabs.
4749 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4750 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4752 * internal-fn.h (internal_load_fn_p): Declare.
4753 (internal_gather_scatter_fn_p): Likewise.
4754 (internal_fn_mask_index): Likewise.
4755 (internal_gather_scatter_fn_supported_p): Likewise.
4756 * internal-fn.c (gather_load_direct): New macro.
4757 (expand_gather_load_optab_fn): New function.
4758 (direct_gather_load_optab_supported_p): New macro.
4759 (direct_internal_fn_optab): New function.
4760 (internal_load_fn_p): Likewise.
4761 (internal_gather_scatter_fn_p): Likewise.
4762 (internal_fn_mask_index): Likewise.
4763 (internal_gather_scatter_fn_supported_p): Likewise.
4764 * optabs-query.c (supports_at_least_one_mode_p): New function.
4765 (supports_vec_gather_load_p): Likewise.
4766 * optabs-query.h (supports_vec_gather_load_p): Declare.
4767 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4768 and memory_type field.
4769 (NUM_PATTERNS): Bump to 15.
4770 * tree-vect-data-refs.c: Include internal-fn.h.
4771 (vect_gather_scatter_fn_p): New function.
4772 (vect_describe_gather_scatter_call): Likewise.
4773 (vect_check_gather_scatter): Try using internal functions for
4774 gather loads. Recognize existing calls to a gather load function.
4775 (vect_analyze_data_refs): Consider using gather loads if
4776 supports_vec_gather_load_p.
4777 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4778 (vect_get_gather_scatter_offset_type): Likewise.
4779 (vect_convert_mask_for_vectype): Likewise.
4780 (vect_add_conversion_to_patterm): Likewise.
4781 (vect_try_gather_scatter_pattern): Likewise.
4782 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4783 (vect_vect_recog_func_ptrs): Add it.
4784 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4785 internal_fn_mask_index and internal_gather_scatter_fn_p.
4786 (check_load_store_masking): Take the gather_scatter_info as an
4787 argument and handle gather loads.
4788 (vect_get_gather_scatter_ops): New function.
4789 (vectorizable_call): Check internal_load_fn_p.
4790 (vectorizable_load): Likewise. Handle gather load internal
4792 (vectorizable_store): Update call to check_load_store_masking.
4793 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4794 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4795 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4796 (aarch64_gather_scale_operand_d): New predicates.
4797 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4798 (mask_gather_load<mode>): New insns.
4800 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4801 Alan Hayward <alan.hayward@arm.com>
4802 David Sherwood <david.sherwood@arm.com>
4804 * optabs.def (fold_left_plus_optab): New optab.
4805 * doc/md.texi (fold_left_plus_@var{m}): Document.
4806 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4807 * internal-fn.c (fold_left_direct): Define.
4808 (expand_fold_left_optab_fn): Likewise.
4809 (direct_fold_left_optab_supported_p): Likewise.
4810 * fold-const-call.c (fold_const_fold_left): New function.
4811 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4812 * tree-parloops.c (valid_reduction_p): New function.
4813 (gather_scalar_reductions): Use it.
4814 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4815 (vect_finish_replace_stmt): Declare.
4816 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4817 (needs_fold_left_reduction_p): New function, split out from...
4818 (vect_is_simple_reduction): ...here. Accept reductions that
4819 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4820 (vect_force_simple_reduction): Also store the reduction type in
4821 the assignment's STMT_VINFO_REDUC_TYPE.
4822 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4823 (merge_with_identity): New function.
4824 (vect_expand_fold_left): Likewise.
4825 (vectorize_fold_left_reduction): Likewise.
4826 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4827 scalar phi in place for it. Check for target support and reject
4828 cases that would reassociate the operation. Defer the transform
4829 phase to vectorize_fold_left_reduction.
4830 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4831 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4832 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4836 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4837 call to ifc_temp_var.
4839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4840 Alan Hayward <alan.hayward@arm.com>
4841 David Sherwood <david.sherwood@arm.com>
4843 * target.def (legitimize_address_displacement): Take the original
4844 offset as a poly_int.
4845 * targhooks.h (default_legitimize_address_displacement): Update
4847 * targhooks.c (default_legitimize_address_displacement): Likewise.
4848 * doc/tm.texi: Regenerate.
4849 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4850 as an argument, moving assert of ad->disp == ad->disp_term to...
4851 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4852 Try calling targetm.legitimize_address_displacement before expanding
4853 the address rather than afterwards, and adjust for the new interface.
4854 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4855 Match the new hook interface. Handle SVE addresses.
4856 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4859 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4861 * Makefile.in (OBJS): Add early-remat.o.
4862 * target.def (select_early_remat_modes): New hook.
4863 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4864 * doc/tm.texi: Regenerate.
4865 * targhooks.h (default_select_early_remat_modes): Declare.
4866 * targhooks.c (default_select_early_remat_modes): New function.
4867 * timevar.def (TV_EARLY_REMAT): New timevar.
4868 * passes.def (pass_early_remat): New pass.
4869 * tree-pass.h (make_pass_early_remat): Declare.
4870 * early-remat.c: New file.
4871 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4873 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4880 vfm1 with a bound_epilog parameter.
4881 (vect_do_peeling): Update calls accordingly, and move the prologue
4882 call earlier in the function. Treat the base bound_epilog as 0 for
4883 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4884 this base when peeling for gaps.
4885 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4886 with fully-masked loops.
4887 (vect_estimate_min_profitable_iters): Handle the single peeled
4888 iteration in that case.
4890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4891 Alan Hayward <alan.hayward@arm.com>
4892 David Sherwood <david.sherwood@arm.com>
4894 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4895 single-element interleaving even if the size is not a power of 2.
4896 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4897 accesses for single-element interleaving if the group size is
4900 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4901 Alan Hayward <alan.hayward@arm.com>
4902 David Sherwood <david.sherwood@arm.com>
4904 * doc/md.texi (fold_extract_last_@var{m}): Document.
4905 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4906 * optabs.def (fold_extract_last_optab): New optab.
4907 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4908 * internal-fn.c (fold_extract_direct): New macro.
4909 (expand_fold_extract_optab_fn): Likewise.
4910 (direct_fold_extract_optab_supported_p): Likewise.
4911 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4912 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4913 EXTRACT_LAST_REDUCTION.
4914 (get_initial_def_for_reduction): Do not create an initial vector
4915 for EXTRACT_LAST_REDUCTION reductions.
4916 (vectorizable_reduction): Leave the scalar phi in place for
4917 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4918 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4919 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4920 transform phase to vectorizable_condition.
4921 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4923 (vect_finish_stmt_generation): ...here.
4924 (vect_finish_replace_stmt): New function.
4925 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4926 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4928 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4931 Alan Hayward <alan.hayward@arm.com>
4932 David Sherwood <david.sherwood@arm.com>
4934 * doc/md.texi (extract_last_@var{m}): Document.
4935 * optabs.def (extract_last_optab): New optab.
4936 * internal-fn.def (EXTRACT_LAST): New internal function.
4937 * internal-fn.c (cond_unary_direct): New macro.
4938 (expand_cond_unary_optab_fn): Likewise.
4939 (direct_cond_unary_optab_supported_p): Likewise.
4940 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4941 loops using EXTRACT_LAST.
4942 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4943 (extract_last_<mode>): ...this optab.
4944 (vec_extract<mode><Vel>): Update accordingly.
4946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4947 Alan Hayward <alan.hayward@arm.com>
4948 David Sherwood <david.sherwood@arm.com>
4950 * target.def (empty_mask_is_expensive): New hook.
4951 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4952 * doc/tm.texi: Regenerate.
4953 * targhooks.h (default_empty_mask_is_expensive): Declare.
4954 * targhooks.c (default_empty_mask_is_expensive): New function.
4955 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4956 if the target says that empty masks are expensive.
4957 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4959 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4961 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4962 Alan Hayward <alan.hayward@arm.com>
4963 David Sherwood <david.sherwood@arm.com>
4965 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4966 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4967 (vect_use_loop_mask_for_alignment_p): New function.
4968 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4969 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4970 niters_skip argument. Make sure that the first niters_skip elements
4971 of the first iteration are inactive.
4972 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4973 Update call to vect_set_loop_masks_directly.
4974 (get_misalign_in_elems): New function, split out from...
4975 (vect_gen_prolog_loop_niters): ...here.
4976 (vect_update_init_of_dr): Take a code argument that specifies whether
4977 the adjustment should be added or subtracted.
4978 (vect_update_init_of_drs): Likewise.
4979 (vect_prepare_for_masked_peels): New function.
4980 (vect_do_peeling): Skip prologue peeling if we're using a mask
4981 instead. Update call to vect_update_inits_of_drs.
4982 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4984 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4985 alignment. Do not include the number of peeled iterations in
4986 the minimum threshold in that case.
4987 (vectorizable_induction): Adjust the start value down by
4988 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4989 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4990 Take the number of skipped iterations into account when calculating
4992 * tree-vect-stmts.c (vect_gen_while_not): New function.
4994 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4995 Alan Hayward <alan.hayward@arm.com>
4996 David Sherwood <david.sherwood@arm.com>
4998 * doc/sourcebuild.texi (vect_fully_masked): Document.
4999 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5001 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5003 (vect_analyze_loop_2): ...here. Don't check the vectorization
5004 factor against the number of loop iterations if the loop is
5007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5008 Alan Hayward <alan.hayward@arm.com>
5009 David Sherwood <david.sherwood@arm.com>
5011 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5012 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5013 (dump_groups): Update accordingly.
5014 (iv_use::mem_type): New member variable.
5015 (address_p): New function.
5016 (record_use): Add a mem_type argument and initialize the new
5018 (record_group_use): Add a mem_type argument. Use address_p.
5019 Remove obsolete null checks of base_object. Update call to record_use.
5020 (find_interesting_uses_op): Update call to record_group_use.
5021 (find_interesting_uses_cond): Likewise.
5022 (find_interesting_uses_address): Likewise.
5023 (get_mem_type_for_internal_fn): New function.
5024 (find_address_like_use): Likewise.
5025 (find_interesting_uses_stmt): Try find_address_like_use before
5026 calling find_interesting_uses_op.
5027 (addr_offset_valid_p): Use the iv mem_type field as the type
5028 of the addressed memory.
5029 (add_autoinc_candidates): Likewise.
5030 (get_address_cost): Likewise.
5031 (split_small_address_groups_p): Use address_p.
5032 (split_address_groups): Likewise.
5033 (add_iv_candidate_for_use): Likewise.
5034 (autoinc_possible_for_pair): Likewise.
5035 (rewrite_groups): Likewise.
5036 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5037 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5038 (get_alias_ptr_type_for_ptr_address): New function.
5039 (rewrite_use_address): Rewrite address uses in calls that were
5040 identified by find_address_like_use.
5042 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5043 Alan Hayward <alan.hayward@arm.com>
5044 David Sherwood <david.sherwood@arm.com>
5046 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5048 * gimple-expr.h (is_gimple_addressable: Likewise.
5049 * gimple-expr.c (is_gimple_address): Likewise.
5050 * internal-fn.c (expand_call_mem_ref): New function.
5051 (expand_mask_load_optab_fn): Use it.
5052 (expand_mask_store_optab_fn): Likewise.
5054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5055 Alan Hayward <alan.hayward@arm.com>
5056 David Sherwood <david.sherwood@arm.com>
5058 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5059 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5060 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5061 (cond_umax@var{mode}): Document.
5062 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5063 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5064 (cond_umin_optab, cond_umax_optab): New optabs.
5065 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5066 (COND_IOR, COND_XOR): New internal functions.
5067 * internal-fn.h (get_conditional_internal_fn): Declare.
5068 * internal-fn.c (cond_binary_direct): New macro.
5069 (expand_cond_binary_optab_fn): Likewise.
5070 (direct_cond_binary_optab_supported_p): Likewise.
5071 (get_conditional_internal_fn): New function.
5072 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5073 Cope with reduction statements that are vectorized as calls rather
5075 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5076 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5077 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5078 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5079 (UNSPEC_COND_EOR): New unspecs.
5080 (optab): Add mappings for them.
5081 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5082 (sve_int_op, sve_fp_op): New int attributes.
5084 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5085 Alan Hayward <alan.hayward@arm.com>
5086 David Sherwood <david.sherwood@arm.com>
5088 * optabs.def (while_ult_optab): New optab.
5089 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5090 * internal-fn.def (WHILE_ULT): New internal function.
5091 * internal-fn.h (direct_internal_fn_supported_p): New override
5092 that takes two types as argument.
5093 * internal-fn.c (while_direct): New macro.
5094 (expand_while_optab_fn): New function.
5095 (convert_optab_supported_p): Likewise.
5096 (direct_while_optab_supported_p): New macro.
5097 * wide-int.h (wi::udiv_ceil): New function.
5098 * tree-vectorizer.h (rgroup_masks): New structure.
5099 (vec_loop_masks): New typedef.
5100 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5102 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5103 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5104 (vect_max_vf): New function.
5105 (slpeel_make_loop_iterate_ntimes): Delete.
5106 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5107 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5108 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5109 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5110 internal-fn.h, stor-layout.h and optabs-query.h.
5111 (vect_set_loop_mask): New function.
5112 (add_preheader_seq): Likewise.
5113 (add_header_seq): Likewise.
5114 (interleave_supported_p): Likewise.
5115 (vect_maybe_permute_loop_masks): Likewise.
5116 (vect_set_loop_masks_directly): Likewise.
5117 (vect_set_loop_condition_masked): Likewise.
5118 (vect_set_loop_condition_unmasked): New function, split out from
5119 slpeel_make_loop_iterate_ntimes.
5120 (slpeel_make_loop_iterate_ntimes): Rename to..
5121 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5122 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5123 (vect_do_peeling): Update call accordingly.
5124 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5126 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5127 mask_compare_type, can_fully_mask_p and fully_masked_p.
5128 (release_vec_loop_masks): New function.
5129 (_loop_vec_info): Use it to free the loop masks.
5130 (can_produce_all_loop_masks_p): New function.
5131 (vect_get_max_nscalars_per_iter): Likewise.
5132 (vect_verify_full_masking): Likewise.
5133 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5134 retries, and free the mask rgroups before retrying. Check loop-wide
5135 reasons for disallowing fully-masked loops. Make the final decision
5136 about whether use a fully-masked loop or not.
5137 (vect_estimate_min_profitable_iters): Do not assume that peeling
5138 for the number of iterations will be needed for fully-masked loops.
5139 (vectorizable_reduction): Disable fully-masked loops.
5140 (vectorizable_live_operation): Likewise.
5141 (vect_halve_mask_nunits): New function.
5142 (vect_double_mask_nunits): Likewise.
5143 (vect_record_loop_mask): Likewise.
5144 (vect_get_loop_mask): Likewise.
5145 (vect_transform_loop): Handle the case in which the final loop
5146 iteration might handle a partial vector. Call vect_set_loop_condition
5147 instead of slpeel_make_loop_iterate_ntimes.
