1 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7 (tlsdesc_small_<mode>): Turn a define_expand and use
8 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9 (tlsdesc_small_advsimd_<mode>): ...this.
10 (tlsdesc_small_sve_<mode>): New pattern.
12 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
15 (UNSPEC_UMUL_HIGHPART): New constants.
16 (MUL_HIGHPART): New int iteraor.
17 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
18 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
20 (*<su>mul<mode>3_highpart): New define_insn.
22 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
25 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
28 2018-03-13 Martin Liska <mliska@suse.cz>
31 * (sem_item_optimizer::sem_item_optimizer): Initialize new
33 (sem_item_optimizer::~sem_item_optimizer): Release it.
34 (sem_item_optimizer::merge_classes): Register variable aliases.
35 (sem_item_optimizer::fixup_pt_set): New function.
36 (sem_item_optimizer::fixup_points_to_sets): Likewise.
37 * ipa-icf.h: Declare new variables and functions.
39 2018-03-13 Jakub Jelinek <jakub@redhat.com>
42 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
43 integer_pow2p@2 and test integer_pow2p in condition.
44 (A < 0 ? C : 0): Similarly for @1.
47 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
48 characters starting at p contain '\0' character, don't look beyond
52 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
53 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
56 * reg-stack.c (change_stack): Change update_end var from int to
57 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
58 also call set_block_for_insn on the newly added insns and rescan.
61 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
65 * tree.c (stabilize_reference_1): Return save_expr (e) for
66 STATEMENT_LIST even if it doesn't have side-effects.
68 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
70 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
72 2018-03-12 Renlin Li <renlin.li@arm.com>
74 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
75 aarch64_output_scalar_simd_mov_immediate.
77 2018-03-12 Martin Sebor <msebor@redhat.com>
79 PR tree-optimization/83456
80 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
81 for perfectly overlapping calls to memcpy.
82 (gimple_fold_builtin_memory_chk): Same.
83 (gimple_fold_builtin_strcpy): Handle no-warning.
84 (gimple_fold_builtin_stxcpy_chk): Same.
85 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
87 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
89 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
90 parameter. Use it for SFmode.
91 (rs6000_function_arg_advance_1): Adjust.
92 (rs6000_function_arg): Adjust.
93 (rs6000_gimplify_va_arg): Pass false for that new parameter.
95 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
97 PR rtl-optimization/84169
98 PR rtl-optimization/84780
99 * combine.c (can_combine_p): Check for a 2-insn combination whether
100 the destination register is used between the two insns, too.
102 2018-03-12 Richard Biener <rguenther@suse.de>
104 PR tree-optimization/84803
105 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
106 for refs DR analysis didn't process.
108 2018-03-12 Richard Biener <rguenther@suse.de>
110 PR tree-optimization/84777
111 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
112 force-vectorize loops ignore whether we are optimizing for size.
114 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
116 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
117 (TARGET_MD_ASM_ADJUST): Define.
119 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
120 Kito Cheng <kito.cheng@gmail.com>
121 Chung-Ju Wu <jasonwucj@gmail.com>
123 * config/nds32/nds32.c (nds32_compute_stack_frame,
124 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
125 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
126 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
127 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
128 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
129 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
130 * config/nds32/nds32.md (prologue, epilogue): Use macro
131 NDS32_V3PUSH_AVAILABLE_P to do checking.
133 2018-03-11 Jakub Jelinek <jakub@redhat.com>
136 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
137 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
138 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
139 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
140 addition of most attributes on !orig_type_die or the attribute not
141 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
143 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
144 Chung-Ju Wu <jasonwucj@gmail.com>
146 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
148 * config/nds32/nds32.opt (mvh): New option.
150 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
151 Chung-Ju Wu <jasonwucj@gmail.com>
153 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
155 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
156 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
159 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
160 Chung-Ju Wu <jasonwucj@gmail.com>
162 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
164 * config/nds32/nds32-multiple.md (strlensi): New pattern.
165 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
167 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
168 Kito Cheng <kito.cheng@gmail.com>
169 Chung-Ju Wu <jasonwucj@gmail.com>
171 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
172 UNSPEC_FFMISM and UNSPEC_FLMISM.
173 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
174 for ffb, ffmism and flmism.
175 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
176 (unspec_ffmism): Ditto.
177 (unspec_flmism): Ditto.
178 (nds32_expand_builtin_impl): Check if string extension is available.
179 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
180 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
182 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
185 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
188 * lra-assigns.c (assign_by_spills): Return a flag of reload
189 assignment failure. Do not process the reload assignment
190 failures. Do not spill other reload pseudos if they has the same
192 (lra_assign): Add a return arg. Set up from the result of
193 assign_by_spills call.
194 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
195 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
196 usage_insns if it is not NULL.
197 (spill_hard_reg_in_range): New function.
198 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
199 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
201 (lra_assign): Change prototype.
202 * lra.c (lra): Add code to deal with fails by splitting hard reg
205 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
208 * config/i386/i386.opt: Replace Enforcment with Enforcement.
210 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
213 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
214 (dw_val_node): Add val_symbolic_view.
215 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
216 (symview_upper_bound): New.
217 (new_line_info_table): Initialize symviews_since_reset.
218 (dwarf2out_source_line): Count symviews_since_reset and set
220 (dw_val_equal_p): Handle symview.
221 (add_AT_symview): New.
222 (print_dw_val): Handle symview.
223 (attr_checksum, attr_checksum_ordered): Likewise.
224 (same_dw_val_p, size_of_die): Likewise.
225 (value_format, output_die): Likewise.
226 (add_high_low_attributes): Use add_AT_symview for entry_view.
227 (dwarf2out_finish): Reset symview_upper_bound, clear
230 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
233 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
234 Add strict argument and use it.
235 (rs6000_split_multireg_move): Update for new strict argument.
236 (mem_operand_gpr): Disallow all non-offsettable addresses.
237 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
239 2018-03-09 Jakub Jelinek <jakub@redhat.com>
242 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
243 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
244 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
247 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
248 decl, use remap_type if we want to use the type.
250 2018-03-09 Martin Sebor <msebor@redhat.com>
252 PR tree-optimization/84526
253 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
255 (builtin_access::generic_overlap): Be prepared to handle non-array
258 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
260 PR rtl-optimization/84682
261 * lra-constraints.c (process_address_1): Check is_address flag
262 for address constraints.
263 (process_alt_operands): Likewise.
264 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
265 preprocess_constraints.
266 * recog.h (preprocess_constraints): Add oploc parameter.
268 * recog.c (preprocess_constraints): Test address_operand for
269 CT_ADDRESS constraints.
271 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
274 * lra-assigns.c (assign_by_spills): Return a flag of reload
275 assignment failure. Do not process the reload assignment
276 failures. Do not spill other reload pseudos if they has the same
278 (lra_assign): Add a return arg. Set up from the result of
279 assign_by_spills call.
280 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
281 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
282 usage_insns if it is not NULL.
283 (spill_hard_reg_in_range): New function.
284 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
285 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
287 (lra_assign): Change prototype.
288 * lra.c (lra): Add code to deal with fails by splitting hard reg
291 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
294 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
295 Accept complain bool parameter. Only emit errors if it is true.
296 (arm_parse_cpu_option_name): Likewise.
297 (arm_target_thumb_only): Adjust callers of the above.
298 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
299 prototype to take a default true bool parameter.
300 (arm_parse_arch_option_name): Likewise.
302 2018-03-09 David Malcolm <dmalcolm@redhat.com>
303 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
307 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
308 * configure: Regenerate.
309 * configure.ac ("linker --version-script option"): New.
310 ("linker soname option"): New.
312 2018-03-09 Richard Biener <rguenther@suse.de>
314 PR tree-optimization/84775
315 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
316 immediate uses of predicate stmts and mark them modified.
319 PR tree-optimization/84178
320 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
322 (version_loop_for_if_conversion): Delay update_ssa call.
323 (tree_if_conversion): Delay update_ssa until after predicate
326 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
329 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
330 when the function accesses prior frames.
332 2018-03-08 Jakub Jelinek <jakub@redhat.com>
335 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
336 gen_llsym, otherwise call maybe_gen_llsym.
339 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
340 has ',' character inside of it.
342 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
345 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
346 as clobbering CC_REGNUM.
348 2018-03-08 Richard Biener <rguenther@suse.de>
351 * tree-scalar-evolution.c: Include tree-into-ssa.h.
352 (follow_copies_to_constant): Do not follow SSA names registered
355 2018-03-08 Richard Biener <rguenther@suse.de>
357 PR tree-optimization/84178
358 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
360 (version_loop_for_if_conversion): Delay update_ssa call.
361 (tree_if_conversion): Delay update_ssa until after predicate
364 2018-03-08 David Malcolm <dmalcolm@redhat.com>
366 PR tree-optimization/84178
367 * tree-if-conv.c (release_bb_predicate): Remove the
368 the assertion that the stmts have NULL use_ops.
369 Discard the statements, asserting that they haven't
370 yet been added to a BB.
372 2018-03-08 Richard Biener <rguenther@suse.de>
374 PR tree-optimization/84746
375 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
376 (phi_translate): Pass in destination ANTIC_OUT set.
377 (phi_translate_1): Likewise. For a simplified result lookup
378 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
379 (phi_translate_set): Adjust.
380 (do_pre_regular_insertion): Likewise.
381 (do_pre_partial_partial_insertion): Likewise.
383 2018-03-08 Martin Liska <mliska@suse.cz>
385 PR gcov-profile/84735
386 * doc/gcov.texi: Document usage of profile files.
387 * gcov-io.h: Document changes in the format.
389 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
393 * dwarf2out.c (struct dw_line_info_table): Update comments for
395 (FORCE_RESET_NEXT_VIEW): New.
396 (FORCE_RESETTING_VIEW_P): New.
397 (RESETTING_VIEW_P): Check for -1 too.
398 (ZERO_VIEW_P): Likewise.
399 (new_line_info_table): Force-reset next view.
400 (dwarf2out_begin_function): Likewise.
401 (dwarf2out_source_line): Simplify zero_view_p initialization.
402 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
403 view directly. Omit view when omitting .loc at line 0.
405 2018-03-08 Jakub Jelinek <jakub@redhat.com>
407 PR tree-optimization/84740
408 * tree-switch-conversion.c (process_switch): Call build_constructors
409 only if info.phi_count is non-zero.
411 PR tree-optimization/84739
412 * tree-tailcall.c (find_tail_calls): Check call arguments against
413 DECL_ARGUMENTS (current_function_decl) rather than
414 DECL_ARGUMENTS (func) when checking for tail recursion.
416 2018-03-07 Jakub Jelinek <jakub@redhat.com>
418 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
419 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
420 Volker Reichelt's entry and add entries for people that perform
421 GCC fuzzy testing and report numerous bugs.
423 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
426 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
427 readonly data in sdata, if that is disabled.
428 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
429 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
430 -mreadonly-in-sdata option.
432 2018-03-07 Martin Sebor <msebor@redhat.com>
434 PR tree-optimization/84468
435 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
436 basic block when looking for nul assignment.
438 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
441 * except.h (output_function_exception_table): Adjust prototype.
442 * except.c (output_function_exception_table): Remove FNNAME parameter
443 and add SECTION parameter. Ouput one part of the table at a time.
444 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
445 the first part of the exception table and emit unwind directives.
446 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
447 (i386_pe_seh_cold_init): Likewise.
448 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
449 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
450 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
451 (ix86_output_call_insn): Emit a nop in one more case for SEH.
452 * config/i386/winnt.c: Include except.h.
453 (struct seh_frame_state): Add reg_offset, after_prologue and
454 in_cold_section fields.
455 (i386_pe_seh_end_prologue): Set seh->after_prologue.
456 (i386_pe_seh_cold_init): New function.
457 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
458 to seh->in_cold_section.
459 (seh_emit_push): Record the offset of the push.
460 (seh_emit_save): Record the offet of the save.
461 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
462 Test seh->after_prologue to disregard the epilogue.
463 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
464 (i386_pe_end_cold_function): New function.
466 2018-03-07 Jakub Jelinek <jakub@redhat.com>
469 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
470 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
473 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
475 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
476 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
479 * multiple_target.c: Include tree-inline.h and intl.h.
480 (expand_target_clones): Diagnose and fail if node->definition and
481 !tree_versionable_function_p (node->decl).
483 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
485 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
487 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
488 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
489 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
491 2018-03-06 Jakub Jelinek <jakub@redhat.com>
494 * combine.c (try_combine): Use reg_or_subregno instead of handling
495 just paradoxical SUBREGs and REGs.
497 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
499 * config/arc/arc.c (arc_finalize_pic): Remove function.
500 (arc_must_save_register): We use single base PIC register, remove
501 checks to save/restore the PIC register.
502 (arc_expand_prologue): Likewise.
503 * config/arc/arc-protos.h (arc_set_default_type_attributes):
505 (arc_verify_short): Likewise.
506 (arc_attr_type): Likewise.
507 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
508 (walk_stores): Likewise.
509 (arc_address_cost): Make it static.
510 (arc_verify_short): Likewise.
511 (branch_dest): Likewise.
512 (arc_attr_type): Likewise.
513 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
514 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
515 (arc_final_prescan_insn): Remove inserting the nops due to
516 hardware hazards. It is done in reorg step.
517 (insn_length_variant_t): Remove.
518 (insn_length_parameters_t): Likewise.
519 (arc_insn_length_parameters): Likewise.
520 (arc_get_insn_variants): Likewise.
521 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
523 2018-03-06 Jakub Jelinek <jakub@redhat.com>
526 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
529 PR tree-optimization/84687
530 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
532 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
534 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
536 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
537 Rename to ppc_speculation_barrier.
538 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
539 __builtin_ppc_speculation_barrier.
541 2018-03-05 Jakub Jelinek <jakub@redhat.com>
544 * combine.c (combine_simplify_rtx): Don't try to simplify if
545 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
548 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
550 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
551 to 32 bytes when compiling for POWER9.
553 2018-03-05 Jakub Jelinek <jakub@redhat.com>
556 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
557 regparm >= 3 with no arg reg available also for calls with
558 flag_force_indirect_call. Pass decl to ix86_function_regparm.
561 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
563 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
565 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
568 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
570 2018-03-05 Richard Biener <rguenther@suse.de>
572 PR tree-optimization/84486
573 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
574 When inserting a __builtin_assume_aligned call set the LHS
575 SSA name alignment info accordingly.
577 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
579 PR tree-optimization/84114
580 * config/aarch64/aarch64.c (aarch64_reassociation_width)
581 Avoid reassociation of FLOAT_MODE addition.
583 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
585 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
586 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
587 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
588 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
589 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
590 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
591 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
593 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
594 __builtin_ia32_wbinvd): New builtins.
595 (SPECIAL_ARGS2): New.
596 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
597 (SPECIAL_ARGS2): New.
598 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
599 (ix86_valid_target_attribute_inner_p): Ditto.
600 (ix86_init_mmx_sse_builtins): Add special_args2.
601 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
602 TARGET_WBNOINVD_P): New.
603 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
604 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
605 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
606 * config/i386/immintrin.h (_wbinvd): New intrinsic.
607 * config/i386/pconfigintrin.h: New file.
608 * config/i386/wbnoinvdintrin.h: Ditto.
609 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
610 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
612 2018-03-05 Richard Biener <rguenther@suse.de>
614 PR tree-optimization/84670
615 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
617 (BB_VISITED_WITH_VISITED_SUCCS): New define.
618 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
619 (compute_antic_aux): Only assert the number of values in ANTIC_IN
620 doesn't grow if all successors (recursively) were visited at least
623 2018-03-05 Richard Biener <rguenther@suse.de>
625 PR tree-optimization/84650
626 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
627 if executed in the loop pipeline.
629 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
631 * doc/configfiles.texi (Configuration Files): Move info about
632 conditionalizing $target-protos.h to...
633 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
634 differs from $target-protos.h.
636 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
637 Chung-Ju Wu <jasonwucj@gmail.com>
639 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
640 * config/nds32/nds32-multiple.md (setmemsi): Define.
641 * config/nds32/nds32-memory-manipulation.c
642 (nds32_gen_dup_4_byte_to_word_value): New.
643 (emit_setmem_word_loop): New.
644 (emit_setmem_byte_loop): New.
645 (nds32_expand_setmem_loop): New.
646 (nds32_expand_setmem_loop_v3m): New.
647 (nds32_expand_setmem_unroll): New.
648 (nds32_expand_setmem): New.
650 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
651 Chung-Ju Wu <jasonwucj@gmail.com>
653 * config/nds32/nds32-memory-manipulation.c
654 (nds32_emit_load_store): New.
655 (nds32_emit_post_inc_load_store): New.
656 (nds32_emit_mem_move): New.
657 (nds32_emit_mem_move_block): New.
658 (nds32_expand_movmemsi_loop_unknown_size): New.
659 (nds32_expand_movmemsi_loop_known_size): New.
660 (nds32_expand_movmemsi_loop): New.
661 (nds32_expand_movmemsi_unroll): New.
662 (nds32_expand_movmemqi): Rename ...
663 (nds32_expand_movmemsi): ... to this.
664 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
665 (movmemsi): ... to this.
666 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
667 (nds32_expand_movmemsi): ... to this.
669 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
670 Monk Chiang <sh.chiang04@gmail.com>
671 Chung-Ju Wu <jasonwucj@gmail.com>
673 * config/nds32/nds32-protos.h
674 (nds32_expand_load_multiple): New arguments.
675 (nds32_expand_store_multiple): Ditto.
676 (nds32_valid_multiple_load_store): Rename ...
677 (nds32_valid_multiple_load_store_p): ... to this.
678 * config/nds32/nds32-memory-manipulation.c
679 (nds32_expand_load_multiple): Refine implementation.
680 (nds32_expand_store_multiple): Ditto.
681 * config/nds32/nds32-multiple.md
682 (load_multiple): Update nds32_expand_load_multiple interface.
683 (store_multiple): Update nds32_expand_store_multiple interface.
684 * config/nds32/nds32-predicates.c
685 (nds32_valid_multiple_load_store): Rename ...
686 (nds32_valid_multiple_load_store_p): ... to this and refine
688 * config/nds32/predicates.md
689 (nds32_load_multiple_and_update_address_operation): New predicate.
690 (nds32_store_multiple_and_update_address_operation): New predicate.
692 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
693 Chung-Ju Wu <jasonwucj@gmail.com>
695 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
696 (combo): New attribute.
697 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
699 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
701 * config/nds32/nds32.opt: Change -mcmodel= default value.
703 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
704 Monk Chiang <sh.chiang04@gmail.com>
705 Chung-Ju Wu <jasonwucj@gmail.com>
707 * config/nds32/constants.md (unspec_element): New enum.
708 * config/nds32/constraints.md (Umw): New constraint.
709 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
710 * config/nds32/nds32-intrinsic.md: Likewise.
711 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
712 (nds32_valid_smw_lwm_base_p): New.
713 (nds32_output_smw_single_word): New.
714 (nds32_output_lmw_single_word): New.
715 (nds32_expand_unaligned_load): New.
716 (nds32_expand_unaligned_store): New.
717 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
718 (nds32_output_smw_single_word): Declare.
719 (nds32_output_lmw_single_word): Declare.
720 (nds32_expand_unaligned_load): Declare.
721 (nds32_expand_unaligned_store): Declare.
722 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
723 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
724 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
725 NDS32_BUILTIN_UASTORE_DW.
726 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
729 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
730 Kito Cheng <kito.cheng@gmail.com>
731 Chung-Ju Wu <jasonwucj@gmail.com>
733 * config/nds32/nds32-intrinsic.c
734 (nds32_expand_builtin_null_ftype_reg): Delete.
735 (nds32_expand_builtin_reg_ftype_imm): Ditto.
736 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
737 (nds32_read_argument): New.
738 (nds32_legitimize_target): Ditto.
739 (nds32_legitimize_argument): Ditto.
740 (nds32_check_constant_argument): Ditto.
741 (nds32_expand_unop_builtin): Ditto.
742 (nds32_expand_unopimm_builtin): Ditto.
743 (nds32_expand_binop_builtin): Ditto.
744 (nds32_builtin_decl_impl): Ditto.
745 (builtin_description): Ditto.
746 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
747 (nds32_init_builtins_impl): Ditto.
748 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
749 (nds32_builtin_decl): New.
750 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
751 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
753 2018-03-02 Jeff Law <law@redhat.com>
755 * reorg.c (stop_search_p): Handle DEBUG_INSN.
756 (redundant_insn, fill_simple_delay_slots): Likewise.
757 (fill_slots_from_thread): Likewise.
758 * resource.c (mark_referenced_resources): Likewise.
759 (mark_set_resources, find_dead_or_set_registers): Likewise.
761 2018-03-02 Jakub Jelinek <jakub@redhat.com>
763 * substring-locations.h (format_warning_va): Formatting fix for
765 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
767 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
768 * substring-locations.c: Include intl.h.
769 (format_warning_va): Turned into small wrapper around
770 format_warning_n_va, renamed to ...
771 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
772 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
774 (format_warning_at_substring_n): New function.
775 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
776 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
777 format_warning_at_substring with just a shorter name instead of
778 const function pointer.
779 (fmtwarn_n): New function.
780 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
781 appropriate, get rid of all the fmtstr temporaries, move conditionals
782 with G_() wrapped string literals directly into fmtwarn arguments,
783 cast dir.len to (int), formatting fixes.
785 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
787 * doc/invoke.texi: Remove "Cilk Plus" references.
789 2018-03-02 Jakub Jelinek <jakub@redhat.com>
790 Richard Biener <rguenther@suse.de>
793 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
794 for error or warning attributes if CALL_FROM_THUNK_P is set.
797 2018-03-02 Jakub Jelinek <jakub@redhat.com>
800 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
801 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
804 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
805 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
807 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
808 instead of -1U in last predictors element's probability member.
810 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
813 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
814 arguments if they are comparable.
816 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
818 PR tree-optimization/84634
819 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
820 masks and masked_loop_p with a single loop_masks, making sure it's
821 null for bb vectorization.
823 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
825 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
826 (vect_analyze_data_ref_access): Use loop->safe_len rather than
827 loop->force_vectorize to check whether there is no alias.
829 2018-03-02 Jakub Jelinek <jakub@redhat.com>
832 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
834 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
836 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
837 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
838 instead of a loop around prev_real_insn.
839 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
843 * config/i386/i386.c (ix86_print_operand): Use conditional
844 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
847 2018-03-02 Richard Biener <rguenther@suse.de>
849 PR tree-optimization/84427
850 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
851 (bitmap_set_subtract_values): Rewrite to handle multiple
854 (prune_clobbered_mems): Likewise.
855 (phi_translate): Take edge instead of pred/phiblock.
856 (phi_translate_1): Likewise.
857 (phi_translate_set): Likewise. Insert all translated
858 exprs for a value into the set, keeping possibly multiple
859 expressions per value.
860 (compute_antic_aux): Adjust for phi_translate changes.
861 When intersecting union the expressions and prune those
862 not in the final value set, keeping possibly multiple
863 expressions per value. Do not use value-insertion
864 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
865 all expressions. Add verification that the value-sets
866 only shrink during iteration.
867 (compute_partial_antic_aux): Adjust for the phi_translate changes.
868 (do_pre_regular_insertion): Likewise.
869 (do_pre_partial_partial_insertion): Likewise.
871 2018-03-02 Richard Biener <rguenther@suse.de>
874 * config/darwin.c (saved_debug_info_level): New static global.
875 (darwin_asm_lto_start): Disable debug info generation for LTO out.
876 (darwin_asm_lto_end): Restore debug info generation settings.
878 2018-03-01 Martin Liska <mliska@suse.cz>
881 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
884 2018-03-01 Richard Biener <rguenther@suse.de>
887 * dwarf2out.c (gen_variable_die): Properly handle late VLA
888 type annotation with LTO when debug was disabled at compile-time.
890 2018-03-01 Matthew Fortune <mfortune@gmail.com>
892 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
894 (mips_final_postscan_insn): Likewise.
896 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
898 PR rtl-optimization/84528
899 * alias.c (init_alias_target): Add commentary.
900 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
901 a unique base value if the frame pointer is not eliminated
902 to the stack pointer.
904 2018-03-01 Tom de Vries <tom@codesourcery.com>
906 PR rtl-optimization/83327
907 * lra-int.h (hard_regs_spilled_into): Declare.
908 * lra.c (hard_regs_spilled_into): Define.
909 (init_reg_info): Init hard_regs_spilled_into.
910 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
911 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
912 (process_bb_lives): Handle hard_regs_spilled_into.
913 (lra_create_live_ranges_1): Before doing liveness propagation, clear
914 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
916 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
918 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
919 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
920 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
921 * config/rs6000/aix72.h: New file.
923 2018-02-28 Jakub Jelinek <jakub@redhat.com>
925 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
926 instead of warning_at with conditional singular and plural messages
930 * stor-layout.c (update_alignment_for_field): For
931 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
932 && !DECL_PACKED (field), do the alignment update, just use
933 only desired_align instead of MAX (type_align, desired_align)
935 (place_field): Don't do known_align < desired_align handling
936 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
937 is non-NULL, instead do it after rli->prev_field handling and
938 only if not within a bitfield word. For DECL_PACKED (field)
939 use type_align of BITS_PER_UNIT.
941 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
943 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
944 superfluous parentheses and trailing spaces.
946 2018-02-28 Richard Biener <rguenther@suse.de>
948 PR tree-optimization/84584
949 * graphite-scop-detection.c (scop_detection::add_scop): Discard
950 SCoPs with fake exit edge.
952 2018-02-28 Martin Liska <mliska@suse.cz>
955 * timevar.c (timer::print): Fix format to properly print 100%
958 2018-02-28 Richard Biener <rguenther@suse.de>
961 * genmatch.c (capture_info::walk_match): Do not mark
962 captured expressions without operands as expr_p given
963 they act more like predicates and should be subject to
964 "lost tail" side-effect preserving.
966 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
968 PR rtl-optimization/81611
969 * auto-inc-dec.c (attempt_change): Move dead note from
970 mem_insn if it's the next use of regno
971 (find_address): Take address use of reg holding
972 non-incremented value. Add parm to limit search to the named
974 (merge_in_block): Attempt to use a mem insn that is the next
975 use of the original regno.
977 2018-02-27 Martin Sebor <msebor@redhat.com>
980 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
981 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
983 2018-02-27 Martin Sebor <msebor@redhat.com>
986 * diagnostic-core.h (warning_n, error_n, inform_n): Change
987 n argument to unsigned HOST_WIDE_INT.
988 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
989 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
990 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
991 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
993 2018-02-27 Richard Biener <rguenther@suse.de>
995 PR tree-optimization/84512
996 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
997 Do not use the estimate returned from record_stmt_cost for
998 the scalar iteration cost but sum properly using add_stmt_cost.
1000 2018-02-27 Richard Biener <rguenther@suse.de>
1002 PR tree-optimization/84466
1003 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1004 Adjust last change to less strictly validate use operands.
1006 2018-02-27 Martin Liska <mliska@suse.cz>
1008 PR gcov-profile/84548
1009 * gcov.c (process_file): Allow partial overlap and consider it
1010 also as group functions.
1011 (output_lines): Properly calculate range of lines for a group.
1013 2018-02-27 Martin Liska <mliska@suse.cz>
1015 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1016 'ggc' suffixes. Change first column width.
1017 (timer::print): Fix formatting of the column.
1019 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1021 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1022 preserve inline entry blocks for the sake of debug inline
1023 entry point markers alone.
1024 (remove_unused_locals): Suggest in comments a better place to
1025 force the preservation of inline entry blocks that are
1026 otherwise unused, but do not preserve them.
1028 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1030 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1032 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1035 * config/i386/constraints.md (Bs): Replace
1036 ix86_indirect_branch_register with
1037 TARGET_INDIRECT_BRANCH_REGISTER.
1039 * config/i386/i386.md (indirect_jump): Likewise.
1040 (tablejump): Likewise.
1041 (*sibcall_memory): Likewise.
1042 (*sibcall_value_memory): Likewise.
1043 Peepholes of indirect call and jump via memory: Likewise.
1044 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1045 (*sibcall_value_GOT_32): Likewise.
1046 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1047 (GOT_memory_operand): Likewise.
1048 (call_insn_operand): Likewise.
1049 (sibcall_insn_operand): Likewise.
1050 (GOT32_symbol_operand): Likewise.
1051 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1053 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1055 PR rtl-optimization/83496
1056 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1057 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1058 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1059 redundant insn, if any.
1060 (relax_delay_slots): Likewise.
1061 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1063 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1065 PR tree-optimization/83965
1066 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1067 that grouped statements are part of a reduction chain. Return
1068 true if the statement is not marked as a reduction itself but
1070 (vect_recog_dot_prod_pattern): Don't check whether the statement
1071 is part of a group here.
1072 (vect_recog_sad_pattern): Likewise.
1073 (vect_recog_widen_sum_pattern): Likewise.
1075 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1078 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1080 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1083 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1085 (ix86_output_indirect_function_return): New prototype.
1086 (ix86_split_simple_return_pop_internal): Likewise.
1087 * config/i386/i386.c (indirect_return_via_cx): New.
1088 (indirect_return_via_cx_bnd): Likewise.
1089 (indirect_thunk_name): Handle return va CX_REG.
1090 (output_indirect_thunk_function): Create alias for
1091 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1092 (ix86_output_indirect_jmp): Remove the bool argument.
1093 (ix86_output_indirect_function_return): New function.
1094 (ix86_split_simple_return_pop_internal): Likewise.
1095 * config/i386/i386.md (*indirect_jump): Don't pass false
1096 to ix86_output_indirect_jmp.
1097 (*tablejump_1): Likewise.
1098 (simple_return_pop_internal): Change it to define_insn_and_split.
1099 Call ix86_split_simple_return_pop_internal to split it for
1101 (simple_return_indirect_internal): Call
1102 ix86_output_indirect_function_return instead of
1103 ix86_output_indirect_jmp.
1105 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1108 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1109 memset and value initialization afterwards.
1111 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1113 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1115 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1118 * common/config/aarch64/aarch64-common.c
1119 (aarch_option_optimization_table[]): Switch
1120 off fomit-frame-pointer
1122 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1123 Chung-Ju Wu <jasonwucj@gmail.com>
1125 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1127 (store_multiple): Ditto.
1129 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1131 * config.gcc: Add --with-cpu support for nds32 target.
1132 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1133 * config/nds32/nds32.opt: Add -mcpu= option.
1135 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1137 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1138 isel=yes): Warn for these deprecated options.
1140 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1142 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1143 ISA_2_5_MASKS_EMBEDDED.
1145 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1147 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1148 p->max as pointers rather than using iterative_hash_expr.
1150 2018-02-23 Carl Love <cel@us.ibm.com>
1152 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1153 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1155 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1156 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1157 P8V_BUILTIN_VEC_VUNSIGNED2.
1159 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1162 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1163 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1165 * lra-constraints.c (curr_insn_transform): Set up
1166 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1168 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1169 * lra-eliminations.c (spill_pseudos): Ditto.
1170 (process_insn_for_elimination): Ditto.
1171 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1172 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1173 LRA_NON_CLOBBERED_ALT.
1175 2018-02-22 Martin Sebor <msebor@redhat.com>
1177 PR tree-optimization/84480
1178 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1179 to maybe_diag_stxncpy_trunc. Call it.
1180 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1181 from gimple_fold_builtin_strcpy. Print inlining stack.
1182 (handle_builtin_stxncpy): Print inlining stack.
1183 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1185 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1188 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1189 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1190 and -fcheck-pointer-bounds are used together.
1191 (indirect_thunk_prefix): New enum.
1192 (indirect_thunk_need_prefix): New function.
1193 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1194 "_nt" instead of "_bnd" for NOTRACK prefix.
1195 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1196 (output_indirect_thunk_function): Likewise.
1198 (ix86_code_end): Update output_indirect_thunk_function calls.
1199 (ix86_output_indirect_branch_via_reg): Replace
1200 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1201 (ix86_output_indirect_branch_via_push): Likewise.
1202 (ix86_output_function_return): Likewise.
1203 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1204 incompatible with -fcf-protection=branch and
1205 -fcheck-pointer-bounds.
1207 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1210 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1211 Change gcc_assert call to output_operand_lossage.
1213 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1215 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1217 2018-02-22 DJ Delorie <dj@redhat.com>
1218 Sebastian Perta <sebastian.perta@renesas.com>
1219 Oleg Endo <olegendo@gcc.gnu.org>
1221 * config/rx/rx.c (rx_rtx_costs): New function.
1222 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1224 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1226 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1228 2018-02-22 Martin Liska <mliska@suse.cz>
1231 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1232 Add "native" as a possible value.
1234 2018-02-22 Martin Liska <mliska@suse.cz>
1237 * config/i386/i386.c (ix86_option_override_internal):
1238 Add "native" as a possible value for -march and -mtune.
1240 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1243 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1244 to all type variants.
1246 PR tree-optimization/84503
1247 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1248 width as info->bitpos + info->bitsize - start.
1249 (merged_store_group::merge_overlapping): Simplify width computation.
1250 (check_no_overlap): New function.
1251 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1252 start + width and last_order of the group, fail if check_no_overlap
1254 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1255 to group if check_no_overlap fails.