5148 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5149 (check_load_store_masking): New function.
5150 (prepare_load_store_mask): Likewise.
5151 (vectorizable_store): Handle fully-masked loops.
5152 (vectorizable_load): Likewise.
5153 (supportable_widening_operation): Use vect_halve_mask_nunits for
5155 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5156 (vect_gen_while): New function.
5157 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5158 (aarch64_uqdec<mode>): New insn.
5160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5161 Alan Hayward <alan.hayward@arm.com>
5162 David Sherwood <david.sherwood@arm.com>
5164 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5165 (reduc_xor_scal_optab): New optabs.
5166 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5167 (reduc_xor_scal_@var{m}): Document.
5168 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5169 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5171 * fold-const-call.c (fold_const_call): Handle them.
5172 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5173 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5174 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5175 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5176 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5177 (UNSPEC_XORV): New unspecs.
5178 (optab): Add entries for them.
5179 (BITWISEV): New int iterator.
5180 (bit_reduc_op): New int attributes.
5182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5183 Alan Hayward <alan.hayward@arm.com>
5184 David Sherwood <david.sherwood@arm.com>
5186 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5187 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5188 * optabs.def (vec_shl_insert_optab): New optab.
5189 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5190 (duplicate_and_interleave): Likewise.
5191 * tree-vect-loop.c: Include internal-fn.h.
5192 (neutral_op_for_slp_reduction): New function, split out from
5193 get_initial_defs_for_reduction.
5194 (get_initial_def_for_reduction): Handle option 2 for variable-length
5195 vectors by loading the neutral value into a vector and then shifting
5196 the initial value into element 0.
5197 (get_initial_defs_for_reduction): Replace the code argument with
5198 the neutral value calculated by neutral_op_for_slp_reduction.
5199 Use gimple_build_vector for constant-length vectors.
5200 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5201 but the first group_size elements have a neutral value.
5202 Use duplicate_and_interleave otherwise.
5203 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5204 Update call to get_initial_defs_for_reduction. Handle SLP
5205 reductions for variable-length vectors by creating one vector
5206 result for each scalar result, with the elements associated
5207 with other scalar results stubbed out with the neutral value.
5208 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5209 Require IFN_VEC_SHL_INSERT for double reductions on
5210 variable-length vectors, or SLP reductions that have
5211 a neutral value. Require can_duplicate_and_interleave_p
5212 support for variable-length unchained SLP reductions if there
5213 is no neutral value, such as for MIN/MAX reductions. Also require
5214 the number of vector elements to be a multiple of the number of
5215 SLP statements when doing variable-length unchained SLP reductions.
5216 Update call to vect_create_epilog_for_reduction.
5217 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5218 and remove initial values.
5219 (duplicate_and_interleave): Make public.
5220 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5221 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5223 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5224 Alan Hayward <alan.hayward@arm.com>
5225 David Sherwood <david.sherwood@arm.com>
5227 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5228 (can_duplicate_and_interleave_p): New function.
5229 (vect_get_and_check_slp_defs): Take the vector of statements
5230 rather than just the current one. Remove excess parentheses.
5231 Restriction rejectinon of vect_constant_def and vect_external_def
5232 for variable-length vectors to boolean types, or types for which
5233 can_duplicate_and_interleave_p is false.
5234 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5235 (duplicate_and_interleave): New function.
5236 (vect_get_constant_vectors): Use gimple_build_vector for
5237 constant-length vectors and suitable variable-length constant
5238 vectors. Use duplicate_and_interleave for other variable-length
5239 vectors. Don't defer the update when inserting new statements.
5241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5242 Alan Hayward <alan.hayward@arm.com>
5243 David Sherwood <david.sherwood@arm.com>
5245 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5246 min_profitable_iters doesn't go negative.
5248 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5249 Alan Hayward <alan.hayward@arm.com>
5250 David Sherwood <david.sherwood@arm.com>
5252 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5253 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5254 * optabs.def (vec_mask_load_lanes_optab): New optab.
5255 (vec_mask_store_lanes_optab): Likewise.
5256 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5257 (MASK_STORE_LANES): Likewise.
5258 * internal-fn.c (mask_load_lanes_direct): New macro.
5259 (mask_store_lanes_direct): Likewise.
5260 (expand_mask_load_optab_fn): Handle masked operations.
5261 (expand_mask_load_lanes_optab_fn): New macro.
5262 (expand_mask_store_optab_fn): Handle masked operations.
5263 (expand_mask_store_lanes_optab_fn): New macro.
5264 (direct_mask_load_lanes_optab_supported_p): Likewise.
5265 (direct_mask_store_lanes_optab_supported_p): Likewise.
5266 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5268 (vect_load_lanes_supported): Likewise.
5269 * tree-vect-data-refs.c (strip_conversion): New function.
5270 (can_group_stmts_p): Likewise.
5271 (vect_analyze_data_ref_accesses): Use it instead of checking
5272 for a pair of assignments.
5273 (vect_store_lanes_supported): Take a masked_p parameter.
5274 (vect_load_lanes_supported): Likewise.
5275 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5276 vect_store_lanes_supported and vect_load_lanes_supported.
5277 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5278 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5279 parameter. Don't allow gaps for masked accesses.
5280 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5281 and vect_load_lanes_supported.
5282 (get_load_store_type): Take a masked_p parameter and update
5283 call to get_group_load_store_type.
5284 (vectorizable_store): Update call to get_load_store_type.
5285 Handle IFN_MASK_STORE_LANES.
5286 (vectorizable_load): Update call to get_load_store_type.
5287 Handle IFN_MASK_LOAD_LANES.
5289 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5290 Alan Hayward <alan.hayward@arm.com>
5291 David Sherwood <david.sherwood@arm.com>
5293 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5295 * config/aarch64/aarch64-protos.h
5296 (aarch64_sve_struct_memory_operand_p): Declare.
5297 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5298 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5299 (VPRED, vpred): Handle SVE structure modes.
5300 * config/aarch64/constraints.md (Utx): New constraint.
5301 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5302 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5303 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5304 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5305 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5306 structure modes. Split into pieces after RA.
5307 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5308 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5310 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5311 SVE structure modes.
5312 (aarch64_classify_address): Likewise.
5313 (sizetochar): Move earlier in file.
5314 (aarch64_print_operand): Handle SVE register lists.
5315 (aarch64_array_mode): New function.
5316 (aarch64_sve_struct_memory_operand_p): Likewise.
5317 (TARGET_ARRAY_MODE): Redefine.
5319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5320 Alan Hayward <alan.hayward@arm.com>
5321 David Sherwood <david.sherwood@arm.com>
5323 * target.def (array_mode): New target hook.
5324 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5325 * doc/tm.texi: Regenerate.
5326 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5327 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5328 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5330 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5333 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5334 Alan Hayward <alan.hayward@arm.com>
5335 David Sherwood <david.sherwood@arm.com>
5337 * fold-const.c (fold_binary_loc): Check the argument types
5338 rather than the result type when testing for a vector operation.
5340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5342 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5343 * doc/tm.texi: Regenerate.
5345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5346 Alan Hayward <alan.hayward@arm.com>
5347 David Sherwood <david.sherwood@arm.com>
5349 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5350 (sve): Document new AArch64 extension.
5351 * doc/md.texi (w): Extend the description of the AArch64
5352 constraint to include SVE vectors.
5353 (Upl, Upa): Document new AArch64 predicate constraints.
5354 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5356 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5357 (msve-vector-bits=): New option.
5358 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5359 SVE when these are disabled.
5360 (sve): New extension.
5361 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5362 modes. Adjust their number of units based on aarch64_sve_vg.
5363 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5364 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5365 aarch64_addr_query_type.
5366 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5367 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5368 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5369 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5370 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5371 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5372 (aarch64_simd_imm_zero_p): Delete.
5373 (aarch64_check_zero_based_sve_index_immediate): Declare.
5374 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5375 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5376 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5377 (aarch64_sve_float_mul_immediate_p): Likewise.
5378 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5380 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5381 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5382 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5383 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5384 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5385 (aarch64_regmode_natural_size): Likewise.
5386 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5387 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5389 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5390 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5391 for VG and the SVE predicate registers.
5392 (V_ALIASES): Add a "z"-prefixed alias.
5393 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5394 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5395 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5396 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5397 (REG_CLASS_NAMES): Add entries for them.
5398 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5399 and the predicate registers.
5400 (aarch64_sve_vg): Declare.
5401 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5402 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5403 (REGMODE_NATURAL_SIZE): Define.
5404 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5406 * config/aarch64/aarch64.c: Include cfgrtl.h.
5407 (simd_immediate_info): Add a constructor for series vectors,
5408 and an associated step field.
5409 (aarch64_sve_vg): New variable.
5410 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5411 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5412 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5413 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5414 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5415 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5416 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5417 (aarch64_get_mask_mode): New functions.
5418 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5419 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5420 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5421 predicate modes and predicate registers. Explicitly restrict
5422 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5423 to store a vector mode if it is recognized by
5424 aarch64_classify_vector_mode.
5425 (aarch64_regmode_natural_size): New function.
5426 (aarch64_hard_regno_caller_save_mode): Return the original mode
5428 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5429 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5430 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5431 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5433 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5434 does not overlap dest if the function is frame-related. Handle
5436 (aarch64_split_add_offset): New function.
5437 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5438 them aarch64_add_offset.
5439 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5440 and update call to aarch64_sub_sp.
5441 (aarch64_add_cfa_expression): New function.
5442 (aarch64_expand_prologue): Pass extra temporary registers to the
5443 functions above. Handle the case in which we need to emit new
5444 DW_CFA_expressions for registers that were originally saved
5445 relative to the stack pointer, but now have to be expressed
5446 relative to the frame pointer.
5447 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5449 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5450 IP0 and IP1 values for SVE frames.
5451 (aarch64_expand_vec_series): New function.
5452 (aarch64_expand_sve_widened_duplicate): Likewise.
5453 (aarch64_expand_sve_const_vector): Likewise.
5454 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5455 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5456 into the register, rather than emitting a SET directly.
5457 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5458 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5459 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5460 (offset_9bit_signed_scaled_p): New functions.
5461 (aarch64_replicate_bitmask_imm): New function.
5462 (aarch64_bitmask_imm): Use it.
5463 (aarch64_cannot_force_const_mem): Reject expressions involving
5464 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5465 (aarch64_classify_index): Handle SVE indices, by requiring
5466 a plain register index with a scale that matches the element size.
5467 (aarch64_classify_address): Handle SVE addresses. Assert that
5468 the mode of the address is VOIDmode or an integer mode.
5469 Update call to aarch64_classify_symbol.
5470 (aarch64_classify_symbolic_expression): Update call to
5471 aarch64_classify_symbol.
5472 (aarch64_const_vec_all_in_range_p): New function.
5473 (aarch64_print_vector_float_operand): Likewise.
5474 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5475 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5476 and the FP immediates 1.0 and 0.5.
5477 (aarch64_print_address_internal): Handle SVE addresses.
5478 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5479 (aarch64_regno_regclass): Handle predicate registers.
5480 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5482 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5483 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5484 (aarch64_convert_sve_vector_bits): New function.
5485 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5486 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5488 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5489 Handle SVE vector and predicate modes. Accept VL-based constants
5490 that need only one temporary register, and VL offsets that require
5491 no temporary registers.
5492 (aarch64_conditional_register_usage): Mark the predicate registers
5493 as fixed if SVE isn't available.
5494 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5495 Return true for SVE vector and predicate modes.
5496 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5497 rather than an unsigned int. Handle SVE modes.
5498 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5500 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5502 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5503 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5504 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5505 (aarch64_sve_float_mul_immediate_p): New functions.
5506 (aarch64_sve_valid_immediate): New function.
5507 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5508 Explicitly reject structure modes. Check for INDEX constants.
5509 Handle PTRUE and PFALSE constants.
5510 (aarch64_check_zero_based_sve_index_immediate): New function.
5511 (aarch64_simd_imm_zero_p): Delete.
5512 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5513 vector modes. Accept constants in the range of CNT[BHWD].
5514 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5515 ask for an Advanced SIMD mode.
5516 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5517 (aarch64_simd_vector_alignment): Handle SVE predicates.
5518 (aarch64_vectorize_preferred_vector_alignment): New function.
5519 (aarch64_simd_vector_alignment_reachable): Use it instead of
5521 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5522 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5524 (MAX_VECT_LEN): Delete.
5525 (expand_vec_perm_d): Add a vec_flags field.
5526 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5527 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5528 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5530 (aarch64_evpc_rev): Rename to...
5531 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5532 (aarch64_evpc_rev_global): New function.
5533 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5534 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5536 (aarch64_evpc_sve_tbl): New function.
5537 (aarch64_expand_vec_perm_const_1): Update after rename of
5538 aarch64_evpc_rev. Handle SVE permutes too, trying
5539 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5540 than aarch64_evpc_tbl.
5541 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5542 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5543 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5544 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5545 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5546 (aarch64_expand_sve_vcond): New functions.
5547 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5548 of aarch64_vector_mode_p.
5549 (aarch64_dwarf_poly_indeterminate_value): New function.
5550 (aarch64_compute_pressure_classes): Likewise.
5551 (aarch64_can_change_mode_class): Likewise.
5552 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5553 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5554 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5555 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5556 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5557 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5558 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5559 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5561 (Dn, Dl, Dr): Accept const as well as const_vector.
5562 (Dz): Likewise. Compare against CONST0_RTX.
5563 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5564 of "vector" where appropriate.
5565 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5566 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5567 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5568 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5569 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5570 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5571 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5572 (v_int_equiv): Extend to SVE modes.
5573 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5575 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5576 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5577 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5578 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5579 (SVE_COND_FP_CMP): New int iterators.
5580 (perm_hilo): Handle the new unpack unspecs.
5581 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5583 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5584 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5585 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5586 (aarch64_equality_operator, aarch64_constant_vector_operand)
5587 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5588 (aarch64_sve_nonimmediate_operand): Likewise.
5589 (aarch64_sve_general_operand): Likewise.
5590 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5591 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5592 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5593 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5594 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5595 (aarch64_sve_float_arith_immediate): Likewise.
5596 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5597 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5598 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5599 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5600 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5601 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5602 (aarch64_sve_float_arith_operand): Likewise.
5603 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5604 (aarch64_sve_float_mul_operand): Likewise.
5605 (aarch64_sve_vec_perm_operand): Likewise.
5606 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5607 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5608 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5609 as well as const_vector.
5610 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5611 in file. Use CONST0_RTX and CONSTM1_RTX.
5612 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5613 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5614 Use aarch64_simd_imm_zero.
5615 * config/aarch64/aarch64-sve.md: New file.
5616 * config/aarch64/aarch64.md: Include it.
5617 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5618 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5619 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5620 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5621 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5622 (sve): New attribute.
5623 (enabled): Disable instructions with the sve attribute unless
5625 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5626 aarch64_expand_mov_immediate.
5627 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5628 CNT[BHSD] immediates.
5629 (movti): Split CONST_POLY_INT moves into two halves.