1257 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1259 * config/rs6000/altivec.md: Delete contraint arguments to
1260 define_expand, define_split, and define_peephole2, and in
1261 define_insn_and_split if always unused.
1262 * config/rs6000/darwin.md: Ditto.
1263 * config/rs6000/dfp.md: Ditto.
1264 * config/rs6000/rs6000.md: Ditto.
1265 * config/rs6000/sync.md: Ditto.
1266 * config/rs6000/vector.md: Ditto.
1267 * config/rs6000/vsx.md: Ditto.
1269 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1271 * config/rs6000/altivec.md: Write output control strings as braced
1272 blocks instead of double-quoted strings.
1273 * config/rs6000/darwin.md: Ditto.
1274 * config/rs6000/rs6000.md: Ditto.
1275 * config/rs6000/vector.md: Ditto.
1276 * config/rs6000/vsx.md: Ditto.
1278 2018-02-21 Jason Merrill <jason@redhat.com>
1280 PR c++/84314 - ICE with templates and fastcall attribute.
1281 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1283 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1285 * ipa-cp.c (determine_versionability): Fix comment typos.
1287 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1290 * ipa-cp.c (determine_versionability): Do not version functions caling
1293 2018-02-21 Martin Liska <mliska@suse.cz>
1296 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1297 Add "native" as a possible value.
1298 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1299 the macro when native cpu detection is available.
1301 2018-02-21 Martin Liska <mliska@suse.cz>
1304 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1305 Add "native" as a possible value.
1306 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1307 when native cpu detection is available.
1309 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1310 Martin Sebor <msebor@redhat.com>
1312 PR tree-optimization/84478
1313 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1315 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1316 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1317 support which is conservatively correct, for 2 only stay conservative
1318 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1319 argument to the 2 argument get_range_strlen, adjust 6 arg
1320 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1322 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1323 (gimple_fold_builtin_strlen): Pass true as last argument to
1326 2018-02-20 Martin Sebor <msebor@redhat.com>
1329 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1330 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1331 (builtin_memref::builtin_memref): Factor out parts into
1332 set_base_and_offset and call it.
1334 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1337 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1338 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1339 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1340 search at the associated MODE_INT.
1342 2018-02-20 Jeff Law <law@redhat.com>
1345 PR tree-optimization/81592
1347 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1348 for range data rather than using global data.
1349 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1350 range data rather than using global data.
1351 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1352 pass it to children as needed.
1353 (struct directive::fmtresult): Similarly.
1354 (struct directive::set_width): Similarly.
1355 (struct directive::set_precision): Similarly.
1356 (format_integer, format_directive, parse_directive): Similarly.
1357 (format_none): Accept unnamed vr_values parameter.
1358 (format_percent, format_floating, format_character): Similarly.
1359 (format_string, format_plain): Similarly.
1360 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1361 the EVRP range analyzer for range data rather than using global data.
1362 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1363 gimple-ssa-evrp-analyze.h
1364 (class sprintf_dom_walker): Add after_dom_children member function.
1365 Add evrp_range_analyzer member.
1366 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1367 range analyzer as needed.
1368 (sprintf_dom_walker::after_dom_children): New member function.
1369 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1371 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1372 (evrp_range_analyzer::pop_to_marker): Likewise.
1374 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1376 PR tree-optimization/84419
1377 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1378 with the required type if its current type is compatible but
1381 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1384 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1385 after vectorization.
1387 2018-02-20 Martin Liska <mliska@suse.cz>
1390 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1391 possible values if we don't have a hint.
1393 2018-02-20 Martin Liska <mliska@suse.cz>
1397 * final.c (shorten_branches): Build align_tab array with one
1399 * opts.c (finish_options): Add alignment option limit check.
1400 (MAX_CODE_ALIGN): Likewise.
1401 (MAX_CODE_ALIGN_VALUE): Likewise.
1402 * doc/invoke.texi: Document maximum allowed option value for
1403 all -falign-* options.
1405 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1408 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1409 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1410 * var-tracking.c (emit_note_insn_var_location): Remove all references
1411 to NOTE_INSN_CALL_ARG_LOCATION.
1412 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1413 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1414 Use copy_rtx_if_shared.
1415 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1416 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1417 (dwarf2out_var_location): Remove handling of
1418 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1420 * final.c (final_scan_insn): Remove all references to
1421 NOTE_INSN_CALL_ARG_LOCATION.
1422 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1423 before dumping final insns.
1424 * except.c (emit_note_eh_region_end): Remove all references to
1425 NOTE_INSN_CALL_ARG_LOCATION.
1426 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1427 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1428 * config/arc/arc.c (hwloop_optimize): Likewise.
1429 * config/arm/arm.c (create_fix_barrier): Likewise.
1430 * config/s390/s390.c (s390_chunkify_start): Likewise.
1431 * config/sh/sh.c (find_barrier): Likewise.
1432 * config/i386/i386.c (rest_of_insert_endbranch,
1433 ix86_seh_fixup_eh_fallthru): Likewise.
1434 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1435 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1436 * config/frv/frv.c (frv_function_prologue): Likewise.
1437 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1439 (note_outside_basic_block_p): Remove all references to
1440 NOTE_INSN_CALL_ARG_LOCATION.
1441 * gengtype.c (adjust_field_rtx_def): Likewise.
1442 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1444 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1445 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1448 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1451 PR tree-optimization/84452
1452 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1453 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1456 2018-02-19 Martin Liska <mliska@suse.cz>
1459 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1461 2018-02-19 Martin Liska <mliska@suse.cz>
1462 Richard Sandiford <richard.sandiford@linaro.org>
1464 PR tree-optimization/82491
1465 * gimple-fold.c (get_base_constructor): Make earlier bail out
1468 2018-02-19 Carl Love <cel@us.ibm.com>
1470 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1471 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1473 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1474 P8V_BUILTIN_VEC_NEG.
1476 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1478 * config/rl78/rl78.md (movdf): New define expand.
1480 2018-02-19 Martin Liska <mliska@suse.cz>
1483 * doc/invoke.texi: Fix typo.
1484 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1486 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1488 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1489 handle rs6000_single_float and rs6000_double_float specially for
1492 2018-02-16 Jeff Law <law@redhat.com>
1494 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1497 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1500 * ipa-inline.c (inline_small_functions): Fix a typo.
1502 2018-02-16 Nathan Sidwell <nathan@acm.org>
1504 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1506 2018-02-16 Carl Love <cel@us.ibm.com>
1508 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1509 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1510 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1511 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1512 expansion to P8V_BUILTIN_VEC_FLOAT2.
1514 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1516 PR rtl-optimization/70023
1517 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1518 src_regno into account.
1520 2018-02-16 Carl Love <cel@us.ibm.com>
1522 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1523 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1524 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1525 * config/rs6000/rs6000.c: Remove case statements for
1526 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1527 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1528 and P9V_BUILTIN_VEC_VINSERT4B.
1529 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1530 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1531 * config/rs6000/vsx.md:
1532 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1535 2018-02-16 Carl Love <cel@us.ibm.com>
1537 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1539 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1541 * config/rs6000/rs6000-c.c: Add the definitions for
1542 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1543 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1544 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1545 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1546 definition for insert4b and define insn *insert3b_internal.
1547 * doc/extend.texi: Add documentation for vec_extract4b.
1549 2018-02-16 Nathan Sidwell <nathan@acm.org>
1551 * doc/extend.texi (Backwards Compatibility): Mention friend
1552 injection. Note for-scope is deprecated.
1553 * doc/invoke.texi (-ffriend-injection): Deprecate.
1555 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1557 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1558 that moved to I2, also allow destinations that are a paradoxical
1559 subreg (instead of a normal reg).
1561 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1564 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1567 2018-02-16 Richard Biener <rguenther@suse.de>
1569 PR tree-optimization/84037
1570 PR tree-optimization/84016
1572 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1573 Adjust vec_construct for the fact we need additional higher latency
1574 128bit inserts for AVX256 and AVX512 vector builds.
1575 (ix86_add_stmt_cost): Scale vector construction cost for
1578 2018-02-16 Richard Biener <rguenther@suse.de>
1580 PR tree-optimization/84417
1581 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1582 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1583 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1585 2018-02-16 Martin Liska <mliska@suse.cz>
1588 * internal-fn.def (ASAN_CHECK): Set proper flags.
1589 (ASAN_MARK): Likewise.
1591 2018-02-16 Julia Koval <julia.koval@intel.com>
1593 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1594 from PTA_CANNONLAKE.
1596 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1599 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1600 Use ++iter rather than iter++ for std::list iterators.
1601 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1602 defer deleting them until all nodes in the forest are processed. Do
1603 free even leaf nodes. Change to_process into auto_vec.
1606 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1607 * vec.h (vec_default_construct): Use memset instead of placement new
1608 if BROKEN_VALUE_INITIALIZATION is defined.
1609 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1610 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1613 PR rtl-optimization/83723
1614 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1615 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1616 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1618 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1620 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1622 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1624 PR rtl-optimization/81443
1625 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1626 from inner REGs to paradoxical SUBREGs.
1628 2018-02-16 Richard Biener <rguenther@suse.de>
1630 PR tree-optimization/84399
1631 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1632 For operands we can analyze at their definition make sure we can
1633 analyze them at each use as well.
1635 2018-02-16 Richard Biener <rguenther@suse.de>
1637 PR tree-optimization/84190
1638 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1639 volatile accesses if the decl isn't volatile.
1641 2018-02-15 Jason Merrill <jason@redhat.com>
1643 PR c++/84314 - ICE with templates and fastcall attribute.
1644 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1645 TYPE_CANONICAL on an existing type.
1647 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1649 PR tree-optimization/84383
1650 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1651 dstoff nor call operand_equal_p if dstbase is NULL.
1653 PR tree-optimization/84334
1654 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1655 also a CONSTANT_CLASS_P, punt.
1657 2018-02-14 Jim Wilson <jimw@sifive.com>
1659 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1660 first SMALL_OPERAND check. New local min_second_step. Move assert
1661 to where locals are set. Add TARGET_RVC support.
1662 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1664 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1666 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1668 2018-02-14 Martin Sebor <msebor@redhat.com>
1670 PR tree-optimization/83698
1671 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1672 arrays constrain the offset range to their bounds.
1673 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1674 (builtin_access::overlap): Avoid setting the size of overlap if it's
1676 (maybe_diag_overlap): Also consider arrays when deciding what values
1677 of offsets to include in diagnostics.
1679 2018-02-14 Martin Sebor <msebor@redhat.com>
1682 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1683 that correspond to the kind of a declaration.
1685 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1688 * config/pa/pa.md: Load address of PIC label using the linkage table
1689 if the label is nonlocal.
1691 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1693 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1694 warning message if user requests -maltivec=be.
1695 * doc/invoke.texi: Document deprecation of -maltivec=be.
1697 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1700 * config/rs6000/rs6000-c.c: Update definitions for
1701 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1702 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1704 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1707 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1708 add _get_ssp intrinsics. Remove argument from
1709 __builtin_ia32_rdssp[d|q].
1710 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1711 * config/i386/i386-builtin.def: Remove argument from
1712 __builtin_ia32_rdssp[d|q].
1713 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1714 ix86_expand_special_args_builtin for _rdssp[d|q].
1715 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1716 Clear register before usage.
1717 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1718 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1720 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1722 PR tree-optimization/84357
1723 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1724 operand 1 of an ARRAY_REF too.
1726 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1729 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1730 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1732 (set_of_reg): New struct.
1733 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1734 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1735 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1737 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1738 Split into bitclr, bitset, bitinvert patterns if appropriate.
1739 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1740 use rx_fuse_in_memory_bitop.
1741 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1742 to named insn, correct maximum insn length.
1744 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1747 * machmode.def: Define a complex mode for PARTIAL_INT.
1748 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1750 * doc/rtl.texi: Document CSPImode.
1751 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1753 (msp430_hard_regno_nregs_with_padding): Likewise.
1755 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1758 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1760 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1762 PR rtl-optimization/84169
1763 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1764 we generated a parallel as new i3 and we split that to new i2 and i3
1765 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1766 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1767 those to i2, not i1. Partially rewrite this scan code.
1769 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1772 * stor-layout.c (place_field): For variable length fields, adjust
1773 offset_align afterwards not just based on the field's alignment,
1774 but also on the size.
1777 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1778 of exps and logs in the use_exp2 case.
1780 2018-02-13 Jeff Law <law@redhat.com>
1782 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1785 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1788 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1792 * common.opt (gas-loc-support, gas-locview-support): New.
1793 (ginline-points, ginternal-reset-location-views): New.
1794 * doc/invoke.texi: Document them. Use @itemx where intended.
1795 (gvariable-location-views): Adjust.
1796 * target.def (reset_location_view): New.
1797 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1798 (TARGET_RESET_LOCATION_VIEW): New.
1799 * doc/tm.texi: Rebuilt.
1800 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1801 (dwarf2out_default_as_locview_support): New.
1802 (output_asm_line_debug_info): Use option variables.
1803 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1804 (output_loc_list): Likewise.
1805 (add_high_low_attributes): Check option variables.
1806 Don't output entry view attribute in strict mode.
1807 (gen_inlined_subroutine_die): Check option variables.
1808 (dwarf2out_inline_entry): Likewise.
1809 (init_sections_and_labels): Likewise.
1810 (dwarf2out_early_finish): Likewise.
1811 (maybe_reset_location_view): New, from...
1812 (dwarf2out_var_location): ... here. Call it.
1813 * debug.h (dwarf2out_default_as_loc_support): Declare.
1814 (dwarf2out_default_as_locview_support): Declare.
1815 * hooks.c (hook_int_rtx_insn_0): New.
1816 * hooks.h (hook_int_rtx_insn_0): Declare.
1817 * toplev.c (process_options): Take -gas-loc-support and
1818 -gas-locview-support from dwarf2out. Enable
1819 -gvariable-location-views by default only with locview
1820 assembler support. Enable -ginternal-reset-location-views by
1821 default only if the target defines the corresponding hook.
1822 Enable -ginline-points by default if location views are
1823 enabled; force it disabled if statement frontiers are
1825 * tree-inline.c (expand_call_inline): Check option variables.
1826 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1828 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1830 PR tree-optimization/84321
1831 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1832 handling. Also check whether the anti-range contains any values
1833 that satisfy the mask; switch to a VR_RANGE if not.
1835 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1838 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1840 2018-02-13 Martin Jambor <mjambor@suse.cz>
1843 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1844 of call statements, also set location of a load to a temporary.
1846 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1848 * config/rl78/rl78.c (add_vector_labels): New function.
1849 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1850 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1851 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1852 which checks that no arguments are passed.
1853 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1854 * doc/extend.texi: Documentation for the new attribute.
1856 2018-02-13 Andreas Schwab <schwab@suse.de>
1858 * config/riscv/linux.h (CPP_SPEC): Define.
1860 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1863 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1864 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1865 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1866 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1867 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1868 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1869 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1870 temporarily for AES and PCLMUL builtins.
1872 PR tree-optimization/84339
1873 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1874 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1878 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1879 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1880 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1882 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1884 * omp-simd-clone.h: New file.
1885 * omp-simd-clone.c: Include omp-simd-clone.h.
1886 (expand_simd_clones): No longer static.
1887 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1888 cgraph.h and omp-simd-clone.h.
1889 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1890 (vect_recog_widen_shift_pattern): Formatting fix.
1891 (vect_pattern_recog_1): Don't check optab for calls.
1894 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1895 operands[2] into a REG before using gen_lowpart on it.
1897 2018-02-12 Jeff Law <law@redhat.com>
1900 * config/sh/sh.c (find_barrier): Consider a sibling call
1903 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1904 successfully back substituting a reg.
1906 2018-02-12 Richard Biener <rguenther@suse.de>
1908 PR tree-optimization/84037
1909 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1910 parameter, move visited init to caller.
1911 (vect_slp_analyze_operations): Separate cost from validity
1912 check, initialize visited once for all instances.
1913 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1915 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1917 (vect_model_promotion_demotion_cost): Likewise.
1918 (vectorizable_bswap): Guard cost modeling with !slp_node
1919 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1921 (vectorizable_call): Likewise.
1922 (vectorizable_conversion): Likewise.
1923 (vectorizable_assignment): Likewise.
1924 (vectorizable_shift): Likewise.
1925 (vectorizable_operation): Likewise.
1926 (vectorizable_store): Likewise.
1927 (vectorizable_load): Likewise.
1928 (vectorizable_condition): Likewise.
1929 (vectorizable_comparison): Likewise.
1931 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1934 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1935 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1936 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1938 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1941 * params.def (inline-min-speedup): Increase from 8 to 15.
1942 (max-inline-insns-auto): Decrease from 40 to 30.
1943 * ipa-split.c (consider_split): Add some buffer for function to
1944 be considered inlining candidate.
1945 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1948 2018-02-12 Richard Biener <rguenther@suse.de>
1950 PR tree-optimization/84037
1951 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1952 matched stmts if we cannot swap the non-matched ones.
1954 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1956 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1957 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1958 _mm_maskz_scalef_round_ss): New intrinsics.
1959 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1960 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1961 __builtin_ia32_scalefss_round): Remove.
1962 (__builtin_ia32_scalefsd_mask_round,
1963 __builtin_ia32_scalefss_mask_round): New intrinsics.
1964 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1965 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1966 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1967 "<round_constraint>")): Changed to ...
1968 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1969 "<round_scalar_constraint>")): ... this.
1970 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1971 %0, %1, %2<round_op3>}"): Changed to ...
1972 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1973 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1974 %2<round_scalar_mask_op3>}"): ... this.
1975 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1977 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1979 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1980 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1981 (_mm_maskz_sqrt_round_ss): New intrinsics.
1982 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1983 (__builtin_ia32_sqrtsd_mask_round)
1984 (__builtin_ia32_sqrtss_mask_round): New builtins.
1985 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1986 (__builtin_ia32_sqrtss_round): Remove.
1987 (__builtin_ia32_sqrtsd_mask_round)
1988 (__builtin_ia32_sqrtss_mask_round): New builtins.
1989 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1990 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1991 ((match_operand:VF_128 1 "vector_operand"
1992 "xBm,<round_constraint>")): Changed to ...
1993 ((match_operand:VF_128 1 "vector_operand"
1994 "xBm,<round_scalar_constraint>")): ... this.
1995 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1996 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1997 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1998 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1999 %<iptr>1<round_scalar_mask_op3>}): ... this.
2000 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2001 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2003 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2006 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2007 Cast vec_cmpeq result to correct type.
2008 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2009 Cast vec_cmpgt result to correct type.
2011 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2013 * final.c (final_scan_insn_1): Renamed from...
2014 (final_scan_insn): ... this. New wrapper, to recover
2015 seen from the outermost call in recursive ones.
2016 * config/sparc/sparc.c (output_return): Drop seen from call.
2017 (output_sibcall): Likewise.
2018 * config/visium/visium.c (output_branch): Likewise.
2020 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2022 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2025 2018-02-10 Alan Modra <amodra@gmail.com>
2028 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2029 Specify LR as an input.
2031 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2034 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2035 remove_member_access_dummy_vars): New functions.
2036 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2037 lower_omp_1, execute_lower_omp): Use them.
2039 PR rtl-optimization/84308
2040 * shrink-wrap.c (spread_components): Release todo vector.
2042 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2044 PR rtl-optimization/57193
2045 * ira-color.c (struct allocno_color_data): Add member
2046 conflict_allocno_hard_prefs.
2047 (update_conflict_allocno_hard_prefs): New.
2048 (bucket_allocno_compare_func): Add a preference based on
2049 conflict_allocno_hard_prefs.
2050 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2051 (color_allocnos): Remove a dead code. Initiate
2052 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2054 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2057 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2058 constraint from =wa to wa. Avoid a subreg on the output operand,
2059 instead use a pseudo and subreg it in a move.
2060 (p9_xxbrd_<mode>): Changed to ...
2061 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2062 (p9_xxbrd_v2df): New expander.
2063 (p9_xxbrw_<mode>): Changed to ...
2064 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2065 (p9_xxbrw_v4sf): New expander.
2067 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2069 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2071 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2074 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2075 multiply in 32-bit mode.
2076 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2077 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2080 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2082 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2083 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2084 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2085 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2087 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2090 * dwarf2out.c (is_trivial_indirect_ref): New function.
2091 (dwarf2out_late_global_decl): Do not generate a location
2092 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2093 and that are not defined in the current unit.
2095 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2097 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2098 instead of a libcall for UNORDERED.
2100 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2103 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2104 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2106 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2109 * config/s390/s390.c (s390_set_current_function): Invoke
2110 s390_indirect_branch_settings also if fndecl didn't change.
2112 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2114 * config/rs6000/rs6000.md (blockage): Set length to zero.
2116 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2118 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2120 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2123 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2124 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2128 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2129 PARALLEL incoming that failed vt_get_decl_and_offset check.
2132 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2133 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2135 (get_variable_section): For decls in named .bss* sections pass true as
2136 second argument to bss_initializer_p.
2138 2018-02-09 Marek Polacek <polacek@redhat.com>
2139 Jakub Jelinek <jakub@redhat.com>
2142 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2143 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2144 Sync some changes from cxx_fold_indirect_ref.
2146 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2148 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2150 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2151 (BLOCK_INLINE_ENTRY_LABEL): New.
2152 (dwarf2out_var_location): Disregard inline entry markers.
2153 (inline_entry_data): New struct.
2154 (inline_entry_data_hasher): New hashtable type.
2155 (inline_entry_data_hasher::hash): New.
2156 (inline_entry_data_hasher::equal): New.
2157 (inline_entry_data_table): New variable.
2158 (add_high_low_attributes): Add DW_AT_entry_pc and
2159 DW_AT_GNU_entry_view attributes if a pending entry is found
2160 in inline_entry_data_table. Add old entry_pc attribute only
2161 if debug nonbinding markers are disabled.
2162 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2163 markers are enabled.
2164 (block_within_block_p, dwarf2out_inline_entry): New.
2165 (dwarf2out_finish): Check that no entries remained in
2166 inline_entry_data_table.
2167 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2168 (final_scan_insn, notice_source_line): Likewise.
2169 (rest_of_clean_state): Skip inline entry markers.
2170 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2172 * gimple.c (gimple_build_debug_inline_entry): New.
2173 * gimple.h (enum gimple_debug_subcode): Add
2174 GIMPLE_DEBUG_INLINE_ENTRY.
2175 (gimple_build_debug_inline_entry): Declare.
2176 (gimple_debug_inline_entry_p): New.
2177 (gimple_debug_nonbind_marker_p): Adjust.
2178 * insn-notes.def (INLINE_ENTRY): New.
2179 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2180 inline entry marker notes.
2181 (print_insn): Likewise.
2182 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2183 (INSN_DEBUG_MARKER_KIND): Likewise.
2184 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2185 * tree-inline.c (expand_call_inline): Build and insert
2186 debug_inline_entry stmt.
2187 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2188 inline entry blocks early, if nonbind markers are enabled.
2189 (dump_scope_block): Dump fragment info.
2190 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2191 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2192 (gimple_build_debug_inline_entry): New.
2193 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2194 Enable/disable inline entry points too.
2195 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2196 (DEBUG_INSN): Describe inline entry markers.
2198 * common.opt (gvariable-location-views): New.
2199 (gvariable-location-views=incompat5): New.
2200 * config.in: Rebuilt.
2201 * configure: Rebuilt.
2202 * configure.ac: Test assembler for view support.
2203 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2204 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2205 * dwarf2out.c (var_loc_view): New typedef.
2206 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2207 (dwarf2out_locviews_in_attribute): New.
2208 (dwarf2out_locviews_in_loclist): New.
2209 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2210 (enum dw_line_info_opcode): Add LI_adv_address.
2211 (struct dw_line_info_table): Add view.
2212 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2213 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2214 (zero_view_p): New variable.
2215 (ZERO_VIEW_P): New macro.
2216 (output_asm_line_debug_info): New.
2217 (struct var_loc_node): Add view.
2218 (add_AT_view_list, AT_loc_list): New.
2219 (add_var_loc_to_decl): Add view param. Test it against last.
2220 (new_loc_list): Add view params. Record them.
2221 (AT_loc_list_ptr): Handle loc and view lists.
2222 (view_list_to_loc_list_val_node): New.
2223 (print_dw_val): Handle dw_val_class_view_list.
2224 (size_of_die): Likewise.
2225 (value_format): Likewise.
2226 (loc_list_has_views): New.
2227 (gen_llsym): Set vl_symbol too.
2228 (maybe_gen_llsym, skip_loc_list_entry): New.
2229 (dwarf2out_maybe_output_loclist_view_pair): New.
2230 (output_loc_list): Output view list or entries too.
2231 (output_view_list_offset): New.
2232 (output_die): Handle dw_val_class_view_list.
2233 (output_dwarf_version): New.
2234 (output_compilation_unit_header): Use it.
2235 (output_skeleton_debug_sections): Likewise.
2236 (output_rnglists, output_line_info): Likewise.
2237 (output_pubnames, output_aranges): Update version comments.
2238 (output_one_line_info_table): Output view numbers in asm comments.
2239 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2240 Call maybe_gen_llsym.
2241 (loc_list_from_tree_1): Adjust.
2242 (add_AT_location_description): Create view list attribute if
2243 needed, check it's absent otherwise.
2244 (convert_cfa_to_fb_loc_list): Adjust.
2245 (maybe_emit_file): Call output_asm_line_debug_info for test.
2246 (dwarf2out_var_location): Reset views as needed. Precompute
2247 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2248 attribute. Set view.
2249 (new_line_info_table): Reset next view.
2250 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2251 (dwarf2out_source_line): Likewise. Output view resets and labels to
2252 the assembler, or select appropriate line info opcodes.
2253 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2254 (optimize_string_length): Catch it. Adjust.
2255 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2256 dw_val_class_view_list, and remove it if no longer needed.
2257 (hash_loc_list): Hash view numbers.
2258 (loc_list_hasher::equal): Compare them.
2259 (optimize_location_lists): Check whether a view list symbol is
2260 needed, and whether the locview attribute is present, and
2261 whether they match. Remove the locview attribute if no longer
2263 (index_location_lists): Call skip_loc_list_entry for test.
2264 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2265 Use output_dwarf_version.
2266 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2267 (struct dw_val_node): Add val_view_list.
2268 * final.c (SEEN_NEXT_VIEW): New.
2269 (set_next_view_needed): New.
2270 (clear_next_view_needed): New.
2271 (maybe_output_next_view): New.
2272 (final_start_function): Rename to...
2273 (final_start_function_1): ... this. Take pointer to FIRST,
2274 add SEEN parameter. Emit param bindings in the initial view.
2275 (final_start_function): Reintroduce SEEN-less interface.
2276 (final): Rename to...
2277 (final_1): ... this. Take SEEN parameter. Output final pending
2278 next view at the end.
2279 (final): Reintroduce seen-less interface.
2280 (final_scan_insn): Output pending next view before switching
2281 sections or ending a block. Mark the next view as needed when
2282 outputting variable locations. Notify debug backend of section
2283 changes, and of location view changes.
2284 (rest_of_handle_final): Adjust.
2285 * toplev.c (process_options): Autodetect value for debug variable
2286 location views option. Warn on incompat5 without -gdwarf-5.
2287 * doc/invoke.texi (gvariable-location-views): New.
2288 (gvariable-location-views=incompat5): New.
2289 (gno-variable-location-views): New.
2291 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2293 PR tree-optimization/84136
2294 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2295 that the result of find_edge is non-NULL.
2297 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2300 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2301 storing integer register in SImode. Fix cost of 256 and 512
2302 byte aligned SSE register store.
2304 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2306 * config/i386/i386.c (ix86_multiplication_cost): Fix
2307 multiplication cost for TARGET_AVX512DQ.
2309 2018-02-08 Marek Polacek <polacek@redhat.com>
2311 PR tree-optimization/84238
2312 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2315 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2317 PR tree-optimization/84265
2318 * tree-vect-stmts.c (vectorizable_store): Don't treat
2319 VMAT_CONTIGUOUS accesses as grouped.
2320 (vectorizable_load): Likewise.
2322 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2324 PR tree-optimization/81635
2325 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2326 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2327 (test_round_for_mask): New functions.
2328 (wide_int_cc_tests): Call test_round_for_mask.
2329 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2330 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2331 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2332 range returned by get_range_info.
2334 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2337 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2338 * symtab.c: Include builtins.h
2339 (symtab_node::output_to_lto_symbol_table_p): Move here
2340 from lto-streamer-out.c:output_symbol_p.
2341 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2342 (output_symbol_p): Move all logic to symtab.c
2343 (produce_symtab): Update.
2345 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2347 * config/s390/s390-opts.h (enum indirect_branch): Define.
2348 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2349 (s390_indirect_branch_via_thunk)
2350 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2351 (enum s390_indirect_branch_type): Define.
2352 * config/s390/s390.c (struct s390_frame_layout, struct
2353 machine_function): Remove.
2354 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2355 (indirect_branch_table_label_no, indirect_branch_table_name):
2357 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2358 (enum s390_indirect_branch_option): Define.
2359 (s390_return_addr_from_memory): New function.
2360 (s390_handle_string_attribute): New function.
2361 (s390_attribute_table): Add new attribute handler.
2362 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2363 (s390_indirect_branch_via_thunk): New function.
2364 (s390_indirect_branch_via_inline_thunk): New function.
2365 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2366 sibling call optimization for non z10 compiles.
2367 (s390_emit_call): Force indirect branch target to be a single
2368 register. Add r1 clobber for non-z10 compiles.
2369 (s390_emit_epilogue): Emit return jump via return_use expander.
2370 (s390_reorg): Handle JUMP_INSNs as execute targets.
2371 (s390_option_override_internal): Perform validity checks for the
2372 new command line options.
2373 (s390_indirect_branch_attrvalue): New function.
2374 (s390_indirect_branch_settings): New function.
2375 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2376 (s390_output_indirect_thunk_function): New function.
2377 (s390_code_end): Implement target hook.
2378 (s390_case_values_threshold): Implement target hook.
2379 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2381 * config/s390/s390.h (struct s390_frame_layout)
2382 (struct machine_function): Move here from s390.c.
2383 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2384 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2385 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2386 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2387 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2388 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2389 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2390 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2391 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2392 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2393 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2394 (mnemonic attribute): Add values which aren't recognized
2396 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2397 pattern for branch conversion. Fix mnemonic attribute.
2398 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2399 indirect branch via thunk if requested.
2400 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2401 ("*indirect_jump"): Disable for branch conversion using out of
2403 ("indirect_jump_via_thunk<mode>_z10")
2404 ("indirect_jump_via_thunk<mode>")
2405 ("indirect_jump_via_inlinethunk<mode>_z10")
2406 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2407 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2408 ("casesi_jump_via_inlinethunk<mode>_z10")
2409 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2410 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2411 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2412 ("*indirect2_jump"): Disable for branch conversion.
2413 ("casesi_jump"): Turn into expander and expand patterns for branch
2415 ("return_use"): New expander.
2416 ("*return"): Emit return via thunk and rename it to ...
2417 ("*return<mode>"): ... this one.
2418 * config/s390/s390.opt: Add new options and and enum for the
2421 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2423 * lra-constraints.c (match_reload): Unconditionally use
2424 gen_lowpart_SUBREG, rather than selecting between that
2425 and equivalent gen_rtx_SUBREG code.
2427 2018-02-08 Richard Biener <rguenther@suse.de>
2429 PR tree-optimization/84233
2430 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2431 changed flag instead of boguously re-using phi_inserted.
2433 2018-02-08 Martin Jambor <mjambor@suse.cz>
2435 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2436 static local variables.
2438 2018-02-08 Richard Biener <rguenther@suse.de>
2440 PR tree-optimization/84278
2441 * tree-vect-stmts.c (vectorizable_store): When looking for
2442 smaller vector types to perform grouped strided loads/stores
2443 make sure the mode is supported by the target.
2444 (vectorizable_load): Likewise.
2446 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2448 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2449 Increase LDP/STP opportunities by adding adjacent callee-saves.
2451 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2453 PR rtl-optimization/84068
2454 PR rtl-optimization/83459
2455 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2457 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2459 PR tree-optimization/84224
2460 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2461 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2464 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2467 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2468 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2469 count, remove one USE.
2471 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2473 * doc/install.texi (Configuration): Document the
2474 --with-long-double-format={ibm,ieee} PowerPC configuration
2478 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2479 Convert from define_expand to be define_insn_and_split. Rework
2480 float/double/_Float128 conversions to QI/HI/SImode to work with
2481 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2482 conversions to QI/HImode types did a store and then a load to
2483 truncate the value. For conversions to VSX registers, don't split
2484 the insn, instead emit the code directly. Use the code iterator
2485 any_fix to combine signed and unsigned conversions.
2486 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2487 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2488 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2489 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2490 (fix_<mode>di2_hw): Likewise.
2491 (fixuns_<mode>di2_hw): Likewise.
2492 (fix_<mode>si2_hw): Likewise.
2493 (fixuns_<mode>si2_hw): Likewise.
2494 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2495 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2496 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2497 fix<uns>_trunc<SFDF:mode>si2_p8.
2498 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2500 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2501 (fix<uns>_<mode>_mem): Likewise.
2502 (fctiw<u>z_<mode>_mem): Likewise.
2503 (fix<uns>_<mode>_mem): Likewise.