5630 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5631 Split additions that need a temporary here if the destination
5632 is the stack pointer.
5633 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5634 (*add<mode>3_poly_1): New instruction.
5635 (set_clobber_cc): New expander.
5637 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5639 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5640 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5641 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5642 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5643 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5644 Change innermode from fixed_mode_size to machine_mode.
5645 (simplify_subreg): Update call accordingly. Handle a constant-sized
5646 subreg of a variable-length CONST_VECTOR.
5648 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5649 Alan Hayward <alan.hayward@arm.com>
5650 David Sherwood <david.sherwood@arm.com>
5652 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5653 (add_offset_to_base): New function, split out from...
5654 (create_mem_ref): ...here. When handling a scale other than 1,
5655 check first whether the address is valid without the offset.
5656 Add it into the base if so, leaving the index and scale as-is.
5658 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5661 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5662 fold_for_warn before checking if arg2 is INTEGER_CST.
5664 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5666 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5667 (store_multiple_operation): Delete.
5668 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5669 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5670 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5671 guarded by TARGET_STRING.
5672 (rs6000_output_load_multiple): Delete.
5673 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5674 OPTION_MASK_STRING / TARGET_STRING handling.
5675 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5676 (const rs6000_opt_masks) <"string">: Change mask to 0.
5677 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5678 (MASK_STRING): Delete.
5679 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5681 (load_multiple): Delete.
5688 (store_multiple): Delete.
5695 (movmemsi_8reg): Delete.
5696 (corresponding unnamed define_insn): Delete.
5697 (movmemsi_6reg): Delete.
5698 (corresponding unnamed define_insn): Delete.
5699 (movmemsi_4reg): Delete.
5700 (corresponding unnamed define_insn): Delete.
5701 (movmemsi_2reg): Delete.
5702 (corresponding unnamed define_insn): Delete.
5703 (movmemsi_1reg): Delete.
5704 (corresponding unnamed define_insn): Delete.
5705 * config/rs6000/rs6000.opt (mno-string): New.
5706 (mstring): Replace by deprecation warning stub.
5707 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5709 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5711 * regrename.c (regrename_do_replace): If replacing the same
5712 reg multiple times, try to reuse last created gen_raw_REG.
5715 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5716 main to workaround a bug in GDB.
5718 2018-01-12 Tom de Vries <tom@codesourcery.com>
5721 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5723 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5725 PR rtl-optimization/80481
5726 * ira-color.c (get_cap_member): New function.
5727 (allocnos_conflict_by_live_ranges_p): Use it.
5728 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5729 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5731 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5734 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5735 (*saddl_se_1): Ditto.
5737 (*ssubl_se_1): Ditto.
5739 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5741 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5742 rather than wi::to_widest for DR_INITs.
5743 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5744 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5745 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5747 (vect_analyze_group_access_1): Note that here.
5749 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5751 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5752 polynomial type sizes.
5754 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5756 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5757 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5758 (gimple_add_tmp_var): Likewise.
5760 2018-01-12 Martin Liska <mliska@suse.cz>
5762 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5763 (gimple_alloc_sizes): Likewise.
5764 (dump_gimple_statistics): Use PRIu64 in printf format.
5765 * gimple.h: Change uint64_t to int.
5767 2018-01-12 Martin Liska <mliska@suse.cz>
5769 * tree-core.h: Use uint64_t instead of int.
5770 * tree.c (tree_node_counts): Likewise.
5771 (tree_node_sizes): Likewise.
5772 (dump_tree_statistics): Use PRIu64 in printf format.
5774 2018-01-12 Martin Liska <mliska@suse.cz>
5776 * Makefile.in: As qsort_chk is implemented in vec.c, add
5777 vec.o to linkage of gencfn-macros.
5778 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5779 passing the info to record_node_allocation_statistics.
5780 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5782 * ggc-common.c (struct ggc_usage): Add operator== and use
5783 it in operator< and compare function.
5784 * mem-stats.h (struct mem_usage): Likewise.
5785 * vec.c (struct vec_usage): Remove operator< and compare
5786 function. Can be simply inherited.
5788 2018-01-12 Martin Jambor <mjambor@suse.cz>
5791 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5792 * tree-ssa-math-opts.c: Include domwalk.h.
5793 (convert_mult_to_fma_1): New function.
5794 (fma_transformation_info): New type.
5795 (fma_deferring_state): Likewise.
5796 (cancel_fma_deferring): New function.
5797 (result_of_phi): Likewise.
5798 (last_fma_candidate_feeds_initial_phi): Likewise.
5799 (convert_mult_to_fma): Added deferring logic, split actual
5800 transformation to convert_mult_to_fma_1.
5801 (math_opts_dom_walker): New type.
5802 (math_opts_dom_walker::after_dom_children): New method, body moved
5803 here from pass_optimize_widening_mul::execute, added deferring logic
5805 (pass_optimize_widening_mul::execute): Moved most of code to
5806 math_opts_dom_walker::after_dom_children.
5807 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5808 * config/i386/i386.c (ix86_option_override_internal): Added
5809 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5811 2018-01-12 Richard Biener <rguenther@suse.de>
5814 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5815 inline instance vars.
5817 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5820 * config/rx/rx.c (rx_is_restricted_memory_address):
5823 2018-01-12 Richard Biener <rguenther@suse.de>
5825 PR tree-optimization/80846
5826 * target.def (split_reduction): New target hook.
5827 * targhooks.c (default_split_reduction): New function.
5828 * targhooks.h (default_split_reduction): Declare.
5829 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5830 target requests first reduce vectors by combining low and high
5832 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5833 (get_vectype_for_scalar_type_and_size): Export.
5834 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5835 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5836 * doc/tm.texi: Regenerate.
5837 * config/i386/i386.c (ix86_split_reduction): Implement
5838 TARGET_VECTORIZE_SPLIT_REDUCTION.
5840 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5843 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5844 in PIC mode except for TARGET_VXWORKS_RTP.
5845 * config/sparc/sparc.c: Include cfgrtl.h.
5846 (TARGET_INIT_PIC_REG): Define.
5847 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5848 (sparc_pic_register_p): New predicate.
5849 (sparc_legitimate_address_p): Use it.
5850 (sparc_legitimize_pic_address): Likewise.
5851 (sparc_delegitimize_address): Likewise.
5852 (sparc_mode_dependent_address_p): Likewise.
5853 (gen_load_pcrel_sym): Remove 4th parameter.
5854 (load_got_register): Adjust call to above. Remove obsolete stuff.
5855 (sparc_expand_prologue): Do not call load_got_register here.
5856 (sparc_flat_expand_prologue): Likewise.
5857 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5858 (sparc_use_pseudo_pic_reg): New function.
5859 (sparc_init_pic_reg): Likewise.
5860 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5861 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5863 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5865 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5866 Add item for branch_cost.
5868 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5870 PR rtl-optimization/83565
5871 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5872 not extend the result to a larger mode for rotate operations.
5873 (num_sign_bit_copies1): Likewise.
5875 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5878 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5880 Use values-Xc.o for -pedantic.
5881 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5883 2018-01-12 Martin Liska <mliska@suse.cz>
5886 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5888 (possible_polymorphic_call_targets): Use it.
5889 (ipa_devirt): Likewise.
5891 2018-01-12 Martin Liska <mliska@suse.cz>
5893 * profile-count.h (enum profile_quality): Use 0 as invalid
5894 enum value of profile_quality.
5896 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5898 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5899 -mext-string options.
5901 2018-01-12 Richard Biener <rguenther@suse.de>
5903 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5904 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5905 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5907 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5909 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5911 * configure.ac (--with-long-double-format): Add support for the
5912 configuration option to change the default long double format on
5914 * config.gcc (powerpc*-linux*-*): Likewise.
5915 * configure: Regenerate.
5916 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5917 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5918 used without modification.
5920 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5922 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5923 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5924 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5925 MISC_BUILTIN_SPEC_BARRIER.
5926 (rs6000_init_builtins): Likewise.
5927 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5929 (speculation_barrier): New define_insn.
5930 * doc/extend.texi: Document __builtin_speculation_barrier.
5932 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5935 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5936 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5937 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5939 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5940 integral modes instead of "ss" and "sd".
5941 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5942 vectors with 32-bit and 64-bit elements.
5943 (vecdupssescalarmodesuffix): New mode attribute.
5944 (vec_dup<mode>): Use it.
5946 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5949 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5950 frame if argument is passed on stack.
5952 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5955 * ree.c (combine_reaching_defs): Optimize also
5956 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5957 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5959 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5962 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5964 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5967 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5968 after they are computed.
5970 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5972 PR tree-optimization/83695
5973 * gimple-loop-linterchange.cc
5974 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5975 reset cached scev information after interchange.
5976 (pass_linterchange::execute): Remove call to scev_reset_htab.
5978 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5980 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5981 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5982 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5983 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5984 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5985 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5986 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5987 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5988 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5989 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5990 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5991 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5992 (V_lane_reg): Likewise.
5993 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5995 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5996 (vfmal_lane_low<mode>_intrinsic,
5997 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5998 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5999 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6000 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6001 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6002 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6004 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6006 * config/arm/arm-cpus.in (fp16fml): New feature.
6007 (ALL_SIMD): Add fp16fml.
6008 (armv8.2-a): Add fp16fml as an option.
6009 (armv8.3-a): Likewise.
6010 (armv8.4-a): Add fp16fml as part of fp16.
6011 * config/arm/arm.h (TARGET_FP16FML): Define.
6012 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6014 * config/arm/arm-modes.def (V2HF): Define.
6015 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6016 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6017 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6018 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6019 vfmsl_low, vfmsl_high): New set of builtins.
6020 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6021 (vfml_op): New code attribute.
6022 (VFMLHALVES): New int iterator.
6023 (VFML, VFMLSEL): New mode attributes.
6024 (V_reg): Define mapping for V2HF.
6025 (V_hi, V_lo): New mode attributes.
6026 (VF_constraint): Likewise.
6027 (vfml_half, vfml_half_selector): New int attributes.
6028 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6030 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6031 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6033 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6034 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6035 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6036 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6038 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6039 Document new effective target and option set.
6041 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6043 * config/arm/arm-cpus.in (armv8_4): New feature.
6044 (ARMv8_4a): New fgroup.
6045 (armv8.4-a): New arch.
6046 * config/arm/arm-tables.opt: Regenerate.
6047 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6048 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6049 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6050 Add matching rules for -march=armv8.4-a and extensions.
6051 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6053 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6056 * config/rx/rx.md (BW): New mode attribute.
6057 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6059 2018-01-11 Richard Biener <rguenther@suse.de>
6061 PR tree-optimization/83435
6062 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6063 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6064 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6066 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6067 Alan Hayward <alan.hayward@arm.com>
6068 David Sherwood <david.sherwood@arm.com>
6070 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6072 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6073 (aarch64_print_address_internal): Use it to check for a zero offset.
6075 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6076 Alan Hayward <alan.hayward@arm.com>
6077 David Sherwood <david.sherwood@arm.com>
6079 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6080 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6081 Return a poly_int64 rather than a HOST_WIDE_INT.
6082 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6083 rather than a HOST_WIDE_INT.
6084 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6085 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6086 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6087 final_offset from HOST_WIDE_INT to poly_int64.
6088 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6089 to_constant when getting the number of units in an Advanced SIMD
6091 (aarch64_builtin_vectorized_function): Check for a constant number
6093 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6095 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6096 attribute instead of GET_MODE_NUNITS.
6097 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6098 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6099 GET_MODE_SIZE for fixed-size registers.
6100 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6101 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6102 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6103 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6104 (aarch64_print_operand, aarch64_print_address_internal)
6105 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6106 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6107 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6108 Handle polynomial GET_MODE_SIZE.
6109 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6110 wider than SImode without modification.
6111 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6112 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6113 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6114 passing and returning SVE modes.
6115 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6116 rather than GEN_INT.
6117 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6118 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6119 (aarch64_allocate_and_probe_stack_space): Likewise.
6120 (aarch64_layout_frame): Cope with polynomial offsets.
6121 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6122 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6124 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6125 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6126 poly_int64 rather than a HOST_WIDE_INT.
6127 (aarch64_get_separate_components, aarch64_process_components)
6128 (aarch64_expand_prologue, aarch64_expand_epilogue)
6129 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6130 (aarch64_anchor_offset): New function, split out from...
6131 (aarch64_legitimize_address): ...here.
6132 (aarch64_builtin_vectorization_cost): Handle polynomial
6133 TYPE_VECTOR_SUBPARTS.
6134 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6136 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6137 number of elements from the PARALLEL rather than the mode.
6138 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6139 rather than GET_MODE_BITSIZE.
6140 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6141 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6142 (aarch64_expand_vec_perm_const_1): Handle polynomial
6143 d->perm.length () and d->perm elements.
6144 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6145 Apply to_constant to d->perm elements.
6146 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6147 polynomial CONST_VECTOR_NUNITS.
6148 (aarch64_move_pointer): Take amount as a poly_int64 rather
6150 (aarch64_progress_pointer): Avoid temporary variable.
6151 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6152 the mode attribute instead of GET_MODE.
6154 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6155 Alan Hayward <alan.hayward@arm.com>
6156 David Sherwood <david.sherwood@arm.com>
6158 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6159 x exists before using it.
6160 (aarch64_add_constant_internal): Rename to...
6161 (aarch64_add_offset_1): ...this. Replace regnum with separate
6162 src and dest rtxes. Handle the case in which they're different,
6163 including when the offset is zero. Replace scratchreg with an rtx.
6164 Use 2 additions if there is no spare register into which we can
6165 move a 16-bit constant.
6166 (aarch64_add_constant): Delete.
6167 (aarch64_add_offset): Replace reg with separate src and dest
6168 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6169 Use aarch64_add_offset_1.
6170 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6171 an rtx rather than an int. Take the delta as a poly_int64
6172 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6173 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6174 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6175 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6176 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6178 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6179 aarch64_add_constant.
6181 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6183 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6184 Use scalar_float_mode.
6186 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6188 * config/aarch64/aarch64-simd.md
6189 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6190 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6191 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6192 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6193 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6194 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6195 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6196 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6197 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6198 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6200 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6203 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6204 targ_options->x_arm_arch_string is non NULL.
6206 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6208 * config/aarch64/aarch64.h
6209 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6211 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6214 * expmed.c (emit_store_flag_force): Swap if const op0
6215 and change VOIDmode to mode of op0.
6217 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6219 PR rtl-optimization/83761
6220 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6221 than bytes to mode_for_size.
6223 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6226 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6227 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6230 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6233 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6234 when in layout mode.
6235 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6236 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6239 2018-01-10 Michael Collison <michael.collison@arm.com>
6241 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6242 * config/aarch64/aarch64-option-extension.def: Add
6243 AARCH64_OPT_EXTENSION of 'fp16fml'.
6244 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6245 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6246 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6247 * config/aarch64/constraints.md (Ui7): New constraint.
6248 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6249 (VFMLA_SEL_W): Ditto.
6252 (VFMLA16_LOW): New int iterator.
6253 (VFMLA16_HIGH): Ditto.
6254 (UNSPEC_FMLAL): New unspec.
6255 (UNSPEC_FMLSL): Ditto.
6256 (UNSPEC_FMLAL2): Ditto.
6257 (UNSPEC_FMLSL2): Ditto.
6258 (f16mac): New code attribute.
6259 * config/aarch64/aarch64-simd-builtins.def
6260 (aarch64_fmlal_lowv2sf): Ditto.
6261 (aarch64_fmlsl_lowv2sf): Ditto.
6262 (aarch64_fmlalq_lowv4sf): Ditto.
6263 (aarch64_fmlslq_lowv4sf): Ditto.
6264 (aarch64_fmlal_highv2sf): Ditto.
6265 (aarch64_fmlsl_highv2sf): Ditto.
6266 (aarch64_fmlalq_highv4sf): Ditto.
6267 (aarch64_fmlslq_highv4sf): Ditto.
6268 (aarch64_fmlal_lane_lowv2sf): Ditto.
6269 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6270 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6271 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6272 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6273 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6274 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6275 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6276 (aarch64_fmlal_lane_highv2sf): Ditto.
6277 (aarch64_fmlsl_lane_highv2sf): Ditto.
6278 (aarch64_fmlal_laneq_highv2sf): Ditto.
6279 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6280 (aarch64_fmlalq_lane_highv4sf): Ditto.
6281 (aarch64_fmlsl_lane_highv4sf): Ditto.
6282 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6283 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6284 * config/aarch64/aarch64-simd.md:
6285 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6286 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6287 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6288 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6289 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6290 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6291 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6292 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6293 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6294 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6295 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6296 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6297 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6298 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6299 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6300 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6301 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6302 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6303 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6304 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6305 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6306 (vfmlsl_low_u32): Ditto.
6307 (vfmlalq_low_u32): Ditto.
6308 (vfmlslq_low_u32): Ditto.
6309 (vfmlal_high_u32): Ditto.
6310 (vfmlsl_high_u32): Ditto.
6311 (vfmlalq_high_u32): Ditto.
6312 (vfmlslq_high_u32): Ditto.
6313 (vfmlal_lane_low_u32): Ditto.
6314 (vfmlsl_lane_low_u32): Ditto.
6315 (vfmlal_laneq_low_u32): Ditto.
6316 (vfmlsl_laneq_low_u32): Ditto.
6317 (vfmlalq_lane_low_u32): Ditto.
6318 (vfmlslq_lane_low_u32): Ditto.
6319 (vfmlalq_laneq_low_u32): Ditto.
6320 (vfmlslq_laneq_low_u32): Ditto.
6321 (vfmlal_lane_high_u32): Ditto.
6322 (vfmlsl_lane_high_u32): Ditto.
6323 (vfmlal_laneq_high_u32): Ditto.
6324 (vfmlsl_laneq_high_u32): Ditto.
6325 (vfmlalq_lane_high_u32): Ditto.
6326 (vfmlslq_lane_high_u32): Ditto.
6327 (vfmlalq_laneq_high_u32): Ditto.
6328 (vfmlslq_laneq_high_u32): Ditto.
6329 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6330 (AARCH64_FL_FOR_ARCH8_4): New.
6331 (AARCH64_ISA_F16FML): New ISA flag.
6332 (TARGET_F16FML): New feature flag for fp16fml.
6333 (doc/invoke.texi): Document new fp16fml option.
6335 2018-01-10 Michael Collison <michael.collison@arm.com>
6337 * config/aarch64/aarch64-builtins.c:
6338 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6339 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6340 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6341 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6342 (AARCH64_ISA_SHA3): New ISA flag.
6343 (TARGET_SHA3): New feature flag for sha3.
6344 * config/aarch64/iterators.md (sha512_op): New int attribute.
6345 (CRYPTO_SHA512): New int iterator.
6346 (UNSPEC_SHA512H): New unspec.
6347 (UNSPEC_SHA512H2): Ditto.
6348 (UNSPEC_SHA512SU0): Ditto.
6349 (UNSPEC_SHA512SU1): Ditto.
6350 * config/aarch64/aarch64-simd-builtins.def
6351 (aarch64_crypto_sha512hqv2di): New builtin.
6352 (aarch64_crypto_sha512h2qv2di): Ditto.
6353 (aarch64_crypto_sha512su0qv2di): Ditto.
6354 (aarch64_crypto_sha512su1qv2di): Ditto.
6355 (aarch64_eor3qv8hi): Ditto.
6356 (aarch64_rax1qv2di): Ditto.
6357 (aarch64_xarqv2di): Ditto.
6358 (aarch64_bcaxqv8hi): Ditto.
6359 * config/aarch64/aarch64-simd.md:
6360 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6361 (aarch64_crypto_sha512su0qv2di): Ditto.
6362 (aarch64_crypto_sha512su1qv2di): Ditto.
6363 (aarch64_eor3qv8hi): Ditto.
6364 (aarch64_rax1qv2di): Ditto.
6365 (aarch64_xarqv2di): Ditto.
6366 (aarch64_bcaxqv8hi): Ditto.
6367 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6368 (vsha512h2q_u64): Ditto.
6369 (vsha512su0q_u64): Ditto.
6370 (vsha512su1q_u64): Ditto.
6371 (veor3q_u16): Ditto.
6372 (vrax1q_u64): Ditto.
6374 (vbcaxq_u16): Ditto.
6375 * config/arm/types.md (crypto_sha512): New type attribute.
6376 (crypto_sha3): Ditto.
6377 (doc/invoke.texi): Document new sha3 option.
6379 2018-01-10 Michael Collison <michael.collison@arm.com>
6381 * config/aarch64/aarch64-builtins.c:
6382 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6383 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6384 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6385 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6386 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6387 (AARCH64_ISA_SM4): New ISA flag.
6388 (TARGET_SM4): New feature flag for sm4.
6389 * config/aarch64/aarch64-simd-builtins.def
6390 (aarch64_sm3ss1qv4si): Ditto.
6391 (aarch64_sm3tt1aq4si): Ditto.
6392 (aarch64_sm3tt1bq4si): Ditto.
6393 (aarch64_sm3tt2aq4si): Ditto.
6394 (aarch64_sm3tt2bq4si): Ditto.
6395 (aarch64_sm3partw1qv4si): Ditto.
6396 (aarch64_sm3partw2qv4si): Ditto.
6397 (aarch64_sm4eqv4si): Ditto.
6398 (aarch64_sm4ekeyqv4si): Ditto.
6399 * config/aarch64/aarch64-simd.md:
6400 (aarch64_sm3ss1qv4si): Ditto.
6401 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6402 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6403 (aarch64_sm4eqv4si): Ditto.
6404 (aarch64_sm4ekeyqv4si): Ditto.
6405 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6406 (sm3part_op): Ditto.
6407 (CRYPTO_SM3TT): Ditto.
6408 (CRYPTO_SM3PART): Ditto.
6409 (UNSPEC_SM3SS1): New unspec.
6410 (UNSPEC_SM3TT1A): Ditto.
6411 (UNSPEC_SM3TT1B): Ditto.
6412 (UNSPEC_SM3TT2A): Ditto.
6413 (UNSPEC_SM3TT2B): Ditto.
6414 (UNSPEC_SM3PARTW1): Ditto.
6415 (UNSPEC_SM3PARTW2): Ditto.
6416 (UNSPEC_SM4E): Ditto.
6417 (UNSPEC_SM4EKEY): Ditto.
6418 * config/aarch64/constraints.md (Ui2): New constraint.
6419 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6420 * config/arm/types.md (crypto_sm3): New type attribute.
6421 (crypto_sm4): Ditto.
6422 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6423 (vsm3tt1aq_u32): Ditto.
6424 (vsm3tt1bq_u32): Ditto.
6425 (vsm3tt2aq_u32): Ditto.
6426 (vsm3tt2bq_u32): Ditto.
6427 (vsm3partw1q_u32): Ditto.
6428 (vsm3partw2q_u32): Ditto.
6429 (vsm4eq_u32): Ditto.
6430 (vsm4ekeyq_u32): Ditto.
6431 (doc/invoke.texi): Document new sm4 option.
6433 2018-01-10 Michael Collison <michael.collison@arm.com>
6435 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6436 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6437 (AARCH64_FL_FOR_ARCH8_4): New.
6438 (AARCH64_FL_V8_4): New flag.
6439 (doc/invoke.texi): Document new armv8.4-a option.
6441 2018-01-10 Michael Collison <michael.collison@arm.com>
6443 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6444 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6445 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6446 * config/aarch64/aarch64-option-extension.def: Add
6447 AARCH64_OPT_EXTENSION of 'sha2'.
6448 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6449 (crypto): Disable sha2 and aes if crypto disabled.
6450 (crypto): Enable aes and sha2 if enabled.
6451 (simd): Disable sha2 and aes if simd disabled.
6452 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6454 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6455 (TARGET_SHA2): New feature flag for sha2.
6456 (TARGET_AES): New feature flag for aes.
6457 * config/aarch64/aarch64-simd.md:
6458 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6459 conditional on TARGET_AES.
6460 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6461 (aarch64_crypto_sha1hsi): Make pattern conditional
6463 (aarch64_crypto_sha1hv4si): Ditto.
6464 (aarch64_be_crypto_sha1hv4si): Ditto.
6465 (aarch64_crypto_sha1su1v4si): Ditto.
6466 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6467 (aarch64_crypto_sha1su0v4si): Ditto.
6468 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6469 (aarch64_crypto_sha256su0v4si): Ditto.
6470 (aarch64_crypto_sha256su1v4si): Ditto.
6471 (doc/invoke.texi): Document new aes and sha2 options.
6473 2018-01-10 Martin Sebor <msebor@redhat.com>
6475 PR tree-optimization/83781
6476 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6479 2018-01-11 Martin Sebor <msebor@gmail.com>
6480 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6482 PR tree-optimization/83501
6483 PR tree-optimization/81703
6485 * tree-ssa-strlen.c (get_string_cst): Rename...
6486 (get_string_len): ...to this. Handle global constants.
6487 (handle_char_store): Adjust.
6489 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6490 Jim Wilson <jimw@sifive.com>
6492 * config/riscv/riscv-protos.h (riscv_output_return): New.
6493 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6494 (riscv_attribute_table, riscv_output_return),
6495 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6496 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6497 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6499 (riscv_expand_prologue): Add early return for naked function.
6500 (riscv_expand_epilogue): Likewise.
6501 (riscv_function_ok_for_sibcall): Return false for naked function.
6502 (riscv_set_current_function): New.
6503 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6504 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6505 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6506 * doc/extend.texi (RISC-V Function Attributes): New.
6508 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6510 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6511 check for 128-bit long double before checking TCmode.
6512 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6513 128-bit long doubles before checking TFmode or TCmode.
6514 (FLOAT128_IBM_P): Likewise.
6516 2018-01-10 Martin Sebor <msebor@redhat.com>
6518 PR tree-optimization/83671
6519 * builtins.c (c_strlen): Unconditionally return zero for the empty
6521 Use -Warray-bounds for warnings.
6522 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6523 for non-constant array indices with COMPONENT_REF, arrays of
6524 arrays, and pointers to arrays.
6525 (gimple_fold_builtin_strlen): Determine and set length range for
6526 non-constant character arrays.
6528 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6531 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6534 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6536 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6538 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6541 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6542 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6543 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6544 indexed_or_indirect_operand predicate.
6545 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6546 (*vsx_le_perm_load_v8hi): Likewise.
6547 (*vsx_le_perm_load_v16qi): Likewise.
6548 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6549 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6550 (*vsx_le_perm_store_v8hi): Likewise.
6551 (*vsx_le_perm_store_v16qi): Likewise.
6552 (eight unnamed splitters): Likewise.
6554 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6556 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6557 * config/rs6000/emmintrin.h: Likewise.
6558 * config/rs6000/mmintrin.h: Likewise.
6559 * config/rs6000/xmmintrin.h: Likewise.
6561 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6564 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6566 * tree.c (tree_nop_conversion): Return true for location wrapper
6568 (maybe_wrap_with_location): New function.
6569 (selftest::check_strip_nops): New function.
6570 (selftest::test_location_wrappers): New function.
6571 (selftest::tree_c_tests): Call it.
6572 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6573 (maybe_wrap_with_location): New decl.
6574 (EXPR_LOCATION_WRAPPER_P): New macro.
6575 (location_wrapper_p): New inline function.
6576 (tree_strip_any_location_wrapper): New inline function.
6578 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6581 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6582 stack_realign_offset for the largest alignment of stack slot
6584 (ix86_find_max_used_stack_alignment): New function.
6585 (ix86_finalize_stack_frame_flags): Use it. Set
6586 max_used_stack_alignment if we don't realign stack.
6587 * config/i386/i386.h (machine_function): Add
6588 max_used_stack_alignment.
6590 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6592 * config/arm/arm.opt (-mbranch-cost): New option.
6593 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6596 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6599 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6600 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6602 2018-01-10 Richard Biener <rguenther@suse.de>
6605 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6606 early out so it also covers the case where we have a non-NULL
6609 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6611 PR tree-optimization/83753
6612 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6613 for non-strided grouped accesses if the number of elements is 1.
6615 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6618 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6619 * i386.h (TARGET_USE_GATHER): Define.
6620 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6622 2018-01-10 Martin Liska <mliska@suse.cz>
6625 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6626 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6628 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6629 CLEANUP_NO_PARTITIONING is not set.
6631 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6633 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6634 for vectors, as a partial revert of r254296.
6635 * rtl.h (const_vec_p): Delete.
6636 (const_vec_duplicate_p): Don't test for vector CONSTs.
6637 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6638 * expmed.c (make_tree): Likewise.
6641 * common.md (E, F): Use CONSTANT_P instead of checking for
6643 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6644 checking for CONST_VECTOR.
6646 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6649 * predict.c (force_edge_cold): Handle in more sane way edges
6652 2018-01-09 Carl Love <cel@us.ibm.com>
6654 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6656 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6657 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6658 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6659 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6660 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6661 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6662 * config/rs6000/rs6000-protos.h: Add extern defition for
6663 rs6000_generate_float2_double_code.
6664 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6666 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6667 (float2_v2df): Add define_expand.
6669 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6672 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6673 op_mode in the force_to_mode call.
6675 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6677 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6678 instead of checking each element individually.
6679 (aarch64_evpc_uzp): Likewise.
6680 (aarch64_evpc_zip): Likewise.
6681 (aarch64_evpc_ext): Likewise.
6682 (aarch64_evpc_rev): Likewise.
6683 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6684 instead of checking each element individually. Return true without
6686 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6687 whether all selected elements come from the same input, instead of
6688 checking each element individually. Remove calls to gen_rtx_REG,
6689 start_sequence and end_sequence and instead assert that no rtl is
6692 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6694 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6695 order of HIGH and CONST checks.
6697 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6699 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6700 if the destination isn't an SSA_NAME.