2504 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2505 the register allocator from doing a direct move to the GPRs to do
2506 a store, and instead use the ISA 3.0 store byte/half-word from
2507 vector register instruction. For IEEE 128-bit floating point,
2508 also optimize stores of 32-bit ints.
2509 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2511 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2513 * genextract.c (push_pathstr_operand): New function to support
2515 (walk_rtx): Call push_pathstr_operand.
2516 (print_path): Support [a-zA-Z].
2518 2018-02-07 Richard Biener <rguenther@suse.de>
2520 PR tree-optimization/84037
2521 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2522 (cse_and_gimplify_to_preheader): Declare.
2523 (vect_get_place_in_interleaving_chain): Likewise.
2524 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2526 (_loop_vec_info::~_loop_vec_info): Delete it.
2527 (cse_and_gimplify_to_preheader): New function.
2528 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2529 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2530 (vectorizable_load): Likewise. For grouped stores always base
2531 the IV on the first element.
2532 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2533 condition before gimplifying.
2535 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2537 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2538 *DIV_EXPR and *MOD_EXPR.
2540 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2543 * config/i386/i386.c (ix86_option_override_internal): Mask out
2544 the CF_SET bit when checking -fcf-protection.
2546 2018-02-07 Tom de Vries <tom@codesourcery.com>
2549 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2552 2018-02-07 Richard Biener <rguenther@suse.de>
2554 PR tree-optimization/84204
2555 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2558 PR tree-optimization/84205
2559 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2560 special-case isl_ast_op_zdiv_r.
2562 PR tree-optimization/84223
2563 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2564 Only add conditions from within the region.
2565 (gather_bbs::after_dom_children): Adjust.
2567 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2570 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2571 * config/avr/avr.md: Only post-reload split REG-REG moves if
2572 either register is GENERAL_REG_P.
2574 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2576 PR tree-optimization/84235
2577 * tree-ssa-scopedtables.c
2578 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2579 if the subtraction is performed in floating point type where NaNs are
2580 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2581 build 1. Formatting fix.
2583 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2586 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2587 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2588 and skip it regardless of bb boundaries. Use CALL_P macro,
2589 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2590 unnecessarily, formatting fix.
2592 2018-02-06 Michael Collison <michael.collison@arm.com>
2594 * config/arm/thumb2.md:
2595 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2596 (*thumb_mov_notscc): Ditto.
2598 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2601 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2602 unsigned_fix, not "s".
2604 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2606 * configure.ac (gcc_fn_eh_frame_ro): New function.
2607 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2608 correct .eh_frame permissions.
2609 * configure: Regenerate.
2611 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2613 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2616 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2618 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2619 Display warning message for -mno-speculate-indirect-jumps.
2621 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2623 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2625 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2627 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2629 PR tree-optimization/84225
2630 * tree-eh.c (find_trapping_overflow): Only call
2631 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2633 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2636 * config/i386/i386.c: Reimplement the check of possible options
2637 -mibt/-mshstk conbination. Change error messages.
2638 * doc/invoke.texi: Fix a typo: remove extra '='.
2640 2018-02-06 Marek Polacek <polacek@redhat.com>
2642 PR tree-optimization/84228
2643 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2645 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2648 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2649 emitted arch directives.
2650 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2651 __ARM_FEATURE_COPROC before changing architectures.
2653 2018-02-06 Richard Biener <rguenther@suse.de>
2655 * config/i386/i386.c (print_reg): Fix typo.
2656 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2658 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2660 * configure: Regenerate.
2662 2018-02-05 Martin Sebor <msebor@redhat.com>
2664 PR tree-optimization/83369
2665 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2668 2018-02-05 Martin Liska <mliska@suse.cz>
2670 * doc/invoke.texi: Cherry-pick upstream r323995.
2672 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2674 * ira.c (ira_init_register_move_cost): Adjust comment.
2676 2018-02-05 Martin Liska <mliska@suse.cz>
2678 PR gcov-profile/84137
2679 * doc/gcov.texi: Fix typo in documentation.
2681 2018-02-05 Martin Liska <mliska@suse.cz>
2683 PR gcov-profile/83879
2684 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2685 using dlopen functionality.
2687 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2689 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2690 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2691 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2692 _mm_maskz_range_round_ss): New intrinsics.
2693 (__builtin_ia32_rangesd128_round)
2694 (__builtin_ia32_rangess128_round): Remove.
2695 (__builtin_ia32_rangesd128_mask_round,
2696 __builtin_ia32_rangess128_mask_round): New builtins.
2697 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2698 __builtin_ia32_rangess128_round): Remove.
2699 (__builtin_ia32_rangesd128_mask_round,
2700 __builtin_ia32_rangess128_mask_round): New builtins.
2701 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2702 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2703 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2704 "<round_saeonly_constraint>")): Changed to ...
2705 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2706 "<round_saeonly_scalar_constraint>")): ... this.
2707 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2708 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2709 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2710 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2711 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2713 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2715 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2717 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2718 Remove all values except native, 8540 and 8548.
2720 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2722 * config/i386/i386.c (ix86_output_function_return): Pass
2723 INVALID_REGNUM, instead of -1, as invalid register number to
2724 indirect_thunk_name and output_indirect_thunk.
2726 2018-02-02 Julia Koval <julia.koval@intel.com>
2728 * config.gcc: Add -march=icelake.
2729 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2730 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2731 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2732 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2733 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2734 (processor_target_table): Add icelake.
2735 (ix86_option_override_internal): Handle new PTAs.
2736 (get_builtin_code_for_version): Handle icelake.
2737 (M_INTEL_COREI7_ICELAKE): New.
2738 (fold_builtin_cpu): Handle icelake.
2739 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2740 * doc/invoke.texi: Add -march=icelake.
2742 2018-02-02 Julia Koval <julia.koval@intel.com>
2744 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2745 to wide_int_bitmask.
2746 * wide-int-bitmask.h: New.
2748 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2751 * config/i386/i386.md: Replace Pmode with word_mode in
2752 builtin_setjmp_setup and builtin_longjmp to support x32.
2754 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2758 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2760 (rs6000_supported_cpu_names): New static variable.
2761 (linux_cpu_translation_table): Likewise.
2762 (elf_platform) <cpu>: Define new static variable and use it.
2763 Translate kernel AT_PLATFORM name to canonical name if needed.
2764 Error if platform name is unknown.
2766 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2769 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2771 2018-02-01 Jeff Law <law@redhat.com>
2774 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2775 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2776 the scratch if RELEASE_VIA_POP is false.
2777 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2778 If we have to save a temporary register, decrement SIZE appropriately.
2779 Pass new arguments to release_scratch_register_on_entry.
2780 (ix86_adjust_stack_and_probe): Likewise.
2781 (ix86_emit_probe_stack_range): Pass new arguments to
2782 release_scratch_register_on_entry.
2784 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2786 PR rtl-optimization/84157
2787 * combine.c (change_zero_ext): Use REG_P predicate in
2788 front of HARD_REGISTER_P predicate.
2790 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2792 * config/avr/avr.c (avr_option_override): Move disabling of
2793 -fdelete-null-pointer-checks to...
2794 * common/config/avr/avr-common.c (avr_option_optimization_table):
2797 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2799 PR tree-optimization/81635
2800 * tree-data-ref.c (split_constant_offset_1): For types that
2801 wrap on overflow, try to use range info to prove that wrapping
2804 2018-02-01 Renlin Li <renlin.li@arm.com>
2807 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2809 (aarch64_register_move_cost): Likewise.
2810 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2812 (REG_CLASS_NAMES): Likewise.
2813 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2814 TAILCALL_ADDR_REGS. Remove IP registers.
2815 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2817 2018-02-01 Richard Biener <rguenther@suse.de>
2819 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2820 for specifying RPO order and allow NULL for that.
2821 * domwalk.c (dom_walker::dom_walker): Likewise.
2822 (dom_walker::walk): Handle NULL RPO order.
2823 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2825 (rewrite_update_dom_walker): Likewise.
2826 (mark_def_dom_walker): Likewise.
2828 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2830 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2831 (aarch64_maybe_expand_sve_subreg_move): Declare.
2832 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2833 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2835 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2836 that are semantically a reverse operation.
2837 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2838 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2839 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2841 (aarch64_can_change_mode_class): For big-endian, forbid changes
2842 between two SVE modes if they have different element sizes.
2844 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2846 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2847 the TImode handling for big-endian targets.
2849 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2851 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2852 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2854 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2855 Remove BSWAP handing for big-endian targets and use the form of
2856 LD1RQ appropariate for the mode.
2858 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2860 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2861 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2864 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2867 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2868 check for operands that need to go through aarch64_sve_reload_be.
2870 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2872 PR tree-optimization/81661
2873 PR tree-optimization/84117
2874 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2875 * tree-eh.c: Include gimplify.h.
2876 (find_trapping_overflow, replace_trapping_overflow,
2877 rewrite_to_non_trapping_overflow): New functions.
2878 * tree-vect-loop.c: Include tree-eh.h.
2879 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2880 * tree-data-ref.c: Include tree-eh.h.
2881 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2883 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2885 PR rtl-optimization/84123
2886 * combine.c (change_zero_ext): Check if hard register satisfies
2887 can_change_dest_mode before calling gen_lowpart_SUBREG.
2889 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2892 * ira.c (ira_init_register_move_cost): Remove assert.
2894 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2896 PR rtl-optimization/84071
2897 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2898 * doc/tm.texi: Regenerate.
2900 2018-01-31 Richard Biener <rguenther@suse.de>
2902 PR tree-optimization/84132
2903 * tree-data-ref.c (analyze_miv_subscript): Properly
2904 check whether evolution_function_is_affine_multivariate_p
2905 before calling gcd_of_steps_may_divide_p.
2907 2018-01-31 Julia Koval <julia.koval@intel.com>
2910 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2911 * config/i386/i386.md (rdpid_rex64) New.
2912 (rdpid): Make 32bit only.
2914 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2917 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2918 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2920 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2923 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2925 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2927 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2929 PR rtl-optimization/84071
2930 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2931 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2933 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2935 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2936 (arc_attribute_table): Add 'aux' attribute.
2937 (arc_in_small_data_p): Consider aux like variables.
2938 (arc_is_aux_reg_p): New function.
2939 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2940 (arc_get_aux_arg): New function.
2941 (prepare_move_operands): Handle aux-register access.
2942 (arc_handle_aux_attribute): New function.
2943 * doc/extend.texi (ARC Variable attributes): Add subsection.
2945 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2947 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2948 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2949 (arc_attribute_table): Add 'uncached' attribute.
2950 (arc_print_operand): Print '.di' flag for uncached memory
2952 (arc_in_small_data_p): Do not consider for small data the uncached
2954 (arc_is_uncached_mem_p): New function.
2955 * config/arc/predicates.md (compact_store_memory_operand): Check
2956 for uncached memory accesses.
2957 (nonvol_nonimm_operand): Likewise.
2958 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2960 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2963 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2964 falign-loops=): Add Optimization flag.
2966 2018-01-30 Jeff Law <law@redhat.com>
2969 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2970 INT_REGISTERS_SAVED. Check it prior to calling
2971 get_scratch_register_on_entry.
2972 (ix86_adjust_stack_and_probe): Similarly.
2973 (ix86_emit_probe_stack_range): Similarly.
2974 (ix86_expand_prologue): Corresponding changes.
2976 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2979 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2980 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2982 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2985 * lra-constraints.c (curr_insn_transform): Process AND in the
2988 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2990 PR rtl-optimization/83986
2991 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2992 dependence against last_pending_memory_flush in addition to
2995 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2997 PR tree-optimization/81611
2998 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3001 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3004 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3007 2018-01-30 Richard Biener <rguenther@suse.de>
3008 Jakub Jelinek <jakub@redhat.com>
3010 PR tree-optimization/84111
3011 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3012 inner loops added during recursion, as they don't have up-to-date
3015 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3018 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3019 (can_inline_edge_by_limits_p): ... here.
3020 (can_early_inline_edge_p, check_callers,
3021 update_caller_keys, update_callee_keys, recursive_inlining,
3022 add_new_edges_to_heap, speculation_useful_p,
3023 inline_small_functions,
3024 inline_small_functions, flatten_function,
3025 inline_to_all_callers_1): Update.
3027 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3029 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3032 2018-01-30 Richard Biener <rguenther@suse.de>
3034 PR tree-optimization/83008
3035 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3036 invariant and constant vector uses in stmts when they need
3039 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3042 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3043 * configure: Regenerate.
3045 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3047 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3049 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3050 Use gen_rtx_REG rather than gen_lowpart.
3052 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3054 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3055 rather than 0 when creating partial subregs.
3057 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3059 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3062 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3065 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3066 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3067 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3068 flags. This restores the settings used before the 2017-07-24.
3069 Turning off pre increment/decrement/modify allows IVOPTS to
3070 optimize DF/SF loops where the index is an int.
3072 2018-01-29 Richard Biener <rguenther@suse.de>
3073 Kelvin Nilsen <kelvin@gcc.gnu.org>
3076 * tree-vect-stmts.c (vectorizable_call): Don't call
3077 targetm.vectorize_builtin_md_vectorized_function if callee is
3080 2018-01-22 Carl Love <cel@us.ibm.com>
3082 * doc/extend.tex: Fix typo in second arg in
3083 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3085 2018-01-29 Richard Biener <rguenther@suse.de>
3087 PR tree-optimization/84086
3088 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3089 (flush_ssaname_freelist): When SSA names were released reset
3090 the SCEV hash table.
3092 2018-01-29 Richard Biener <rguenther@suse.de>
3094 PR tree-optimization/84057
3095 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3096 removed paths when removing edges.
3098 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3100 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3101 -mfunction-return=@var{choice}.
3103 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3106 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3108 (layout::print_source_line): Likewise.
3109 (test_get_line_width_without_trailing_whitespace): Add test cases.
3111 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3114 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3117 2018-01-26 Jim Wilson <jimw@sifive.com>
3119 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3121 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3124 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3126 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3127 and CMP + SUB-immediate -> SUBS.
3129 2018-01-26 Martin Sebor <msebor@redhat.com>
3131 PR tree-optimization/83896
3132 * tree-ssa-strlen.c (get_string_len): Rename...
3133 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3134 Avoid assuming length is constant.
3135 (handle_char_store): Use HOST_WIDE_INT for string length.
3137 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3140 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3141 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3143 2018-01-26 Richard Biener <rguenther@suse.de>
3145 PR rtl-optimization/84003
3146 * dse.c (record_store): Only record redundant stores when
3147 the earlier store aliases at least all accesses the later one does.
3149 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3151 PR rtl-optimization/83985
3152 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3153 REG_CFA_RESTORE insns.
3154 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3155 REG_CFA_RESTORE insns here.
3158 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3159 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3161 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3163 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3164 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3165 (arc_init): Likewise.
3166 (arc_override_options): Likewise.
3167 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3169 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3171 * config/arc/arc.h (TARGET_DBNZ): Define.
3172 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3173 properly set the tune attribute.
3174 (dbnz): Use TARGET_DBNZ guard.
3175 * config/arc/arc.opt (mtune): Add core3 option.
3177 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3179 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3180 recognize new pic like addresses.
3181 (arc_delegitimize_address): Clean up.
3183 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3185 * config/arc/arc-arches.def: Option mrf16 valid for all
3187 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3188 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3189 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3190 * config/arc/arc-tables.opt: Regenerate.
3191 * config/arc/arc.c (arc_conditional_register_usage): Handle
3192 reduced register file case.
3193 (arc_file_start): Set must have build attributes.
3194 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3196 * config/arc/arc.opt (mrf16): Add new option.
3197 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3198 * config/arc/genmultilib.awk: Handle new mrf16 option.
3199 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3200 * config/arc/t-multilib: Regenerate.
3201 * doc/invoke.texi (ARC Options): Document mrf16 option.
3203 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3205 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3206 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3207 (arc_attribute_table): Add 'secure_call' attribute.
3208 (arc_print_operand): Print secure call operand.
3209 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3211 (arc_is_secure_call_p): New function. * config/arc/arc.md
3212 (call_i): Add support for sjli instruction.
3213 (call_value_i): Likewise.
3214 * config/arc/constraints.md (Csc): New constraint.
3216 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3217 John Eric Martin <John.Martin@emmicro-us.com>
3219 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3220 * config/arc/arc.c (_arc_jli_section): New struct.
3221 (arc_jli_section): New type.
3222 (rc_jli_sections): New static variable.
3223 (arc_handle_jli_attribute): New function.
3224 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3225 (arc_file_end): New function.
3226 (TARGET_ASM_FILE_END): Define.
3227 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3228 (arc_add_jli_section): New function.
3229 (jli_call_scan): Likewise.
3230 (arc_reorg): Call jli_call_scan.
3231 (arc_output_addsi): Remove 'S' from printing asm operand.