6702 2018-01-09 Richard Biener <rguenther@suse.de>
6704 PR tree-optimization/83668
6705 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6707 (canonicalize_loop_form): ... here, renamed from ...
6708 (canonicalize_loop_closed_ssa_form): ... this and amended to
6709 swap successor edges for loop exit blocks to make us use
6710 the RPO order we need for initial schedule generation.
6712 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6714 PR tree-optimization/64811
6715 * match.pd: When optimizing comparisons with Inf, avoid
6716 introducing or losing exceptions from comparisons with NaN.
6718 2018-01-09 Martin Liska <mliska@suse.cz>
6721 * asan.c (shadow_mem_size): Add gcc_assert.
6723 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6725 Don't save registers in main().
6728 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6729 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6730 * config/avr/avr.c (avr_set_current_function): Don't error if
6731 naked, OS_task or OS_main are specified at the same time.
6732 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6734 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6736 * common/config/avr/avr-common.c (avr_option_optimization_table):
6737 Switch on -mmain-is-OS_task for optimizing compilations.
6739 2018-01-09 Richard Biener <rguenther@suse.de>
6741 PR tree-optimization/83572
6742 * graphite.c: Include cfganal.h.
6743 (graphite_transform_loops): Connect infinite loops to exit
6744 and remove fake edges at the end.
6746 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6748 * ipa-inline.c (edge_badness): Revert accidental checkin.
6750 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6753 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6754 symbols; not inline clones.
6756 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6759 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6760 hard registers. Formatting fixes.
6762 PR preprocessor/83722
6763 * gcc.c (try_generate_repro): Pass
6764 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6765 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6768 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6769 Kito Cheng <kito.cheng@gmail.com>
6771 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6772 (riscv_leaf_function_p): Delete.
6773 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6775 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6777 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6779 (do_ifelse): New function.
6780 (do_isel): New function.
6781 (do_sub3): New function.
6782 (do_add3): New function.
6783 (do_load_mask_compare): New function.
6784 (do_overlap_load_compare): New function.
6785 (expand_compare_loop): New function.
6786 (expand_block_compare): Call expand_compare_loop() when appropriate.
6787 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6789 (-mblock-compare-inline-loop-limit): New option.
6791 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6794 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6795 Reverse order of second and third operands in first alternative.
6796 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6797 of first and second elements in UNSPEC_VPERMR vector.
6798 (altivec_expand_vec_perm_le): Likewise.
6800 2018-01-08 Jeff Law <law@redhat.com>
6802 PR rtl-optimizatin/81308
6803 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6804 (process_switch): If group_case_labels makes a change, then set
6806 (pass_convert_switch::execute): If a switch is converted, then
6807 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6809 PR rtl-optimization/81308
6810 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6813 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6815 PR target/83663 - Revert r255946
6816 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6817 generation for cases where splatting a value is not useful.
6818 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6819 across a vec_duplicate and a paradoxical subreg forming a vector
6820 mode to a vec_concat.
6822 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6824 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6825 -march=armv8.3-a variants.
6826 * config/arm/t-multilib: Likewise.
6827 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6829 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6831 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6833 (cceq_ior_compare_complement): Give it a name so I can use it, and
6834 change boolean_or_operator predicate to boolean_operator so it can
6835 be used to generate a crand.
6836 (eqne): New code iterator.
6837 (bd/bd_neg): New code_attrs.
6838 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6839 a single define_insn.
6840 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6841 decrement (bdnzt/bdnzf/bdzt/bdzf).
6842 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6843 with the new names of the branch decrement patterns, and added the
6844 names of the branch decrement conditional patterns.
6846 2018-01-08 Richard Biener <rguenther@suse.de>
6848 PR tree-optimization/83563
6849 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6852 2018-01-08 Richard Biener <rguenther@suse.de>
6855 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6857 2018-01-08 Richard Biener <rguenther@suse.de>
6859 PR tree-optimization/83685
6860 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6861 references to abnormals.
6863 2018-01-08 Richard Biener <rguenther@suse.de>
6866 * dwarf2out.c (output_indirect_strings): Handle empty
6867 skeleton_debug_str_hash.
6868 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6870 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6872 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6873 (emit_store_direct): Likewise.
6874 (arc_trampoline_adjust_address): Likewise.
6875 (arc_asm_trampoline_template): New function.
6876 (arc_initialize_trampoline): Use asm_trampoline_template.
6877 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6878 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6879 * config/arc/arc.md (flush_icache): Delete pattern.
6881 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6883 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6884 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6887 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6890 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6891 by not USED_FOR_TARGET.
6892 (make_pass_resolve_sw_modes): Likewise.
6894 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6896 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6899 2018-01-08 Richard Biener <rguenther@suse.de>
6902 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6904 2018-01-08 Richard Biener <rguenther@suse.de>
6907 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6909 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6912 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6913 basic blocks with a small number of successors.
6914 (convert_control_dep_chain_into_preds): Improve handling of
6916 (dump_predicates): Split apart into...
6917 (dump_pred_chain): ...here...
6918 (dump_pred_info): ...and here.
6919 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6920 (can_chain_union_be_invalidated_p): Improve check for invalidation
6922 (uninit_uses_cannot_happen): Avoid unnecessary if
6923 convert_control_dep_chain_into_preds yielded nothing.
6925 2018-01-06 Martin Sebor <msebor@redhat.com>
6927 PR tree-optimization/83640
6928 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6929 subtracting negative offset from size.
6930 (builtin_access::overlap): Adjust offset bounds of the access to fall
6931 within the size of the object if possible.
6933 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6935 PR rtl-optimization/83699
6936 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6937 extract_bit_field_as_subreg to cases in which the extracted
6938 value is also a vector.
6940 * lra-constraints.c (process_alt_operands): Test for the equivalence
6941 substitutions when detecting a possible reload cycle.
6943 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6946 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6947 by default if flag_selective_schedling{,2}. Formatting fixes.
6949 PR rtl-optimization/83682
6950 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6951 if it has non-VECTOR_MODE element mode.
6952 (vec_duplicate_p): Likewise.
6955 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6956 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6958 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6961 * config/i386/i386-builtin.def
6962 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6963 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6964 Require also OPTION_MASK_ISA_AVX512F in addition to
6965 OPTION_MASK_ISA_GFNI.
6966 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6967 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6968 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6969 to OPTION_MASK_ISA_GFNI.
6970 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6971 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6972 OPTION_MASK_ISA_AVX512BW.
6973 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6974 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6975 addition to OPTION_MASK_ISA_GFNI.
6976 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6977 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6978 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6979 to OPTION_MASK_ISA_GFNI.
6980 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6981 a requirement for all ISAs rather than any of them with a few
6983 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6985 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6986 bitmasks to be enabled with 3 exceptions, instead of requiring any
6987 enabled ISA with lots of exceptions.
6988 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6989 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6990 Change avx512bw in isa attribute to avx512f.
6991 * config/i386/sgxintrin.h: Add license boilerplate.
6992 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6993 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6994 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6995 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6997 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6998 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6999 temporarily sse2 rather than sse if not enabled already.
7002 * config/i386/sse.md (VI248_VLBW): Rename to ...
7003 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7004 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7005 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7006 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7007 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7008 mode iterator instead of VI248_VLBW.
7010 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7012 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7013 (record_modified): Skip clobbers; add debug output.
7014 (param_change_prob): Use sreal frequencies.
7016 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7018 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7019 punt for user-aligned variables.
7021 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7023 * tree-chrec.c (chrec_contains_symbols): Return true for
7026 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7029 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7030 of (x|y) == x for BICS pattern.
7032 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7034 PR tree-optimization/83605
7035 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7036 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7039 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7041 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7042 * config/epiphany/rtems.h: New file.
7044 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7045 Uros Bizjak <ubizjak@gmail.com>
7048 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7049 QIreg_operand instead of register_operand predicate.
7050 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7051 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7052 comments instead of -fmitigate[-_]rop.
7054 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7057 * cgraphunit.c (symbol_table::compile): Switch to text_section
7058 before calling assembly_start debug hook.
7059 * run-rtl-passes.c (run_rtl_passes): Likewise.
7062 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7064 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7065 range_int_cst_p rather than !symbolic_range_p before calling
7066 extract_range_from_multiplicative_op_1.
7068 2018-01-04 Jeff Law <law@redhat.com>
7070 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7071 redundant test in assertion.
7073 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7075 * doc/rtl.texi: Document machine_mode wrapper classes.
7077 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7079 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7082 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7084 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7085 the VEC_PERM_EXPR fold to fail.
7087 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7090 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7091 to switched_sections.
7093 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7096 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7099 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7102 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7103 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7105 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7108 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7109 is BLKmode and bitpos not zero or mode change is needed.
7111 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7114 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7117 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7120 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7121 instead of MULT rtx. Update all corresponding splitters.
7123 (*ssub<modesuffix>): Ditto.
7125 (*cmp_sadd_di): Update split patterns.
7126 (*cmp_sadd_si): Ditto.
7127 (*cmp_sadd_sidi): Ditto.
7128 (*cmp_ssub_di): Ditto.
7129 (*cmp_ssub_si): Ditto.
7130 (*cmp_ssub_sidi): Ditto.
7131 * config/alpha/predicates.md (const23_operand): New predicate.
7132 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7133 Look for ASHIFT, not MULT inner operand.
7134 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7136 2018-01-04 Martin Liska <mliska@suse.cz>
7138 PR gcov-profile/83669
7139 * gcov.c (output_intermediate_file): Add version to intermediate
7141 * doc/gcov.texi: Document new field 'version' in intermediate
7142 file format. Fix location of '-k' option of gcov command.
7144 2018-01-04 Martin Liska <mliska@suse.cz>
7147 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7149 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7151 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7153 2018-01-03 Martin Sebor <msebor@redhat.com>
7155 PR tree-optimization/83655
7156 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7157 checking calls with invalid arguments.
7159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7161 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7162 (vectorizable_mask_load_store): Delete.
7163 (vectorizable_call): Return false for masked loads and stores.
7164 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7165 instead of gimple_assign_rhs1.
7166 (vectorizable_load): Handle IFN_MASK_LOAD.
7167 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7171 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7173 (vectorizable_mask_load_store): ...here.
7174 (vectorizable_load): ...and here.
7176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7178 * tree-vect-stmts.c (vect_build_all_ones_mask)
7179 (vect_build_zero_merge_argument): New functions, split out from...
7180 (vectorizable_load): ...here.
7182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7184 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7186 (vectorizable_mask_load_store): ...here.
7187 (vectorizable_store): ...and here.
7189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7191 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7193 (vectorizable_mask_load_store): ...here.
7195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7197 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7198 (vect_model_store_cost): Take a vec_load_store_type instead of a
7200 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7201 (vect_model_store_cost): Take a vec_load_store_type instead of a
7203 (vectorizable_mask_load_store): Update accordingly.
7204 (vectorizable_store): Likewise.
7205 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7209 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7210 IFN_MASK_LOAD calls here rather than...
7211 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7214 Alan Hayward <alan.hayward@arm.com>
7215 David Sherwood <david.sherwood@arm.com>
7217 * expmed.c (extract_bit_field_1): For vector extracts,
7218 fall back to extract_bit_field_as_subreg if vec_extract
7221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7222 Alan Hayward <alan.hayward@arm.com>
7223 David Sherwood <david.sherwood@arm.com>
7225 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7226 they are variable or constant sized.
7227 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7228 slots for constant-sized data.
7230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7231 Alan Hayward <alan.hayward@arm.com>
7232 David Sherwood <david.sherwood@arm.com>
7234 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7235 handling COND_EXPRs with boolean comparisons, try to find a better
7236 basis for the mask type than the boolean itself.
7238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7240 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7241 is calculated and how it can be overridden.
7242 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7243 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7245 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7249 Alan Hayward <alan.hayward@arm.com>
7250 David Sherwood <david.sherwood@arm.com>
7252 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7253 Remove the mode argument.
7254 (aarch64_simd_valid_immediate): Remove the mode and inverse
7256 * config/aarch64/iterators.md (bitsize): New iterator.
7257 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7258 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7259 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7260 aarch64_simd_valid_immediate.
7261 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7262 (aarch64_reg_or_bic_imm): Likewise.
7263 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7264 with an insn_type enum and msl with a modifier_type enum.
7265 Replace element_width with a scalar_mode. Change the shift
7266 to unsigned int. Add constructors for scalar_float_mode and
7267 scalar_int_mode elements.
7268 (aarch64_vect_float_const_representable_p): Delete.
7269 (aarch64_can_const_movi_rtx_p)
7270 (aarch64_simd_scalar_immediate_valid_for_move)
7271 (aarch64_simd_make_constant): Update call to
7272 aarch64_simd_valid_immediate.
7273 (aarch64_advsimd_valid_immediate_hs): New function.
7274 (aarch64_advsimd_valid_immediate): Likewise.
7275 (aarch64_simd_valid_immediate): Remove mode and inverse
7276 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7277 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7278 and aarch64_float_const_representable_p on the result.
7279 (aarch64_output_simd_mov_immediate): Remove mode argument.
7280 Update call to aarch64_simd_valid_immediate and use of
7281 simd_immediate_info.
7282 (aarch64_output_scalar_simd_mov_immediate): Update call
7285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7286 Alan Hayward <alan.hayward@arm.com>
7287 David Sherwood <david.sherwood@arm.com>
7289 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7290 (mode_nunits): Likewise CONST_MODE_NUNITS.
7291 * machmode.def (ADJUST_NUNITS): Document.
7292 * genmodes.c (mode_data::need_nunits_adj): New field.
7293 (blank_mode): Update accordingly.
7294 (adj_nunits): New variable.
7295 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7297 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7298 listed in adj_nunits.
7299 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7300 listed in adj_nunits. Don't emit case statements for such modes.
7301 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7302 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7303 nothing if adj_nunits is nonnull.
7304 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7305 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7306 (emit_mode_fbit): Update use of print_maybe_const_decl.
7307 (emit_move_size): Likewise. Treat the array as non-const
7309 (emit_mode_adjustments): Handle adj_nunits.
7311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7313 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7314 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7315 (VECTOR_MODES): Use it.
7316 (make_vector_modes): Take the prefix as an argument.
7318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7319 Alan Hayward <alan.hayward@arm.com>
7320 David Sherwood <david.sherwood@arm.com>
7322 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7323 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7324 for MODE_VECTOR_BOOL.
7325 * machmode.def (VECTOR_BOOL_MODE): Document.
7326 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7327 (make_vector_bool_mode): New function.
7328 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7330 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7331 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7333 * stor-layout.c (int_mode_for_mode): Likewise.
7334 * tree.c (build_vector_type_for_mode): Likewise.
7335 * varasm.c (output_constant_pool_2): Likewise.
7336 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7337 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7338 for MODE_VECTOR_BOOL.
7339 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7340 of mode class checks.
7341 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7342 instead of a list of mode class checks.
7343 (expand_vector_scalar_condition): Likewise.
7344 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7347 Alan Hayward <alan.hayward@arm.com>
7348 David Sherwood <david.sherwood@arm.com>
7350 * machmode.h (mode_size): Change from unsigned short to
7352 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7353 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7354 or if measurement_type is not polynomial.
7355 (fixed_size_mode::includes_p): Check for constant-sized modes.
7356 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7357 return a poly_uint16 rather than an unsigned short.
7358 (emit_mode_size): Change the type of mode_size from unsigned short
7359 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7360 (emit_mode_adjustments): Cope with polynomial vector sizes.
7361 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7363 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7365 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7366 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7367 * caller-save.c (setup_save_areas): Likewise.
7368 (replace_reg_with_saved_mem): Likewise.
7369 * calls.c (emit_library_call_value_1): Likewise.
7370 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7371 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7372 (gen_lowpart_for_combine): Likewise.
7373 * convert.c (convert_to_integer_1): Likewise.
7374 * cse.c (equiv_constant, cse_insn): Likewise.
7375 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7376 (cselib_subst_to_values): Likewise.
7377 * dce.c (word_dce_process_block): Likewise.
7378 * df-problems.c (df_word_lr_mark_ref): Likewise.
7379 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7380 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7381 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7382 (rtl_for_decl_location): Likewise.
7383 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7384 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7385 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7386 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7387 (expand_expr_real_1): Likewise.
7388 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7389 (pad_below): Likewise.
7390 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7391 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7392 * ira.c (get_subreg_tracking_sizes): Likewise.
7393 * ira-build.c (ira_create_allocno_objects): Likewise.
7394 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7395 (ira_sort_regnos_for_alter_reg): Likewise.
7396 * ira-costs.c (record_operand_costs): Likewise.
7397 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7398 (resolve_simple_move): Likewise.
7399 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7400 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7401 (lra_constraints): Likewise.
7402 (CONST_POOL_OK_P): Reject variable-sized modes.
7403 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7404 (add_pseudo_to_slot, lra_spill): Likewise.
7405 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7406 * optabs-query.c (get_best_extraction_insn): Likewise.
7407 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7408 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7409 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7410 * recog.c (offsettable_address_addr_space_p): Likewise.
7411 * regcprop.c (maybe_mode_change): Likewise.
7412 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7413 * regrename.c (build_def_use): Likewise.
7414 * regstat.c (dump_reg_info): Likewise.
7415 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7416 (find_reloads, find_reloads_subreg_address): Likewise.
7417 * reload1.c (eliminate_regs_1): Likewise.
7418 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7419 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7420 (simplify_binary_operation_1, simplify_subreg): Likewise.
7421 * targhooks.c (default_function_arg_padding): Likewise.
7422 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7423 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7424 (verify_gimple_assign_ternary): Likewise.
7425 * tree-inline.c (estimate_move_cost): Likewise.
7426 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7427 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7428 (get_address_cost_ainc): Likewise.
7429 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7430 (vect_supportable_dr_alignment): Likewise.
7431 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7432 (vectorizable_reduction): Likewise.
7433 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7434 (vectorizable_operation, vectorizable_load): Likewise.
7435 * tree.c (build_same_sized_truth_vector_type): Likewise.
7436 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7437 * var-tracking.c (emit_note_insn_var_location): Likewise.
7438 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7439 (ADDR_VEC_ALIGN): Likewise.
7441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7442 Alan Hayward <alan.hayward@arm.com>
7443 David Sherwood <david.sherwood@arm.com>
7445 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7447 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7448 or if measurement_type is polynomial.
7449 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7450 * combine.c (make_extraction): Likewise.
7451 * dse.c (find_shift_sequence): Likewise.
7452 * dwarf2out.c (mem_loc_descriptor): Likewise.
7453 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7454 (extract_bit_field, extract_low_bits): Likewise.
7455 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7456 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7457 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7458 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7459 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7460 * reload.c (find_reloads): Likewise.
7461 * reload1.c (alter_reg): Likewise.
7462 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7463 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7464 * tree-if-conv.c (predicate_mem_writes): Likewise.
7465 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7466 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7467 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7468 * valtrack.c (dead_debug_insert_temp): Likewise.
7469 * varasm.c (mergeable_constant_section): Likewise.
7470 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7473 Alan Hayward <alan.hayward@arm.com>
7474 David Sherwood <david.sherwood@arm.com>
7476 * expr.c (expand_assignment): Cope with polynomial mode sizes
7477 when assigning to a CONCAT.
7479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7480 Alan Hayward <alan.hayward@arm.com>
7481 David Sherwood <david.sherwood@arm.com>
7483 * machmode.h (mode_precision): Change from unsigned short to
7485 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7487 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7488 or if measurement_type is not polynomial.
7489 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7490 in which the mode is already known to be a scalar_int_mode.
7491 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7492 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7494 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7495 for GET_MODE_PRECISION.
7496 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7497 for GET_MODE_PRECISION.
7498 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7500 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7501 (expand_field_assignment, make_extraction): Likewise.
7502 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7503 (get_last_value): Likewise.
7504 * convert.c (convert_to_integer_1): Likewise.
7505 * cse.c (cse_insn): Likewise.
7506 * expr.c (expand_expr_real_1): Likewise.
7507 * lra-constraints.c (simplify_operand_subreg): Likewise.
7508 * optabs-query.c (can_atomic_load_p): Likewise.
7509 * optabs.c (expand_atomic_load): Likewise.
7510 (expand_atomic_store): Likewise.
7511 * ree.c (combine_reaching_defs): Likewise.
7512 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7513 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7514 * tree.h (type_has_mode_precision_p): Likewise.
7515 * ubsan.c (instrument_si_overflow): Likewise.
7517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7518 Alan Hayward <alan.hayward@arm.com>
7519 David Sherwood <david.sherwood@arm.com>
7521 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7522 polynomial numbers of units.
7523 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7524 (valid_vector_subparts_p): New function.
7525 (build_vector_type): Remove temporary shim and take the number
7526 of units as a poly_uint64 rather than an int.
7527 (build_opaque_vector_type): Take the number of units as a
7528 poly_uint64 rather than an int.
7529 * tree.c (build_vector_from_ctor): Handle polynomial
7530 TYPE_VECTOR_SUBPARTS.
7531 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7532 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7533 (build_vector_from_val): If the number of units is variable,
7534 use build_vec_duplicate_cst for constant operands and
7535 VEC_DUPLICATE_EXPR otherwise.
7536 (make_vector_type): Remove temporary is_constant ().
7537 (build_vector_type, build_opaque_vector_type): Take the number of
7538 units as a poly_uint64 rather than an int.
7539 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7541 * cfgexpand.c (expand_debug_expr): Likewise.
7542 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7543 (store_constructor, expand_expr_real_1): Likewise.
7544 (const_scalar_mask_from_tree): Likewise.
7545 * fold-const-call.c (fold_const_reduction): Likewise.
7546 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7547 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7548 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7549 (fold_relational_const): Likewise.
7550 (native_interpret_vector): Likewise. Change the size from an
7551 int to an unsigned int.
7552 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7553 TYPE_VECTOR_SUBPARTS.
7554 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7555 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7556 duplicating a non-constant operand into a variable-length vector.
7557 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7558 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7559 * ipa-icf.c (sem_variable::equals): Likewise.
7560 * match.pd: Likewise.
7561 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7562 * print-tree.c (print_node): Likewise.
7563 * stor-layout.c (layout_type): Likewise.
7564 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7565 * tree-cfg.c (verify_gimple_comparison): Likewise.
7566 (verify_gimple_assign_binary): Likewise.
7567 (verify_gimple_assign_ternary): Likewise.
7568 (verify_gimple_assign_single): Likewise.
7569 * tree-pretty-print.c (dump_generic_node): Likewise.
7570 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7571 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7572 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7573 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7574 (vect_shift_permute_load_chain): Likewise.
7575 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7576 (expand_vector_condition, optimize_vector_constructor): Likewise.
7577 (lower_vec_perm, get_compute_type): Likewise.
7578 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7579 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7580 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7581 (vect_recog_mask_conversion_pattern): Likewise.
7582 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7583 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7584 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7585 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7586 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7587 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7588 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7589 (supportable_widening_operation): Likewise.
7590 (supportable_narrowing_operation): Likewise.
7591 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7593 * varasm.c (output_constant): Likewise.
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 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7600 so that both the length == 3 and length != 3 cases set up their
7601 own permute vectors. Add comments explaining why we know the
7602 number of elements is constant.
7603 (vect_permute_load_chain): Likewise.
7605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7606 Alan Hayward <alan.hayward@arm.com>
7607 David Sherwood <david.sherwood@arm.com>
7609 * machmode.h (mode_nunits): Change from unsigned char to
7611 (ONLY_FIXED_SIZE_MODES): New macro.
7612 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7613 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7614 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7616 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7617 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7618 or if measurement_type is not polynomial.
7619 * genmodes.c (ZERO_COEFFS): New macro.
7620 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7622 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7623 Use ZERO_COEFFS when emitting initializers.
7624 * data-streamer.h (bp_pack_poly_value): New function.
7625 (bp_unpack_poly_value): Likewise.
7626 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7627 for GET_MODE_NUNITS.
7628 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7629 for GET_MODE_NUNITS.
7630 * tree.c (make_vector_type): Remove temporary shim and make
7631 the real function take the number of units as a poly_uint64
7633 (build_vector_type_for_mode): Handle polynomial nunits.
7634 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7635 * emit-rtl.c (const_vec_series_p_1): Likewise.
7636 (gen_rtx_CONST_VECTOR): Likewise.
7637 * fold-const.c (test_vec_duplicate_folding): Likewise.
7638 * genrecog.c (validate_pattern): Likewise.
7639 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7640 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7641 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7642 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7643 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7644 * rtlanal.c (subreg_get_info): Likewise.
7645 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7646 (vect_grouped_load_supported): Likewise.
7647 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7648 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7649 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7650 (simplify_const_unary_operation, simplify_binary_operation_1)
7651 (simplify_const_binary_operation, simplify_ternary_operation)
7652 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7653 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7654 instead of CONST_VECTOR_NUNITS.
7655 * varasm.c (output_constant_pool_2): Likewise.
7656 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7657 explicit-encoded elements in the XVEC for variable-length vectors.
7659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7661 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7664 Alan Hayward <alan.hayward@arm.com>
7665 David Sherwood <david.sherwood@arm.com>
7667 * coretypes.h (fixed_size_mode): Declare.
7668 (fixed_size_mode_pod): New typedef.
7669 * builtins.h (target_builtins::x_apply_args_mode)
7670 (target_builtins::x_apply_result_mode): Change type to
7671 fixed_size_mode_pod.
7672 * builtins.c (apply_args_size, apply_result_size, result_vector)
7673 (expand_builtin_apply_args_1, expand_builtin_apply)
7674 (expand_builtin_return): Update accordingly.
7676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7678 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7679 * cselib.c (cselib_hash_rtx): Likewise.
7680 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7681 CONST_VECTOR encoding.
7683 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7684 Jeff Law <law@redhat.com>
7687 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7688 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7689 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7690 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7693 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7694 explicitly probe *sp in a noreturn function if there were any callee
7695 register saves or frame pointer is needed.
7697 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7700 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7701 BLKmode for ternary, binary or unary expressions.
7704 * var-tracking.c (delete_vta_debug_insn): New inline function.
7705 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7706 insns from get_insns () to NULL instead of each bb separately.
7707 Use delete_vta_debug_insn. No longer static.
7708 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7709 delete_vta_debug_insns callers.
7710 * rtl.h (delete_vta_debug_insns): Declare.
7711 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7712 instead of variable_tracking_main.
7714 2018-01-03 Martin Sebor <msebor@redhat.com>
7716 PR tree-optimization/83603
7717 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7718 arguments past the endof the argument list in functions declared
7719 without a prototype.
7720 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7721 Avoid checking when arguments are null.
7723 2018-01-03 Martin Sebor <msebor@redhat.com>
7726 * doc/extend.texi (attribute const): Fix a typo.
7727 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7728 issuing -Wsuggest-attribute for void functions.
7730 2018-01-03 Martin Sebor <msebor@redhat.com>
7732 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7733 offset_int::from instead of wide_int::to_shwi.
7734 (maybe_diag_overlap): Remove assertion.
7735 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7736 * gimple-ssa-sprintf.c (format_directive): Same.
7737 (parse_directive): Same.
7738 (sprintf_dom_walker::compute_format_length): Same.
7739 (try_substitute_return_value): Same.
7741 2018-01-03 Jeff Law <law@redhat.com>
7744 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7745 non-constant residual for zero at runtime and avoid probing in
7746 that case. Reorganize code for trailing problem to mirror handling
7749 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7751 PR tree-optimization/83501
7752 * tree-ssa-strlen.c (get_string_cst): New.
7753 (handle_char_store): Call get_string_cst.
7755 2018-01-03 Martin Liska <mliska@suse.cz>
7757 PR tree-optimization/83593
7758 * tree-ssa-strlen.c: Include tree-cfg.h.
7759 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7760 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7761 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7763 (strlen_dom_walker::before_dom_children): Call
7764 gimple_purge_dead_eh_edges. Dump tranformation with details
7766 (strlen_dom_walker::before_dom_children): Update call by adding
7767 new argument cleanup_eh.
7768 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7770 2018-01-03 Martin Liska <mliska@suse.cz>
7773 * cif-code.def (VARIADIC_THUNK): New enum value.
7774 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7777 2018-01-03 Jan Beulich <jbeulich@suse.com>
7779 * sse.md (mov<mode>_internal): Tighten condition for when to use
7780 vmovdqu<ssescalarsize> for TI and OI modes.
7782 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7784 Update copyright years.
7786 2018-01-03 Martin Liska <mliska@suse.cz>
7789 * ipa-visibility.c (function_and_variable_visibility): Skip
7790 functions with noipa attribure.
7792 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7794 * gcc.c (process_command): Update copyright notice dates.
7795 * gcov-dump.c (print_version): Ditto.
7796 * gcov.c (print_version): Ditto.
7797 * gcov-tool.c (print_version): Ditto.
7798 * gengtype.c (create_file): Ditto.
7799 * doc/cpp.texi: Bump @copying's copyright year.
7800 * doc/cppinternals.texi: Ditto.
7801 * doc/gcc.texi: Ditto.
7802 * doc/gccint.texi: Ditto.
7803 * doc/gcov.texi: Ditto.
7804 * doc/install.texi: Ditto.
7805 * doc/invoke.texi: Ditto.
7807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7809 * vector-builder.h (vector_builder::m_full_nelts): Change from
7810 unsigned int to poly_uint64.
7811 (vector_builder::full_nelts): Update prototype accordingly.
7812 (vector_builder::new_vector): Likewise.
7813 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7814 (vector_builder::operator ==): Likewise.
7815 (vector_builder::finalize): Likewise.
7816 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7817 Take the number of elements as a poly_uint64 rather than an
7819 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7820 from unsigned int to poly_uint64.
7821 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7822 (vec_perm_indices::new_vector): Likewise.