3232 (arc_is_jli_call_p): New function.
3233 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3235 (movhi_insn): Likewise.
3236 (movsi_insn): Likewise.
3237 (movsi_set_cc_insn): Likewise.
3238 (loadqi_update): Likewise.
3239 (load_zeroextendqisi_update): Likewise.
3240 (load_signextendqisi_update): Likewise.
3241 (loadhi_update): Likewise.
3242 (load_zeroextendhisi_update): Likewise.
3243 (load_signextendhisi_update): Likewise.
3244 (loadsi_update): Likewise.
3245 (loadsf_update): Likewise.
3246 (movsicc_insn): Likewise.
3247 (bset_insn): Likewise.
3248 (bxor_insn): Likewise.
3249 (bclr_insn): Likewise.
3250 (bmsk_insn): Likewise.
3251 (bicsi3_insn): Likewise.
3252 (cmpsi_cc_c_insn): Likewise.
3253 (movsi_ne): Likewise.
3254 (movsi_cond_exec): Likewise.
3255 (clrsbsi2): Likewise.
3266 (call_i): Remove 'S' asm letter, add jli instruction.
3267 (call_value_i): Likewise.
3268 * config/arc/arc.op (mjli-always): New option.
3269 * config/arc/constraints.md (Cji): New constraint.
3270 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3272 (subsf3_fpx): Likewise.
3273 (mulsf3_fpx): Likewise.
3274 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3276 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3278 * doc/invoke.texi (ARC): Document mjli-always option.
3280 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3282 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3283 avoid addition with 0 and use incw and decw where possible.
3285 2018-01-26 Richard Biener <rguenther@suse.de>
3287 PR tree-optimization/81082
3288 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3289 association if it requires casting to unsigned.
3290 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3291 from fold_plusminus_mult_expr to catch important cases late when
3292 range info is available.
3294 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3296 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3297 * configure.ac (hidden_linkonce): New test.
3298 * configure: Regenerate.
3299 * config.in: Regenerate.
3301 2018-01-26 Julia Koval <julia.koval@intel.com>
3303 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3304 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3305 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3306 _mm_mask_bitshuffle_epi64_mask): Fix type.
3307 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3308 USI_FTYPE_V4DI_V4DI_USI): Remove.
3309 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3310 __builtin_ia32_vpshufbitqmb256_mask,
3311 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3312 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3313 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3315 2018-01-26 Alan Modra <amodra@gmail.com>
3318 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3319 UNSPEC_VBPERMQ. Sort other unspecs.
3321 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3323 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3325 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3328 * predict.c (drop_profile): Do not push/pop cfun; update also
3330 (handle_missing_profiles): Fix logic looking for zero profiles.
3332 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3335 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3336 on functions with #pragma omp declare simd or functions with simd
3338 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3339 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3340 Remove trailing \n from warning_at calls.
3342 2018-01-25 Tom de Vries <tom@codesourcery.com>
3345 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3346 for neutered workers.
3348 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3351 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3352 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3354 2018-01-24 Jeff Law <law@redhat.com>
3357 * i386.c (get_probe_interval): Move to earlier point.
3358 (ix86_compute_frame_layout): If -fstack-clash-protection and
3359 the frame is larger than the probe interval, then use pushes
3360 to save registers rather than reg->mem moves.
3361 (ix86_expand_prologue): Remove conditional for int_registers_saved
3364 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3367 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3368 min/max for never referenced object.
3370 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3373 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3375 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3376 attributes from DECL_ATTRIBUTES (decl) without affecting
3377 DECL_ATTRIBUTES (current_function_decl).
3378 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3379 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3381 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3383 PR tree-optimization/83979
3384 * fold-const.c (fold_comparison): Use constant_boolean_node
3385 instead of boolean_{true,false}_node.
3387 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3389 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3392 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3394 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3395 Simplify the clause that sets the length attribute.
3396 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3397 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3398 clause that sets the length attribute.
3399 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3401 2018-01-24 Tom de Vries <tom@codesourcery.com>
3404 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3405 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3406 Add strict parameter.
3407 (prevent_branch_around_nothing): Insert dummy insn between branch to
3408 label and label with no ptx insn inbetween.
3409 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3411 2018-01-24 Tom de Vries <tom@codesourcery.com>
3414 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3415 for neutered threads in warp.
3416 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3418 2018-01-24 Richard Biener <rguenther@suse.de>
3420 PR tree-optimization/83176
3421 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3424 2018-01-24 Richard Biener <rguenther@suse.de>
3426 PR tree-optimization/82819
3427 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3428 code generating pluses that are no-ops in the target precision.
3430 2018-01-24 Richard Biener <rguenther@suse.de>
3433 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3435 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3437 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3438 to merge probabilities.
3439 * predict.c (probably_never_executed): Also mark as cold functions
3440 with global 0 profile and guessed local profile.
3441 * profile-count.c (profile_probability::combine_with_count): New
3443 * profile-count.h (profile_probability::operator*,
3444 profile_probability::operator*=, profile_probability::operator/,
3445 profile_probability::operator/=): Reduce precision to adjusted
3446 and set value to guessed on contradictory divisions.
3447 (profile_probability::combine_with_freq): Remove.
3448 (profile_probability::combine_wiht_count): Declare.
3449 (profile_count::force_nonzero):: Set to adjusted.
3450 (profile_count::probability_in):: Set quality to adjusted.
3451 * tree-ssa-tail-merge.c (replace_block_by): Use
3454 2018-01-23 Andrew Waterman <andrew@sifive.com>
3455 Jim Wilson <jimw@sifive.com>
3457 * config/riscv/riscv.c (riscv_stack_boundary): New.
3458 (riscv_option_override): Set riscv_stack_boundary. Handle
3459 riscv_preferred_stack_boundary_arg.
3460 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3461 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3462 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3463 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3464 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3465 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3467 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3470 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3471 of struct ix86_frame.
3472 (ix86_expand_epilogue): Likewise. Add a local variable for
3473 the reg_save_offset field in struct ix86_frame.
3475 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3477 PR tree-optimization/82604
3478 * tree-loop-distribution.c (enum partition_kind): New enum item
3479 PKIND_PARTIAL_MEMSET.
3480 (partition_builtin_p): Support above new enum item.
3481 (generate_code_for_partition): Ditto.
3482 (compute_access_range): Differentiate cases that equality can be
3483 proven at all loops, the innermost loops or no loops.
3484 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3485 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3486 (finalize_partitions, distribute_loop): Don't fuse partition of
3487 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3488 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3491 2018-01-23 Martin Liska <mliska@suse.cz>
3493 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3494 order to ignore the predictor.
3495 (PRED_POLYMORPHIC_CALL): Likewise.
3496 (PRED_RECURSIVE_CALL): Likewise.
3498 2018-01-23 Martin Liska <mliska@suse.cz>
3500 * tree-profile.c (tree_profiling): Print function header to
3501 aware reader which function we are working on.
3502 * value-prof.c (gimple_find_values_to_profile): Do not print
3503 not interesting value histograms.
3505 2018-01-23 Martin Liska <mliska@suse.cz>
3507 * profile-count.h (enum profile_quality): Add
3508 profile_uninitialized as the first value. Do not number values
3509 as they are zero based.
3510 (profile_count::verify): Update sanity check.
3511 (profile_probability::verify): Likewise.
3513 2018-01-23 Nathan Sidwell <nathan@acm.org>
3515 * doc/invoke.texi (ffor-scope): Deprecate.
3517 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3519 PR tree-optimization/83510
3520 * domwalk.c (set_all_edges_as_executable): New function.
3521 (dom_walker::dom_walker): Convert bool param
3522 "skip_unreachable_blocks" to enum reachability. Move setup of
3523 edge flags to set_all_edges_as_executable and only do it when
3524 reachability is REACHABLE_BLOCKS.
3525 * domwalk.h (enum dom_walker::reachability): New enum.
3526 (dom_walker::dom_walker): Convert bool param
3527 "skip_unreachable_blocks" to enum reachability.
3528 (set_all_edges_as_executable): New decl.
3529 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3530 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3532 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3533 but converting true to REACHABLE_BLOCKS.
3534 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3536 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3537 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3538 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3540 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3541 if check_all_array_refs will be called.
3543 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3545 * tree.c (selftest::test_location_wrappers): Add more test
3548 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3550 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3551 (selftest::test_bit_in_range): Likewise.
3553 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3556 * doc/sourcebuild.texi (vect_float): Say that the selector
3557 only describes the situation when -funsafe-math-optimizations is on.
3558 (vect_float_strict): Document.
3560 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3562 PR tree-optimization/83965
3563 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3564 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3565 instead of checking only for a reduction.
3566 (vect_recog_widen_sum_pattern): Likewise.
3568 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3570 * predict.c (probably_never_executed): Only use precise profile info.
3571 (compute_function_frequency): Skip after inlining hack since we now
3572 have quality checking.
3574 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3576 * profile-count.h (profile_probability::very_unlikely,
3577 profile_probability::unlikely, profile_probability::even): Set
3578 precision to guessed.
3580 2018-01-23 Richard Biener <rguenther@suse.de>
3582 PR tree-optimization/83963
3583 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3584 Properly terminate dominator walk when crossing the exit edge not
3585 when visiting its source block.
3587 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3590 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3591 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3593 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3595 PR tree-optimization/83957
3596 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3597 semicolon after for body surrounded by braces.
3599 PR tree-optimization/83081
3600 * profile-count.h (profile_probability::split): New method.
3601 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3602 Use profile_probability::split.
3603 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3604 when splitting a single conditional jump into 2.
3606 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3608 PR tree-optimization/69452
3609 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3612 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3614 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3615 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3616 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3618 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3620 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3621 * config/rl78/rl78.md (movdi): New define_expand.
3622 * config/rl78/rl78.c (rl78_split_movdi): New function.
3624 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3627 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3629 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3630 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3631 128-bit to produce an UNSPEC move to get the double word with the
3632 signbit and then a shift directly to do signbit.
3633 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3634 implementation with a new version that just does either a direct
3635 move or a regular move. Move memory interface to separate insns.
3636 Move insns so they are next to the expander.
3637 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3638 with signbit move. Split big and little endian case.
3639 (signbit<mode>2_dm_mem_le): Likewise.
3640 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3641 (signbit<mode>2_dm2): Likewise.
3643 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3645 * config/rl78/rl78.md (anddi3): New define_expand.
3647 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3649 * config/rl78/rl78.md (umindi3): New define_expand.
3651 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3653 * config/rl78/rl78.md (smindi3): New define_expand.
3655 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3657 * config/rl78/rl78.md (smaxdi3): New define_expand.
3659 2018-01-22 Carl Love <cel@us.ibm.com>
3661 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3662 LVX_V1TI): Add macro expansion.
3663 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3664 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3665 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3666 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3667 Change check to determine if the instruction is a byte reversing
3668 entry. Fix typo in comment.
3669 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3670 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3671 Add def_builtin calls for new builtins.
3672 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3673 Add define_insn expansion.
3675 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3677 * config/rl78/rl78.md (umaxdi3): New define_expand.
3679 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3681 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3682 for non-QImode registers.
3684 2018-01-22 Richard Biener <rguenther@suse.de>
3686 PR tree-optimization/83963
3687 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3688 including the loop exit block.
3689 (scop_detection::merge_sese): Likewise.
3690 (scop_detection::add_scop): Do it here instead.
3692 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3694 * doc/sourcebuild.texi (arm_softfloat): Document.
3696 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3699 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3700 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3701 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3703 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3704 David Edelsohn <dje.gcc@gmail.com>
3707 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3708 Change "crset eq" to "crset 2".
3709 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3710 (*call_indirect_aix<mode>_nospec): Likewise.
3711 (*call_value_indirect_aix<mode>_nospec): Likewise.
3712 (*call_indirect_elfv2<mode>_nospec): Likewise.
3713 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3714 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3715 change assembly output from . to $.
3716 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3717 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3718 (*tablejump<mode>_internal1_nospec): Likewise.
3720 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3723 * config/sh/sh_optimize_sett_clrt.cc:
3724 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3726 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3728 PR tree-optimization/83940
3729 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3730 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3731 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3732 use it to pass back the definition type.
3733 (vect_check_store_rhs): Likewise rhs_dt_out.
3734 (vect_build_gather_load_calls): Add a mask_dt argument and use
3735 it instead of a call to vect_is_simple_use.
3736 (vectorizable_store): Update calls to vect_check_load_store_mask
3737 and vect_check_store_rhs. Use the dt returned by the latter instead
3738 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3739 instead of calls to vect_is_simple_use. Pass the scalar rather
3740 than the vector operand to vect_is_simple_use when handling
3741 second and subsequent copies of an rhs value.
3742 (vectorizable_load): Update calls to vect_check_load_store_mask
3743 and vect_build_gather_load_calls. Use the cached mask_dt and
3744 gs_info.offset_dt instead of calls to vect_is_simple_use.
3746 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3749 * tree-emutls.c: Include gimplify.h.
3750 (lower_emutls_2): New function.
3751 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3752 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3753 it before further processing.
3756 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3757 UINTVAL (trueop1) instead of INTVAL (op1).
3759 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3763 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3764 INCOMING_FRAME_SP_OFFSET if not defined.
3765 (scan_trace): Add ENTRY argument. If true and
3766 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3767 emit a note to adjust the CFA offset.
3768 (create_cfi_notes): Adjust scan_trace callers.
3769 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3770 INCOMING_FRAME_SP_OFFSET in the CIE.
3771 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3772 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3774 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3775 * doc/tm.texi: Regenerated.
3777 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3779 PR rtl-optimization/83147
3780 * lra-constraints.c (remove_inheritance_pseudos): Use
3781 lra_substitute_pseudo_within_insn.
3783 2018-01-19 Tom de Vries <tom@codesourcery.com>
3784 Cesar Philippidis <cesar@codesourcery.com>
3787 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3789 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3792 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3793 spaces for function labels.
3795 2018-01-19 Martin Liska <mliska@suse.cz>
3797 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3798 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3799 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3800 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3801 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3802 (PRED_CONST_RETURN): Change from 69 to 65.
3803 (PRED_NULL_RETURN): Change from 91 to 71.
3804 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3805 (PRED_LOOP_GUARD): Change from 66 to 73.
3807 2018-01-19 Martin Liska <mliska@suse.cz>
3809 * predict.c (predict_insn_def): Add new assert.
3810 (struct branch_predictor): Change type to signed integer.
3811 (test_prediction_value_range): Amend test to cover
3813 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3814 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3815 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3816 (PRED_LOOP_IV_COMPARE): Likewise.
3817 * predict.h (PROB_UNINITIALIZED): Define new constant.
3819 2018-01-19 Martin Liska <mliska@suse.cz>
3821 * predict.c (dump_prediction): Add new format for
3822 analyze_brprob.py script which is enabled with -details
3824 * profile-count.h (precise_p): New function.
3826 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3828 PR tree-optimization/83922
3829 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3830 there are no statements that need masking.
3831 (vect_active_double_reduction_p): New function.
3832 (vect_analyze_loop_operations): Use it when handling phis that
3833 are not in the loop header.
3835 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3837 PR tree-optimization/83914
3838 * tree-vect-loop.c (vectorizable_induction): Don't convert
3839 init_expr or apply the peeling adjustment for inductions
3840 that are nested within the vectorized loop.
3842 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3844 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3847 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3851 * function.h (gimplify_parameters): Add gimple_seq * argument.
3852 * function.c: Include gimple.h and options.h.
3853 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3854 for the added local temporaries if needed.
3855 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3856 if there are any parameter cleanups, wrap whole body into a
3857 try/finally with the cleanups.
3859 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3862 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3863 Use GET_MODE_CLASS for scalar floating point.
3865 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3869 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3870 Fix call of call_cgraph_insertion_hooks.
3872 2018-01-18 Martin Sebor <msebor@redhat.com>
3874 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3876 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3879 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3882 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3885 * common.opt: (-ffile-prefix-map): New option.
3886 * opts.c (common_handle_option): Defer it.
3887 * opts-global.c (handle_common_deferred_options): Handle it.
3888 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3889 * file-prefix-map.h: New file.
3890 (remap_debug_filename, add_debug_prefix_map): ...here.
3891 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3892 * final.c (debug_prefix_map, add_debug_prefix_map
3893 remap_debug_filename): Move to...
3894 * file-prefix-map.c: New file.
3895 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3896 generalize, get rid of alloca(), use strrchr() instead of strchr().
3897 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3898 Implement in terms of add_prefix_map().
3899 (remap_macro_filename, remap_debug_filename): Implement in term of
3901 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3902 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3903 * dbxout.c: Include file-prefix-map.h.
3904 * varasm.c: Likewise.
3905 * vmsdbgout.c: Likewise.
3906 * xcoffout.c: Likewise.
3907 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3908 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3909 * doc/invoke.texi (-ffile-prefix-map): Document.