7823 (vec_perm_indices::length): Likewise.
7824 (vec_perm_indices::nelts_per_input): Likewise.
7825 (vec_perm_indices::input_nelts): Likewise.
7826 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7827 number of elements per input as a poly_uint64 rather than an
7828 unsigned int. Use the original encoding for variable-length
7829 vectors, rather than clamping each individual element.
7830 For the second and subsequent elements in each pattern,
7831 clamp the step and base before clamping their sum.
7832 (vec_perm_indices::series_p): Handle polynomial element counts.
7833 (vec_perm_indices::all_in_range_p): Likewise.
7834 (vec_perm_indices_to_tree): Likewise.
7835 (vec_perm_indices_to_rtx): Likewise.
7836 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7837 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7838 (tree_vector_builder::new_binary_operation): Handle polynomial
7839 element counts. Return false if we need to know the number
7840 of elements at compile time.
7841 * fold-const.c (fold_vec_perm): Punt if the number of elements
7842 isn't known at compile time.
7844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7846 * vec-perm-indices.h (vec_perm_builder): Change element type
7847 from HOST_WIDE_INT to poly_int64.
7848 (vec_perm_indices::element_type): Update accordingly.
7849 (vec_perm_indices::clamp): Handle polynomial element_types.
7850 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7851 (vec_perm_indices::all_in_range_p): Likewise.
7852 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7854 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7855 polynomial vec_perm_indices element types.
7856 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7857 * fold-const.c (fold_vec_perm): Likewise.
7858 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7859 * tree-vect-generic.c (lower_vec_perm): Likewise.
7860 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7861 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7862 element type to HOST_WIDE_INT.
7864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7865 Alan Hayward <alan.hayward@arm.com>
7866 David Sherwood <david.sherwood@arm.com>
7868 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7869 rather than an int. Use plus_constant.
7870 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7871 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7874 Alan Hayward <alan.hayward@arm.com>
7875 David Sherwood <david.sherwood@arm.com>
7877 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7878 a HOST_WIDE_INT to a poly_int64.
7880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7881 Alan Hayward <alan.hayward@arm.com>
7882 David Sherwood <david.sherwood@arm.com>
7884 * calls.c (load_register_parameters): Cope with polynomial
7885 mode sizes. Require a constant size for BLKmode parameters
7886 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7887 forces a parameter to be padded at the lsb end in order to
7888 fill a complete number of words, require the parameter size
7889 to be ordered wrt UNITS_PER_WORD.
7891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7892 Alan Hayward <alan.hayward@arm.com>
7893 David Sherwood <david.sherwood@arm.com>
7895 * reload1.c (spill_stack_slot_width): Change element type
7896 from unsigned int to poly_uint64_pod.
7897 (alter_reg): Treat mode sizes as polynomial.
7899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7900 Alan Hayward <alan.hayward@arm.com>
7901 David Sherwood <david.sherwood@arm.com>
7903 * reload.c (complex_word_subreg_p): New function.
7904 (reload_inner_reg_of_subreg, push_reload): Use it.
7906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7907 Alan Hayward <alan.hayward@arm.com>
7908 David Sherwood <david.sherwood@arm.com>
7910 * lra-constraints.c (process_alt_operands): Reject matched
7911 operands whose sizes aren't ordered.
7912 (match_reload): Refer to this check here.
7914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7915 Alan Hayward <alan.hayward@arm.com>
7916 David Sherwood <david.sherwood@arm.com>
7918 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7919 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7926 Use plus_constant instead of gen_rtx_PLUS.
7928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7929 Alan Hayward <alan.hayward@arm.com>
7930 David Sherwood <david.sherwood@arm.com>
7932 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7933 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7934 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7935 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7936 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7937 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7938 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7939 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7940 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7941 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7943 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7944 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7945 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7946 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7947 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7948 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7949 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7950 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7951 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7952 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7954 * expr.c (emit_move_resolve_push): Treat the input and result
7955 of PUSH_ROUNDING as a poly_int64.
7956 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7957 (emit_push_insn): Likewise.
7958 * lra-eliminations.c (mark_not_eliminable): Likewise.
7959 * recog.c (push_operand): Likewise.
7960 * reload1.c (elimination_effects): Likewise.
7961 * rtlanal.c (nonzero_bits1): Likewise.
7962 * calls.c (store_one_arg): Likewise. Require the padding to be
7963 known at compile time.
7965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7966 Alan Hayward <alan.hayward@arm.com>
7967 David Sherwood <david.sherwood@arm.com>
7969 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7970 Use plus_constant instead of gen_rtx_PLUS.
7972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7973 Alan Hayward <alan.hayward@arm.com>
7974 David Sherwood <david.sherwood@arm.com>
7976 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7980 Alan Hayward <alan.hayward@arm.com>
7981 David Sherwood <david.sherwood@arm.com>
7983 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7984 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7985 via stack temporaries. Treat the mode size as polynomial too.
7987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7988 Alan Hayward <alan.hayward@arm.com>
7989 David Sherwood <david.sherwood@arm.com>
7991 * expr.c (expand_expr_real_2): When handling conversions involving
7992 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7993 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7994 as a poly_uint64 too.
7996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7997 Alan Hayward <alan.hayward@arm.com>
7998 David Sherwood <david.sherwood@arm.com>
8000 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8003 Alan Hayward <alan.hayward@arm.com>
8004 David Sherwood <david.sherwood@arm.com>
8006 * combine.c (can_change_dest_mode): Handle polynomial
8007 REGMODE_NATURAL_SIZE.
8008 * expmed.c (store_bit_field_1): Likewise.
8009 * expr.c (store_constructor): Likewise.
8010 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8011 and polynomial REGMODE_NATURAL_SIZE.
8012 (gen_lowpart_common): Likewise.
8013 * reginfo.c (record_subregs_of_mode): Likewise.
8014 * rtlanal.c (read_modify_subreg_p): Likewise.
8016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8017 Alan Hayward <alan.hayward@arm.com>
8018 David Sherwood <david.sherwood@arm.com>
8020 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8021 numbers of elements.
8023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8024 Alan Hayward <alan.hayward@arm.com>
8025 David Sherwood <david.sherwood@arm.com>
8027 * match.pd: Cope with polynomial numbers of vector elements.
8029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8030 Alan Hayward <alan.hayward@arm.com>
8031 David Sherwood <david.sherwood@arm.com>
8033 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8034 in a POINTER_PLUS_EXPR.
8036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8037 Alan Hayward <alan.hayward@arm.com>
8038 David Sherwood <david.sherwood@arm.com>
8040 * omp-simd-clone.c (simd_clone_subparts): New function.
8041 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8042 (ipa_simd_modify_function_body): Likewise.
8044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8045 Alan Hayward <alan.hayward@arm.com>
8046 David Sherwood <david.sherwood@arm.com>
8048 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8049 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8050 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8051 (expand_vector_condition, vector_element): Likewise.
8052 (subparts_gt): New function.
8053 (get_compute_type): Use subparts_gt.
8054 (count_type_subparts): Delete.
8055 (expand_vector_operations_1): Use subparts_gt instead of
8056 count_type_subparts.
8058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8059 Alan Hayward <alan.hayward@arm.com>
8060 David Sherwood <david.sherwood@arm.com>
8062 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8063 (vect_compile_time_alias): ...this new function. Do the calculation
8064 on poly_ints rather than trees.
8065 (vect_prune_runtime_alias_test_list): Update call accordingly.
8067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8068 Alan Hayward <alan.hayward@arm.com>
8069 David Sherwood <david.sherwood@arm.com>
8071 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8073 (vect_schedule_slp_instance): Likewise.
8075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8076 Alan Hayward <alan.hayward@arm.com>
8077 David Sherwood <david.sherwood@arm.com>
8079 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8080 constant and extern definitions for variable-length vectors.
8081 (vect_get_constant_vectors): Note that the number of units
8082 is known to be constant.
8084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8085 Alan Hayward <alan.hayward@arm.com>
8086 David Sherwood <david.sherwood@arm.com>
8088 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8089 of units as polynomial. Choose between WIDE and NARROW based
8092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8093 Alan Hayward <alan.hayward@arm.com>
8094 David Sherwood <david.sherwood@arm.com>
8096 * tree-vect-stmts.c (simd_clone_subparts): New function.
8097 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8100 Alan Hayward <alan.hayward@arm.com>
8101 David Sherwood <david.sherwood@arm.com>
8103 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8104 vectors as polynomial. Use build_index_vector for
8107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8108 Alan Hayward <alan.hayward@arm.com>
8109 David Sherwood <david.sherwood@arm.com>
8111 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8112 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8113 for variable-length vectors.
8114 (vectorizable_mask_load_store): Treat the number of units as
8115 polynomial, asserting that it is constant if the condition has
8116 already been enforced.
8117 (vectorizable_store, vectorizable_load): Likewise.
8119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8120 Alan Hayward <alan.hayward@arm.com>
8121 David Sherwood <david.sherwood@arm.com>
8123 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8124 of units as polynomial. Punt if we can't tell at compile time
8125 which vector contains the final result.
8127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8128 Alan Hayward <alan.hayward@arm.com>
8129 David Sherwood <david.sherwood@arm.com>
8131 * tree-vect-loop.c (vectorizable_induction): Treat the number
8132 of units as polynomial. Punt on SLP inductions. Use an integer
8133 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8134 cast of such a series for variable-length floating-point
8137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8138 Alan Hayward <alan.hayward@arm.com>
8139 David Sherwood <david.sherwood@arm.com>
8141 * tree.h (build_index_vector): Declare.
8142 * tree.c (build_index_vector): New function.
8143 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8144 of units as polynomial, forcibly converting it to a constant if
8145 vectorizable_reduction has already enforced the condition.
8146 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8147 to create a {1,2,3,...} vector.
8148 (vectorizable_reduction): Treat the number of units as polynomial.
8149 Choose vectype_in based on the largest scalar element size rather
8150 than the smallest number of units. Enforce the restrictions
8153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8154 Alan Hayward <alan.hayward@arm.com>
8155 David Sherwood <david.sherwood@arm.com>
8157 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8158 number of units as polynomial.
8160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8161 Alan Hayward <alan.hayward@arm.com>
8162 David Sherwood <david.sherwood@arm.com>
8164 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8165 * target.def (autovectorize_vector_sizes): Return the vector sizes
8166 by pointer, using vector_sizes rather than a bitmask.
8167 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8168 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8169 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8171 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8172 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8173 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8174 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8175 * omp-general.c (omp_max_vf): Likewise.
8176 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8177 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8178 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8179 * tree-vect-slp.c (vect_slp_bb): Likewise.
8180 * doc/tm.texi: Regenerate.
8181 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8183 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8184 the vector size as a poly_uint64 rather than an unsigned int.
8185 (current_vector_size): Change from an unsigned int to a poly_uint64.
8186 (get_vectype_for_scalar_type): Update accordingly.
8187 * tree.h (build_truth_vector_type): Take the size and number of
8188 units as a poly_uint64 rather than an unsigned int.
8189 (build_vector_type): Add a temporary overload that takes
8190 the number of units as a poly_uint64 rather than an unsigned int.
8191 * tree.c (make_vector_type): Likewise.
8192 (build_truth_vector_type): Take the number of units as a poly_uint64
8193 rather than an unsigned int.
8195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8196 Alan Hayward <alan.hayward@arm.com>
8197 David Sherwood <david.sherwood@arm.com>
8199 * target.def (get_mask_mode): Take the number of units and length
8200 as poly_uint64s rather than unsigned ints.
8201 * targhooks.h (default_get_mask_mode): Update accordingly.
8202 * targhooks.c (default_get_mask_mode): Likewise.
8203 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8204 * doc/tm.texi: Regenerate.
8206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8207 Alan Hayward <alan.hayward@arm.com>
8208 David Sherwood <david.sherwood@arm.com>
8210 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8211 * omp-general.c (omp_max_vf): Likewise.
8212 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8213 (expand_omp_simd): Handle polynomial safelen.
8214 * omp-low.c (omplow_simd_context): Add a default constructor.
8215 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8216 (lower_rec_simd_input_clauses): Update accordingly.
8217 (lower_rec_input_clauses): Likewise.
8219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8220 Alan Hayward <alan.hayward@arm.com>
8221 David Sherwood <david.sherwood@arm.com>
8223 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8224 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8225 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8226 (vect_analyze_slp_cost): Likewise.
8227 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8228 (vect_model_load_cost): Likewise.
8230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8231 Alan Hayward <alan.hayward@arm.com>
8232 David Sherwood <david.sherwood@arm.com>
8234 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8235 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8236 from an unsigned int * to a poly_uint64_pod *.
8237 (calculate_unrolling_factor): New function.
8238 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8241 Alan Hayward <alan.hayward@arm.com>
8242 David Sherwood <david.sherwood@arm.com>
8244 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8245 from an unsigned int to a poly_uint64.
8246 (_loop_vec_info::slp_unrolling_factor): Likewise.
8247 (_loop_vec_info::vectorization_factor): Change from an int
8249 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8250 (vect_get_num_vectors): New function.
8251 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8252 (vect_get_num_copies): Use vect_get_num_vectors.
8253 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8254 to an unsigned int *.
8255 (vect_analyze_data_refs): Change min_vf from an int * to a
8257 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8258 than an unsigned HOST_WIDE_INT.
8259 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8260 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8261 to an unsigned int *.
8262 (vect_analyze_data_ref_dependences): Likewise.
8263 (vect_compute_data_ref_alignment): Handle polynomial vf.
8264 (vect_enhance_data_refs_alignment): Likewise.
8265 (vect_prune_runtime_alias_test_list): Likewise.
8266 (vect_shift_permute_load_chain): Likewise.
8267 (vect_supportable_dr_alignment): Likewise.
8268 (dependence_distance_ge_vf): Take the vectorization factor as a
8269 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8270 (vect_analyze_data_refs): Change min_vf from an int * to a
8272 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8273 vfm1 as a poly_uint64 rather than an int. Make the same change
8274 for the returned bound_scalar.
8275 (vect_gen_vector_loop_niters): Handle polynomial vf.
8276 (vect_do_peeling): Likewise. Update call to
8277 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8278 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8280 * tree-vect-loop.c (vect_determine_vectorization_factor)
8281 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8282 (vect_get_known_peeling_cost): Likewise.
8283 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8284 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8285 (vect_transform_loop): Likewise. Use the lowest possible VF when
8286 updating the upper bounds of the loop.
8287 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8289 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8290 polynomial unroll factors.
8291 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8292 (vect_make_slp_decision): Likewise.
8293 (vect_supported_load_permutation_p): Likewise, and polynomial
8295 (vect_analyze_slp_cost): Handle polynomial vf.
8296 (vect_slp_analyze_node_operations): Likewise.
8297 (vect_slp_analyze_bb_1): Likewise.
8298 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8299 than an unsigned HOST_WIDE_INT.
8300 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8301 (vectorizable_load): Handle polynomial vf.
8302 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8304 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8307 Alan Hayward <alan.hayward@arm.com>
8308 David Sherwood <david.sherwood@arm.com>
8310 * match.pd: Handle bit operations involving three constants
8311 and try to fold one pair.
8313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8315 * tree-vect-loop-manip.c: Include gimple-fold.h.
8316 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8317 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8318 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8319 Add a path that uses a step of VF instead of 1, but disable it
8321 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8322 and niters_no_overflow parameters. Update calls to
8323 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8324 Create a new SSA name if the latter choses to use a ste other
8325 than zero, and return it via niters_vector_mult_vf_var.
8326 * tree-vect-loop.c (vect_transform_loop): Update calls to
8327 vect_do_peeling, vect_gen_vector_loop_niters and
8328 slpeel_make_loop_iterate_ntimes.
8329 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8330 (vect_gen_vector_loop_niters): Update declarations after above changes.
8332 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8334 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8335 128-bit round to integer instructions.
8336 (ceil<mode>2): Likewise.
8337 (btrunc<mode>2): Likewise.
8338 (round<mode>2): Likewise.
8340 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8342 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8343 unaligned VSX load/store on P8/P9.
8344 (expand_block_clear): Allow the use of unaligned VSX
8345 load/store on P8/P9.
8347 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8349 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8351 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8352 swap associated with both a load and a store.
8354 2018-01-02 Andrew Waterman <andrew@sifive.com>
8356 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8357 * config/riscv/riscv.md (clear_cache): Use it.
8359 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8361 * web.c: Remove out-of-date comment.
8363 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8365 * expr.c (fixup_args_size_notes): Check that any existing
8366 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8367 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8368 (emit_single_push_insn): ...here.
8370 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8372 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8373 (const_vector_encoded_nelts): New function.
8374 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8375 (const_vector_int_elt, const_vector_elt): Declare.
8376 * emit-rtl.c (const_vector_int_elt_1): New function.
8377 (const_vector_elt): Likewise.
8378 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8379 of CONST_VECTOR_ELT.
8381 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8383 * expr.c: Include rtx-vector-builder.h.
8384 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8385 directly on the tree encoding.
8386 (const_vector_from_tree): Likewise.
8387 * optabs.c: Include rtx-vector-builder.h.
8388 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8389 sequence of "u" values.
8390 * vec-perm-indices.c: Include rtx-vector-builder.h.
8391 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8392 directly on the vec_perm_indices encoding.
8394 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8396 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8397 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8398 * rtx-vector-builder.h: New file.
8399 * rtx-vector-builder.c: Likewise.
8400 * rtl.h (rtx_def::u2): Add a const_vector field.
8401 (CONST_VECTOR_NPATTERNS): New macro.
8402 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8403 (CONST_VECTOR_DUPLICATE_P): Likewise.
8404 (CONST_VECTOR_STEPPED_P): Likewise.
8405 (CONST_VECTOR_ENCODED_ELT): Likewise.
8406 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8407 (unwrap_const_vec_duplicate): Likewise.
8408 (const_vec_series_p): Check for a non-duplicated vector encoding.
8409 Say that the function only returns true for integer vectors.
8410 * emit-rtl.c: Include rtx-vector-builder.h.
8411 (gen_const_vec_duplicate_1): Delete.
8412 (gen_const_vector): Call gen_const_vec_duplicate instead of
8413 gen_const_vec_duplicate_1.
8414 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8415 (gen_const_vec_duplicate): Use rtx_vector_builder.
8416 (gen_const_vec_series): Likewise.
8417 (gen_rtx_CONST_VECTOR): Likewise.
8418 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8419 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8420 Build a new vector rather than modifying a CONST_VECTOR in-place.
8421 (handle_special_swappables): Update call accordingly.
8422 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8423 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8424 Build a new vector rather than modifying a CONST_VECTOR in-place.
8425 (handle_special_swappables): Update call accordingly.
8427 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8429 * simplify-rtx.c (simplify_const_binary_operation): Use
8430 CONST_VECTOR_ELT instead of XVECEXP.
8432 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8434 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8435 the selector elements to be different from the data elements
8436 if the selector is a VECTOR_CST.
8437 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8438 ssizetype for the selector.
8440 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8442 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8443 before testing each element individually.
8444 * tree-vect-generic.c (lower_vec_perm): Likewise.
8446 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8448 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8449 * selftest-run-tests.c (selftest::run_tests): Call it.
8450 * vector-builder.h (vector_builder::operator ==): New function.
8451 (vector_builder::operator !=): Likewise.
8452 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8453 (vec_perm_indices::all_from_input_p): New function.
8454 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8455 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8456 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8457 instead of reading the VECTOR_CST directly. Detect whether both
8458 vector inputs are the same before constructing the vec_perm_indices,
8459 and update the number of inputs argument accordingly. Use the
8460 utility functions added above. Only construct sel2 if we need to.
8462 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8464 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8465 the broadcast of the low byte.
8466 (expand_mult_highpart): Use an explicit encoding for the permutes.
8467 * optabs-query.c (can_mult_highpart_p): Likewise.
8468 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8469 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8470 (vectorizable_bswap): Likewise.
8471 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8472 explicit encoding for the power-of-2 permutes.
8473 (vect_permute_store_chain): Likewise.
8474 (vect_grouped_load_supported): Likewise.
8475 (vect_permute_load_chain): Likewise.
8477 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8479 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8480 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8481 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8482 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8483 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8484 (vect_gen_perm_mask_any): Likewise.
8486 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8488 * int-vector-builder.h: New file.
8489 * vec-perm-indices.h: Include int-vector-builder.h.
8490 (vec_perm_indices): Redefine as an int_vector_builder.
8491 (auto_vec_perm_indices): Delete.
8492 (vec_perm_builder): Redefine as a stand-alone class.
8493 (vec_perm_indices::vec_perm_indices): New function.
8494 (vec_perm_indices::clamp): Likewise.
8495 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8496 (vec_perm_indices::new_vector): New function.
8497 (vec_perm_indices::new_expanded_vector): Update for new
8498 vec_perm_indices class.
8499 (vec_perm_indices::rotate_inputs): New function.
8500 (vec_perm_indices::all_in_range_p): Operate directly on the
8501 encoded form, without computing elided elements.
8502 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8503 encoding. Update for new vec_perm_indices class.
8504 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8505 the given vec_perm_builder.
8506 (expand_vec_perm_var): Update vec_perm_builder constructor.
8507 (expand_mult_highpart): Use vec_perm_builder instead of
8508 auto_vec_perm_indices.
8509 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8510 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8511 or double series encoding as appropriate.
8512 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8513 vec_perm_indices instead of auto_vec_perm_indices.
8514 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8515 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8516 (vect_permute_store_chain): Likewise.
8517 (vect_grouped_load_supported): Likewise.
8518 (vect_permute_load_chain): Likewise.
8519 (vect_shift_permute_load_chain): Likewise.
8520 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8521 (vect_transform_slp_perm_load): Likewise.
8522 (vect_schedule_slp_instance): Likewise.
8523 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8524 (vectorizable_mask_load_store): Likewise.
8525 (vectorizable_bswap): Likewise.
8526 (vectorizable_store): Likewise.
8527 (vectorizable_load): Likewise.
8528 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8529 vec_perm_indices instead of auto_vec_perm_indices. Use
8530 tree_to_vec_perm_builder to read the vector from a tree.
8531 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8532 vec_perm_builder instead of a vec_perm_indices.
8533 (have_whole_vector_shift): Use vec_perm_builder and
8534 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8535 truncation to calc_vec_perm_mask_for_shift.
8536 (vect_create_epilog_for_reduction): Likewise.
8537 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8538 from auto_vec_perm_indices to vec_perm_indices.
8539 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8540 instead of changing individual elements.
8541 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8542 the vector in d.perm.
8543 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8544 from auto_vec_perm_indices to vec_perm_indices.
8545 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8546 instead of changing individual elements.
8547 (arm_vectorize_vec_perm_const): Use new_vector to install
8548 the vector in d.perm.
8549 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8550 Update vec_perm_builder constructor.
8551 (rs6000_expand_interleave): Likewise.
8552 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8553 (rs6000_expand_interleave): Likewise.
8555 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8557 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8558 to qimode could truncate the indices.
8559 * optabs.c (expand_vec_perm_var): Likewise.
8561 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8563 * Makefile.in (OBJS): Add vec-perm-indices.o.
8564 * vec-perm-indices.h: New file.
8565 * vec-perm-indices.c: Likewise.
8566 * target.h (vec_perm_indices): Replace with a forward class
8568 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8569 * optabs.h: Include vec-perm-indices.h.
8570 (expand_vec_perm): Delete.
8571 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8572 (expand_vec_perm_const): Declare.
8573 * target.def (vec_perm_const_ok): Replace with...
8574 (vec_perm_const): ...this new hook.
8575 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8576 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8577 * doc/tm.texi: Regenerate.
8578 * optabs.def (vec_perm_const): Delete.
8579 * doc/md.texi (vec_perm_const): Likewise.
8580 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8581 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8582 expand_vec_perm for constant permutation vectors. Assert that
8583 the mode of variable permutation vectors is the integer equivalent
8584 of the mode that is being permuted.
8585 * optabs-query.h (selector_fits_mode_p): Declare.
8586 * optabs-query.c: Include vec-perm-indices.h.
8587 (selector_fits_mode_p): New function.
8588 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8589 is defined, instead of checking whether the vec_perm_const_optab
8590 exists. Use targetm.vectorize.vec_perm_const instead of
8591 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8592 fit in the vector mode before using a variable permute.
8593 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8594 vec_perm_indices instead of an rtx.
8595 (expand_vec_perm): Replace with...
8596 (expand_vec_perm_const): ...this new function. Take the selector
8597 as a vec_perm_indices rather than an rtx. Also take the mode of
8598 the selector. Update call to shift_amt_for_vec_perm_mask.
8599 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8600 Use vec_perm_indices::new_expanded_vector to expand the original
8601 selector into bytes. Check whether the indices fit in the vector
8602 mode before using a variable permute.
8603 (expand_vec_perm_var): Make global.
8604 (expand_mult_highpart): Use expand_vec_perm_const.
8605 * fold-const.c: Includes vec-perm-indices.h.
8606 * tree-ssa-forwprop.c: Likewise.
8607 * tree-vect-data-refs.c: Likewise.
8608 * tree-vect-generic.c: Likewise.
8609 * tree-vect-loop.c: Likewise.
8610 * tree-vect-slp.c: Likewise.
8611 * tree-vect-stmts.c: Likewise.
8612 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8614 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8615 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8616 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8617 (aarch64_vectorize_vec_perm_const): ...this new function.
8618 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8619 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8620 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8621 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8622 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8623 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8624 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8626 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8627 check for NEON modes.
8628 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8629 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8630 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8631 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8633 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8634 the old VEC_PERM_CONST conditions.
8635 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8636 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8637 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8638 (ia64_vectorize_vec_perm_const_ok): Merge into...
8639 (ia64_vectorize_vec_perm_const): ...this new function.
8640 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8641 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8642 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8643 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8644 * config/mips/mips.c (mips_expand_vec_perm_const)
8645 (mips_vectorize_vec_perm_const_ok): Merge into...
8646 (mips_vectorize_vec_perm_const): ...this new function.
8647 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8648 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8649 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8650 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8651 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8652 (rs6000_expand_vec_perm_const): Delete.
8653 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8655 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8656 (altivec_expand_vec_perm_const_le): Take each operand individually.
8657 Operate on constant selectors rather than rtxes.
8658 (altivec_expand_vec_perm_const): Likewise. Update call to
8659 altivec_expand_vec_perm_const_le.
8660 (rs6000_expand_vec_perm_const): Delete.
8661 (rs6000_vectorize_vec_perm_const_ok): Delete.
8662 (rs6000_vectorize_vec_perm_const): New function.
8663 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8664 an element count and rtx array.
8665 (rs6000_expand_extract_even): Update call accordingly.
8666 (rs6000_expand_interleave): Likewise.
8667 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8668 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8669 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8670 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8671 (rs6000_expand_vec_perm_const): Delete.
8672 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8673 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8674 (altivec_expand_vec_perm_const_le): Take each operand individually.
8675 Operate on constant selectors rather than rtxes.
8676 (altivec_expand_vec_perm_const): Likewise. Update call to
8677 altivec_expand_vec_perm_const_le.
8678 (rs6000_expand_vec_perm_const): Delete.
8679 (rs6000_vectorize_vec_perm_const_ok): Delete.
8680 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8681 reference to the SPE evmerge intructions.
8682 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8683 an element count and rtx array.
8684 (rs6000_expand_extract_even): Update call accordingly.
8685 (rs6000_expand_interleave): Likewise.
8686 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8687 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8689 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8691 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8693 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8694 vector mode and that that mode matches the mode of the data
8696 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8697 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8698 directly using expand_vec_perm_1 when forcing selectors into
8700 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8702 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8704 * optabs-query.h (can_vec_perm_p): Delete.
8705 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8706 * optabs-query.c (can_vec_perm_p): Split into...
8707 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8708 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8709 particular selector is valid.
8710 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8711 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8712 (vect_grouped_load_supported): Likewise.
8713 (vect_shift_permute_load_chain): Likewise.
8714 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8715 (vect_transform_slp_perm_load): Likewise.
8716 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8717 (vectorizable_bswap): Likewise.
8718 (vect_gen_perm_mask_checked): Likewise.
8719 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8720 implementations of variable permutation vectors into account
8721 when deciding which selector to use.
8722 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8723 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8724 with a false third argument.
8725 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8726 to test whether the constant selector is valid and can_vec_perm_var_p
8727 to test whether a variable selector is valid.
8729 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8731 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8732 * optabs-query.c (can_vec_perm_p): Likewise.
8733 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8734 instead of vec_perm_indices.
8735 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8736 (vect_gen_perm_mask_checked): Likewise,
8737 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8738 (vect_gen_perm_mask_checked): Likewise,
8740 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8742 * optabs-query.h (qimode_for_vec_perm): Declare.
8743 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8744 (qimode_for_vec_perm): ...this new function.
8745 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8747 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8749 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8750 does not have a conditional at the top.
8752 2018-01-02 Richard Biener <rguenther@suse.de>
8754 * ipa-inline.c (big_speedup_p): Fix expression.
8756 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8759 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8762 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8766 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8767 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8768 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8769 cond_taken_branch_cost 3->4.
8771 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8773 PR tree-optimization/83581
8774 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8775 TODO_cleanup_cfg if any changes have been made.
8778 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8779 convert_modes if target mode has the right side, but different mode
8783 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8784 last argument when extracting from CONCAT. If either from_real or
8785 from_imag is NULL, use expansion through memory. If result is not
8786 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8787 the parts directly to inner mode, if even that fails, use expansion
8791 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8792 check for bswap in mode rather than HImode and use that in expand_unop
8795 Copyright (C) 2018 Free Software Foundation, Inc.
8797 Copying and distribution of this file, with or without modification,
8798 are permitted in any medium without royalty provided the copyright
8799 notice and this notice are preserved.