3910 (-fdebug-prefix-map): Update description.
3912 2018-01-18 Martin Liska <mliska@suse.cz>
3914 * config/i386/i386.c (indirect_thunk_name): Document that also
3916 (output_indirect_thunk): Document why both instructions
3917 (pause and lfence) are generated.
3919 2018-01-18 Richard Biener <rguenther@suse.de>
3921 PR tree-optimization/83887
3922 * graphite-scop-detection.c
3923 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3924 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3925 (scop_detection::merge_sese): Re-implement with a flood-fill
3926 algorithm that properly finds a SESE region if it exists.
3928 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3931 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3932 pointer_diff optimizations use view_convert instead of convert.
3934 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3936 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3937 Generate different code for -mno-speculate-indirect-jumps.
3938 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3939 (*call_indirect_aix<mode>): Disable for
3940 -mno-speculate-indirect-jumps.
3941 (*call_indirect_aix<mode>_nospec): New define_insn.
3942 (*call_value_indirect_aix<mode>): Disable for
3943 -mno-speculate-indirect-jumps.
3944 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3945 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3946 -mno-speculate-indirect-jumps.
3947 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3949 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3951 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3952 long double type, set the flags for noting the default long double
3953 type, even if we don't pass or return a long double type.
3955 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3958 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3961 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3963 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3964 support for merge[hl].
3965 (fold_mergehl_helper): New helper function.
3966 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3967 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3968 (altivec_vmrglw_direct): Add xxmrglw insn.
3970 2018-01-17 Andrew Waterman <andrew@sifive.com>
3972 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3973 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3975 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3978 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3979 call the lto_location_cache before reading the
3980 DECL_SOURCE_LOCATION of the types.
3982 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3983 Richard Sandiford <richard.sandiford@linaro.org>
3985 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3986 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3987 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3988 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3989 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3991 * config/aarch64/constraints.md (aarch64_movti_operand):
3993 * config/aarch64/predicates.md (Uti): Add new constraint.
3995 2018-01-17 Carl Love <cel@us.ibm.com>
3997 * config/rs6000/vsx.md (define_expand xl_len_r,
3998 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3999 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4001 (define_expand, define_insn): Move the shift left from the
4002 define_insn to the define_expand for lxvl and stxvl instructions.
4003 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4004 and XL_LEN_R definitions to PURE.
4006 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4008 * config/i386/i386.c (indirect_thunk_name): Declare regno
4009 as unsigned int. Compare regno with INVALID_REGNUM.
4010 (output_indirect_thunk): Ditto.
4011 (output_indirect_thunk_function): Ditto.
4012 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4013 in the call to output_indirect_thunk_function.
4015 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4018 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4019 rather than the size of inner_type to determine the stack slot size
4020 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4022 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4025 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4028 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4030 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4031 endian Linux systems to optionally enable multilibs for selecting
4032 the long double type if the user configured an explicit type.
4033 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4034 have no long double multilibs if not defined.
4035 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4036 warn if the user used -mabi={ieee,ibm}longdouble and we built
4037 multilibs for long double.
4038 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4039 appropriate multilib option.
4040 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4042 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4043 for building long double multilibs.
4044 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4046 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4048 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4051 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4053 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4056 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4059 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4062 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4064 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4065 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4067 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4069 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4070 different rtl trees depending on TARGET_64BIT.
4071 (rs6000_gen_lvx): Likewise.
4073 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4075 * config/visium/visium.md (nop): Tweak comment.
4076 (hazard_nop): Likewise.
4078 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4080 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4081 -mspeculate-indirect-jumps.
4082 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4083 for -mno-speculate-indirect-jumps.
4084 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4085 (*call_value_indirect_elfv2<mode>): Disable for
4086 -mno-speculate-indirect-jumps.
4087 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4088 (indirect_jump): Emit different RTL for
4089 -mno-speculate-indirect-jumps.
4090 (*indirect_jump<mode>): Disable for
4091 -mno-speculate-indirect-jumps.
4092 (*indirect_jump<mode>_nospec): New define_insn.
4093 (tablejump): Emit different RTL for
4094 -mno-speculate-indirect-jumps.
4095 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4096 (tablejumpsi_nospec): New define_expand.
4097 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4098 (tablejumpdi_nospec): New define_expand.
4099 (*tablejump<mode>_internal1): Disable for
4100 -mno-speculate-indirect-jumps.
4101 (*tablejump<mode>_internal1_nospec): New define_insn.
4102 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4105 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4107 * caller-save.c (insert_save): Drop unnecessary parameter. All
4110 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4111 Richard Biener <rguenth@suse.de>
4114 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4115 return early, inline manually is_gimple_sizepos. Make sure if we
4116 call gimplify_expr we don't end up with a gimple constant.
4117 * tree.c (variably_modified_type_p): Don't return true for
4118 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4119 * gimplify.h (is_gimple_sizepos): Remove.
4121 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4123 PR tree-optimization/83857
4124 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4125 vectorizable_live_operation for pure SLP statements.
4126 (vectorizable_live_operation): Handle PHIs.
4128 2018-01-16 Richard Biener <rguenther@suse.de>
4130 PR tree-optimization/83867
4131 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4132 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4134 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4137 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4138 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4139 If off is not INTEGER_CST, issue a may not be aligned warning
4140 rather than isn't aligned. Use isn%'t rather than isn't.
4141 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4143 <case MULT_EXPR>: Improve the case when bottom and one of the
4144 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4145 operand, in that case check if the other operand is multiple of
4146 bottom divided by the INTEGER_CST operand.
4148 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4151 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4152 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4153 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4154 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4155 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4156 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4157 (pa_function_arg_size): New function.
4159 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4161 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4162 in a separate statement.
4164 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4166 PR tree-optimization/83847
4167 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4168 group gathers and scatters.
4170 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4172 PR rtl-optimization/86620
4173 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4175 PR rtl-optimization/83213
4176 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4177 to last if both are JUMP_INSNs.
4179 PR tree-optimization/83843
4180 * gimple-ssa-store-merging.c
4181 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4182 store_immediate_info for bswap/nop orig_stores.
4184 2018-01-15 Andrew Waterman <andrew@sifive.com>
4186 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4188 <UDIV>: Increase cost if !TARGET_DIV.
4190 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4192 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4193 (define_attr "cr_logical_3op"): New.
4194 (cceq_ior_compare): Adjust.
4195 (cceq_ior_compare_complement): Adjust.
4196 (*cceq_rev_compare): Adjust.
4197 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4198 (is_cracked_insn): Adjust.
4199 (insn_must_be_first_in_group): Adjust.
4200 * config/rs6000/40x.md: Adjust.
4201 * config/rs6000/440.md: Adjust.
4202 * config/rs6000/476.md: Adjust.
4203 * config/rs6000/601.md: Adjust.
4204 * config/rs6000/603.md: Adjust.
4205 * config/rs6000/6xx.md: Adjust.
4206 * config/rs6000/7450.md: Adjust.
4207 * config/rs6000/7xx.md: Adjust.
4208 * config/rs6000/8540.md: Adjust.
4209 * config/rs6000/cell.md: Adjust.
4210 * config/rs6000/e300c2c3.md: Adjust.
4211 * config/rs6000/e500mc.md: Adjust.
4212 * config/rs6000/e500mc64.md: Adjust.
4213 * config/rs6000/e5500.md: Adjust.
4214 * config/rs6000/e6500.md: Adjust.
4215 * config/rs6000/mpc.md: Adjust.
4216 * config/rs6000/power4.md: Adjust.
4217 * config/rs6000/power5.md: Adjust.
4218 * config/rs6000/power6.md: Adjust.
4219 * config/rs6000/power7.md: Adjust.
4220 * config/rs6000/power8.md: Adjust.
4221 * config/rs6000/power9.md: Adjust.
4222 * config/rs6000/rs64.md: Adjust.
4223 * config/rs6000/titan.md: Adjust.
4225 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4227 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4228 ix86_indirect_branch_register logic.
4230 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4232 * config/i386/constraints.md (Bs): Update
4233 ix86_indirect_branch_register check. Don't check
4234 ix86_indirect_branch_register with GOT_memory_operand.
4236 * config/i386/predicates.md (GOT_memory_operand): Don't check
4237 ix86_indirect_branch_register here.
4238 (GOT32_symbol_operand): Likewise.
4240 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4242 * config/i386/predicates.md (constant_call_address_operand):
4243 Rewrite ix86_indirect_branch_register logic.
4244 (sibcall_insn_operand): Likewise.
4246 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4248 * config/i386/constraints.md (Bs): Replace
4249 ix86_indirect_branch_thunk_register with
4250 ix86_indirect_branch_register.
4252 * config/i386/i386.md (indirect_jump): Likewise.
4253 (tablejump): Likewise.
4254 (*sibcall_memory): Likewise.
4255 (*sibcall_value_memory): Likewise.
4256 Peepholes of indirect call and jump via memory: Likewise.
4257 * config/i386/i386.opt: Likewise.
4258 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4259 (GOT_memory_operand): Likewise.
4260 (call_insn_operand): Likewise.
4261 (sibcall_insn_operand): Likewise.
4262 (GOT32_symbol_operand): Likewise.
4264 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4267 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4268 type rather than type addr's type points to.
4269 (expand_omp_atomic_mutex): Likewise.
4270 (expand_omp_atomic): Likewise.
4272 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4275 * config/i386/i386.c (output_indirect_thunk_function): Use
4276 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4277 for __x86_return_thunk.
4279 2018-01-15 Richard Biener <rguenther@suse.de>
4282 * expmed.c (extract_bit_field_1): Fix typo.
4284 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4287 * config/arm/iterators.md (VF): New mode iterator.
4288 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4289 Remove integer-related logic from pattern.
4290 (neon_vabd<mode>_3): Likewise.
4292 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4295 * common.opt (fstrict-overflow): No longer an alias.
4296 (fwrapv-pointer): New option.
4297 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4298 also for pointer types based on flag_wrapv_pointer.
4299 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4300 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4301 opts->x_flag_wrapv got set.
4302 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4303 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4304 POINTER_TYPE_OVERFLOW_UNDEFINED.
4305 * match.pd: Likewise in address comparison pattern.
4306 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4308 2018-01-15 Richard Biener <rguenther@suse.de>
4311 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4312 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4313 Reset type names to their identifier if their TYPE_DECL doesn't
4314 have linkage (and thus is used for ODR and devirt).
4315 (save_debug_info_for_decl): Remove.
4316 (save_debug_info_for_type): Likewise.
4317 (add_tree_to_fld_list): Adjust.
4318 * tree-pretty-print.c (dump_generic_node): Make dumping of
4319 type names more robust.
4321 2018-01-15 Richard Biener <rguenther@suse.de>
4323 * BASE-VER: Bump to 8.0.1.
4325 2018-01-14 Martin Sebor <msebor@redhat.com>
4328 * builtins.c (check_access): Avoid warning when the no-warning bit
4331 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4333 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4334 * ira-color (allocno_hard_regs_compare): Likewise.
4336 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4339 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4340 Use .pushsection/.popsection.
4342 2018-01-14 Martin Sebor <msebor@redhat.com>
4345 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4347 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4349 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4350 entry from extra_headers.
4351 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4352 extra_headers, make the list bitwise identical to the i?86-*-* one.
4354 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4356 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4357 -mcmodel=large with -mindirect-branch=thunk,
4358 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4359 -mfunction-return=thunk-extern.
4360 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4361 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4362 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4364 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4366 * config/i386/i386.c (print_reg): Print the name of the full
4367 integer register without '%'.
4368 (ix86_print_operand): Handle 'V'.
4369 * doc/extend.texi: Document 'V' modifier.
4371 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4373 * config/i386/constraints.md (Bs): Disallow memory operand for
4374 -mindirect-branch-register.
4376 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4377 (GOT_memory_operand): Likewise.
4378 (call_insn_operand): Likewise.
4379 (sibcall_insn_operand): Likewise.
4380 (GOT32_symbol_operand): Likewise.
4381 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4382 for -mindirect-branch-register.
4383 (tablejump): Likewise.
4384 (*sibcall_memory): Likewise.
4385 (*sibcall_value_memory): Likewise.
4386 Disallow peepholes of indirect call and jump via memory for
4387 -mindirect-branch-register.
4388 (*call_pop): Replace m with Bw.
4389 (*call_value_pop): Likewise.
4390 (*sibcall_pop_memory): Replace m with Bs.
4391 * config/i386/i386.opt (mindirect-branch-register): New option.
4392 * doc/invoke.texi: Document -mindirect-branch-register option.
4394 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4396 * config/i386/i386-protos.h (ix86_output_function_return): New.
4397 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4398 set function_return_type.
4399 (indirect_thunk_name): Add ret_p to indicate thunk for function
4401 (output_indirect_thunk_function): Pass false to
4402 indirect_thunk_name.
4403 (ix86_output_indirect_branch_via_reg): Likewise.
4404 (ix86_output_indirect_branch_via_push): Likewise.
4405 (output_indirect_thunk_function): Create alias for function
4406 return thunk if regno < 0.
4407 (ix86_output_function_return): New function.
4408 (ix86_handle_fndecl_attribute): Handle function_return.
4409 (ix86_attribute_table): Add function_return.
4410 * config/i386/i386.h (machine_function): Add
4411 function_return_type.
4412 * config/i386/i386.md (simple_return_internal): Use
4413 ix86_output_function_return.
4414 (simple_return_internal_long): Likewise.
4415 * config/i386/i386.opt (mfunction-return=): New option.
4416 (indirect_branch): Mention -mfunction-return=.
4417 * doc/extend.texi: Document function_return function attribute.
4418 * doc/invoke.texi: Document -mfunction-return= option.
4420 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4422 * config/i386/i386-opts.h (indirect_branch): New.
4423 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4424 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4425 with local indirect jump when converting indirect call and jump.
4426 (ix86_set_indirect_branch_type): New.
4427 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4428 (indirectlabelno): New.
4429 (indirect_thunk_needed): Likewise.
4430 (indirect_thunk_bnd_needed): Likewise.
4431 (indirect_thunks_used): Likewise.
4432 (indirect_thunks_bnd_used): Likewise.
4433 (INDIRECT_LABEL): Likewise.
4434 (indirect_thunk_name): Likewise.
4435 (output_indirect_thunk): Likewise.
4436 (output_indirect_thunk_function): Likewise.
4437 (ix86_output_indirect_branch_via_reg): Likewise.
4438 (ix86_output_indirect_branch_via_push): Likewise.
4439 (ix86_output_indirect_branch): Likewise.
4440 (ix86_output_indirect_jmp): Likewise.
4441 (ix86_code_end): Call output_indirect_thunk_function if needed.
4442 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4444 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4445 (ix86_attribute_table): Add indirect_branch.
4446 * config/i386/i386.h (machine_function): Add indirect_branch_type
4447 and has_local_indirect_jump.
4448 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4450 (tablejump): Likewise.
4451 (*indirect_jump): Use ix86_output_indirect_jmp.
4452 (*tablejump_1): Likewise.
4453 (simple_return_indirect_internal): Likewise.
4454 * config/i386/i386.opt (mindirect-branch=): New option.
4455 (indirect_branch): New.
4458 (thunk-inline): Likewise.
4459 (thunk-extern): Likewise.
4460 * doc/extend.texi: Document indirect_branch function attribute.
4461 * doc/invoke.texi: Document -mindirect-branch= option.
4463 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4466 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4468 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4470 * ipa-inline.c (want_inline_small_function_p): Return false if
4471 inlining has already failed with CIF_FINAL_ERROR.
4472 (update_caller_keys): Call want_inline_small_function_p before
4474 (update_callee_keys): Likewise.
4476 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4478 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4480 (rs6000_quadword_masked_address_p): Likewise.
4481 (quad_aligned_load_p): Likewise.
4482 (quad_aligned_store_p): Likewise.
4483 (const_load_sequence_p): Add comment to describe the outer-most loop.
4484 (mimic_memory_attributes_and_flags): New function.
4485 (rs6000_gen_stvx): Likewise.
4486 (replace_swapped_aligned_store): Likewise.
4487 (rs6000_gen_lvx): Likewise.
4488 (replace_swapped_aligned_load): Likewise.
4489 (replace_swapped_load_constant): Capitalize argument name in
4490 comment describing this function.
4491 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4492 and stores that access quad-word aligned addresses and replace
4493 with stvx or lvx instructions when appropriate.
4494 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4495 New function prototype.
4496 (rs6000_quadword_masked_address_p): Likewise.
4497 (rs6000_gen_lvx): Likewise.
4498 (rs6000_gen_stvx): Likewise.
4499 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4500 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4501 when memory address is aligned.
4502 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4503 this split to select lvx instruction when memory address is aligned.
4504 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4505 instruction when memory address is aligned.
4506 (*vsx_le_perm_load_v16qi): Likewise.
4507 (four unnamed splitters): Modify to select the stvx instruction
4508 when memory is aligned.
4510 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4512 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4513 which appears in the queue multiple times.
4515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4516 Alan Hayward <alan.hayward@arm.com>
4517 David Sherwood <david.sherwood@arm.com>
4519 * tree-vectorizer.h (vec_lower_bound): New structure.
4520 (_loop_vec_info): Add check_nonzero and lower_bounds.
4521 (LOOP_VINFO_CHECK_NONZERO): New macro.
4522 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4523 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4524 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4525 fields. Make seg_len the distance travelled, not including the
4527 (dr_direction_indicator): Declare.
4528 (dr_zero_step_indicator): Likewise.
4529 (dr_known_forward_stride_p): Likewise.
4530 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4532 (runtime_alias_check_p): Allow runtime alias checks with
4534 (operator ==): Compare access_size and align.
4535 (prune_runtime_alias_test_list): Rework for new distinction between
4536 the access_size and seg_len.
4537 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4539 (get_segment_min_max): New function.
4540 (create_intersect_range_checks): Use it.
4541 (dr_step_indicator): New function.
4542 (dr_direction_indicator): Likewise.
4543 (dr_zero_step_indicator): Likewise.
4544 (dr_known_forward_stride_p): Likewise.
4545 * tree-loop-distribution.c (data_ref_segment_size): Return
4546 DR_STEP * (niters - 1).
4547 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4549 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4550 (vect_preserves_scalar_order_p): New function, split out from...
4551 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4552 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4553 (vect_vfa_access_size): New function.
4554 (vect_vfa_align): Likewise.
4555 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4556 (dump_lower_bound): New function.
4557 (vect_check_lower_bound): Likewise.
4558 (vect_small_gap_p): Likewise.
4559 (vectorizable_with_step_bound_p): Likewise.
4560 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4561 depencies if the vectorization factor is 1. Convert the checks
4562 for nonzero steps into checks on the bounds of DR_STEP. Try using
4563 a bunds check for variable steps if the minimum required step is
4564 relatively small. Update calls to the dr_with_seg_len
4565 constructor and to vect_compile_time_alias.
4566 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4568 (vect_loop_versioning): Call it.
4569 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4571 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4574 Alan Hayward <alan.hayward@arm.com>
4575 David Sherwood <david.sherwood@arm.com>
4577 * doc/sourcebuild.texi (vect_scatter_store): Document.
4578 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4580 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4582 * genopinit.c (main): Add supports_vec_scatter_store and
4583 supports_vec_scatter_store_cached to target_optabs.
4584 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4585 IFN_MASK_SCATTER_STORE.
4586 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4588 * internal-fn.h (internal_store_fn_p): Declare.
4589 (internal_fn_stored_value_index): Likewise.
4590 * internal-fn.c (scatter_store_direct): New macro.
4591 (expand_scatter_store_optab_fn): New function.
4592 (direct_scatter_store_optab_supported_p): New macro.
4593 (internal_store_fn_p): New function.
4594 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4595 IFN_MASK_SCATTER_STORE.
4596 (internal_fn_mask_index): Likewise.
4597 (internal_fn_stored_value_index): New function.
4598 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4600 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4601 * optabs-query.c (supports_vec_scatter_store_p): New function.
4602 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4603 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4604 true for scatter stores.
4605 (vect_gather_scatter_fn_p): Handle scatter stores too.
4606 (vect_check_gather_scatter): Consider using scatter stores if
4607 supports_vec_scatter_store_p.
4608 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4610 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4611 internal_fn_stored_value_index.
4612 (check_load_store_masking): Handle scatter stores too.
4613 (vect_get_store_rhs): Make public.
4614 (vectorizable_call): Use internal_store_fn_p.
4615 (vectorizable_store): Handle scatter store internal functions.
4616 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4617 when deciding whether the end of the group has been reached.
4618 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4619 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4620 (mask_scatter_store<mode>): New insns.
4622 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4623 Alan Hayward <alan.hayward@arm.com>
4624 David Sherwood <david.sherwood@arm.com>
4626 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4627 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4628 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4630 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4631 Use vect_truncate_gather_scatter_offset if we can't treat the
4632 operation as a normal gather load or scatter store.
4633 (get_group_load_store_type): Take the gather_scatter_info
4634 as argument. Try using a gather load or scatter store for
4635 single-element groups.
4636 (get_load_store_type): Update calls to get_group_load_store_type
4637 and vect_use_strided_gather_scatters_p.
4639 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4640 Alan Hayward <alan.hayward@arm.com>
4641 David Sherwood <david.sherwood@arm.com>
4643 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4644 optional tree argument.
4645 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4647 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4648 but continue to use the current value as a fallback.
4649 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4650 to compare the updates.
4651 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4652 (get_load_store_type): Use it when handling a strided access.
4653 (vect_get_strided_load_store_ops): New function.
4654 (vect_get_data_ptr_increment): Likewise.
4655 (vectorizable_load): Handle strided gather loads. Always pass
4656 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4659 Alan Hayward <alan.hayward@arm.com>
4660 David Sherwood <david.sherwood@arm.com>
4662 * doc/md.texi (gather_load@var{m}): Document.
4663 (mask_gather_load@var{m}): Likewise.
4664 * genopinit.c (main): Add supports_vec_gather_load and
4665 supports_vec_gather_load_cached to target_optabs.
4666 * optabs-tree.c (init_tree_optimization_optabs): Use
4667 ggc_cleared_alloc to allocate target_optabs.
4668 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4669 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4671 * internal-fn.h (internal_load_fn_p): Declare.
4672 (internal_gather_scatter_fn_p): Likewise.
4673 (internal_fn_mask_index): Likewise.
4674 (internal_gather_scatter_fn_supported_p): Likewise.
4675 * internal-fn.c (gather_load_direct): New macro.
4676 (expand_gather_load_optab_fn): New function.
4677 (direct_gather_load_optab_supported_p): New macro.
4678 (direct_internal_fn_optab): New function.
4679 (internal_load_fn_p): Likewise.
4680 (internal_gather_scatter_fn_p): Likewise.
4681 (internal_fn_mask_index): Likewise.
4682 (internal_gather_scatter_fn_supported_p): Likewise.
4683 * optabs-query.c (supports_at_least_one_mode_p): New function.
4684 (supports_vec_gather_load_p): Likewise.
4685 * optabs-query.h (supports_vec_gather_load_p): Declare.
4686 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4687 and memory_type field.
4688 (NUM_PATTERNS): Bump to 15.
4689 * tree-vect-data-refs.c: Include internal-fn.h.
4690 (vect_gather_scatter_fn_p): New function.
4691 (vect_describe_gather_scatter_call): Likewise.
4692 (vect_check_gather_scatter): Try using internal functions for
4693 gather loads. Recognize existing calls to a gather load function.
4694 (vect_analyze_data_refs): Consider using gather loads if
4695 supports_vec_gather_load_p.
4696 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4697 (vect_get_gather_scatter_offset_type): Likewise.
4698 (vect_convert_mask_for_vectype): Likewise.
4699 (vect_add_conversion_to_patterm): Likewise.
4700 (vect_try_gather_scatter_pattern): Likewise.
4701 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4702 (vect_vect_recog_func_ptrs): Add it.
4703 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4704 internal_fn_mask_index and internal_gather_scatter_fn_p.
4705 (check_load_store_masking): Take the gather_scatter_info as an
4706 argument and handle gather loads.
4707 (vect_get_gather_scatter_ops): New function.
4708 (vectorizable_call): Check internal_load_fn_p.
4709 (vectorizable_load): Likewise. Handle gather load internal
4711 (vectorizable_store): Update call to check_load_store_masking.
4712 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4713 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4714 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4715 (aarch64_gather_scale_operand_d): New predicates.
4716 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4717 (mask_gather_load<mode>): New insns.
4719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4720 Alan Hayward <alan.hayward@arm.com>
4721 David Sherwood <david.sherwood@arm.com>
4723 * optabs.def (fold_left_plus_optab): New optab.
4724 * doc/md.texi (fold_left_plus_@var{m}): Document.
4725 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4726 * internal-fn.c (fold_left_direct): Define.
4727 (expand_fold_left_optab_fn): Likewise.
4728 (direct_fold_left_optab_supported_p): Likewise.
4729 * fold-const-call.c (fold_const_fold_left): New function.
4730 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4731 * tree-parloops.c (valid_reduction_p): New function.
4732 (gather_scalar_reductions): Use it.
4733 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4734 (vect_finish_replace_stmt): Declare.
4735 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4736 (needs_fold_left_reduction_p): New function, split out from...
4737 (vect_is_simple_reduction): ...here. Accept reductions that
4738 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4739 (vect_force_simple_reduction): Also store the reduction type in
4740 the assignment's STMT_VINFO_REDUC_TYPE.
4741 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4742 (merge_with_identity): New function.
4743 (vect_expand_fold_left): Likewise.
4744 (vectorize_fold_left_reduction): Likewise.
4745 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4746 scalar phi in place for it. Check for target support and reject
4747 cases that would reassociate the operation. Defer the transform
4748 phase to vectorize_fold_left_reduction.
4749 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4750 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4751 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4753 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4755 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4756 call to ifc_temp_var.
4758 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4759 Alan Hayward <alan.hayward@arm.com>
4760 David Sherwood <david.sherwood@arm.com>
4762 * target.def (legitimize_address_displacement): Take the original
4763 offset as a poly_int.
4764 * targhooks.h (default_legitimize_address_displacement): Update
4766 * targhooks.c (default_legitimize_address_displacement): Likewise.
4767 * doc/tm.texi: Regenerate.
4768 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4769 as an argument, moving assert of ad->disp == ad->disp_term to...
4770 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4771 Try calling targetm.legitimize_address_displacement before expanding
4772 the address rather than afterwards, and adjust for the new interface.
4773 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4774 Match the new hook interface. Handle SVE addresses.
4775 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4778 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4780 * Makefile.in (OBJS): Add early-remat.o.
4781 * target.def (select_early_remat_modes): New hook.
4782 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4783 * doc/tm.texi: Regenerate.
4784 * targhooks.h (default_select_early_remat_modes): Declare.
4785 * targhooks.c (default_select_early_remat_modes): New function.
4786 * timevar.def (TV_EARLY_REMAT): New timevar.
4787 * passes.def (pass_early_remat): New pass.
4788 * tree-pass.h (make_pass_early_remat): Declare.
4789 * early-remat.c: New file.
4790 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4792 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4795 Alan Hayward <alan.hayward@arm.com>
4796 David Sherwood <david.sherwood@arm.com>
4798 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4799 vfm1 with a bound_epilog parameter.
4800 (vect_do_peeling): Update calls accordingly, and move the prologue
4801 call earlier in the function. Treat the base bound_epilog as 0 for
4802 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4803 this base when peeling for gaps.
4804 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4805 with fully-masked loops.
4806 (vect_estimate_min_profitable_iters): Handle the single peeled
4807 iteration in that case.
4809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4810 Alan Hayward <alan.hayward@arm.com>
4811 David Sherwood <david.sherwood@arm.com>
4813 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4814 single-element interleaving even if the size is not a power of 2.
4815 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4816 accesses for single-element interleaving if the group size is
4819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4820 Alan Hayward <alan.hayward@arm.com>
4821 David Sherwood <david.sherwood@arm.com>
4823 * doc/md.texi (fold_extract_last_@var{m}): Document.
4824 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4825 * optabs.def (fold_extract_last_optab): New optab.
4826 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4827 * internal-fn.c (fold_extract_direct): New macro.
4828 (expand_fold_extract_optab_fn): Likewise.
4829 (direct_fold_extract_optab_supported_p): Likewise.
4830 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4831 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4832 EXTRACT_LAST_REDUCTION.
4833 (get_initial_def_for_reduction): Do not create an initial vector
4834 for EXTRACT_LAST_REDUCTION reductions.
4835 (vectorizable_reduction): Leave the scalar phi in place for
4836 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4837 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4838 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4839 transform phase to vectorizable_condition.
4840 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4842 (vect_finish_stmt_generation): ...here.
4843 (vect_finish_replace_stmt): New function.
4844 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4845 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4847 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4850 Alan Hayward <alan.hayward@arm.com>
4851 David Sherwood <david.sherwood@arm.com>
4853 * doc/md.texi (extract_last_@var{m}): Document.
4854 * optabs.def (extract_last_optab): New optab.
4855 * internal-fn.def (EXTRACT_LAST): New internal function.
4856 * internal-fn.c (cond_unary_direct): New macro.
4857 (expand_cond_unary_optab_fn): Likewise.
4858 (direct_cond_unary_optab_supported_p): Likewise.
4859 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4860 loops using EXTRACT_LAST.
4861 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4862 (extract_last_<mode>): ...this optab.
4863 (vec_extract<mode><Vel>): Update accordingly.
4865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4866 Alan Hayward <alan.hayward@arm.com>
4867 David Sherwood <david.sherwood@arm.com>
4869 * target.def (empty_mask_is_expensive): New hook.
4870 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4871 * doc/tm.texi: Regenerate.
4872 * targhooks.h (default_empty_mask_is_expensive): Declare.
4873 * targhooks.c (default_empty_mask_is_expensive): New function.
4874 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4875 if the target says that empty masks are expensive.
4876 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4878 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4881 Alan Hayward <alan.hayward@arm.com>
4882 David Sherwood <david.sherwood@arm.com>
4884 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4885 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4886 (vect_use_loop_mask_for_alignment_p): New function.
4887 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4888 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4889 niters_skip argument. Make sure that the first niters_skip elements
4890 of the first iteration are inactive.
4891 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4892 Update call to vect_set_loop_masks_directly.
4893 (get_misalign_in_elems): New function, split out from...
4894 (vect_gen_prolog_loop_niters): ...here.
4895 (vect_update_init_of_dr): Take a code argument that specifies whether
4896 the adjustment should be added or subtracted.
4897 (vect_update_init_of_drs): Likewise.
4898 (vect_prepare_for_masked_peels): New function.
4899 (vect_do_peeling): Skip prologue peeling if we're using a mask
4900 instead. Update call to vect_update_inits_of_drs.
4901 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4903 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4904 alignment. Do not include the number of peeled iterations in
4905 the minimum threshold in that case.
4906 (vectorizable_induction): Adjust the start value down by
4907 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4908 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4909 Take the number of skipped iterations into account when calculating
4911 * tree-vect-stmts.c (vect_gen_while_not): New function.
4913 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4914 Alan Hayward <alan.hayward@arm.com>
4915 David Sherwood <david.sherwood@arm.com>
4917 * doc/sourcebuild.texi (vect_fully_masked): Document.
4918 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4920 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4922 (vect_analyze_loop_2): ...here. Don't check the vectorization
4923 factor against the number of loop iterations if the loop is
4926 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4927 Alan Hayward <alan.hayward@arm.com>
4928 David Sherwood <david.sherwood@arm.com>
4930 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4931 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4932 (dump_groups): Update accordingly.
4933 (iv_use::mem_type): New member variable.
4934 (address_p): New function.
4935 (record_use): Add a mem_type argument and initialize the new
4937 (record_group_use): Add a mem_type argument. Use address_p.
4938 Remove obsolete null checks of base_object. Update call to record_use.
4939 (find_interesting_uses_op): Update call to record_group_use.
4940 (find_interesting_uses_cond): Likewise.
4941 (find_interesting_uses_address): Likewise.
4942 (get_mem_type_for_internal_fn): New function.
4943 (find_address_like_use): Likewise.
4944 (find_interesting_uses_stmt): Try find_address_like_use before
4945 calling find_interesting_uses_op.
4946 (addr_offset_valid_p): Use the iv mem_type field as the type
4947 of the addressed memory.
4948 (add_autoinc_candidates): Likewise.
4949 (get_address_cost): Likewise.
4950 (split_small_address_groups_p): Use address_p.
4951 (split_address_groups): Likewise.
4952 (add_iv_candidate_for_use): Likewise.
4953 (autoinc_possible_for_pair): Likewise.
4954 (rewrite_groups): Likewise.
4955 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4956 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4957 (get_alias_ptr_type_for_ptr_address): New function.
4958 (rewrite_use_address): Rewrite address uses in calls that were
4959 identified by find_address_like_use.
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 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4967 * gimple-expr.h (is_gimple_addressable: Likewise.
4968 * gimple-expr.c (is_gimple_address): Likewise.
4969 * internal-fn.c (expand_call_mem_ref): New function.
4970 (expand_mask_load_optab_fn): Use it.
4971 (expand_mask_store_optab_fn): Likewise.
4973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4974 Alan Hayward <alan.hayward@arm.com>
4975 David Sherwood <david.sherwood@arm.com>
4977 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4978 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4979 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4980 (cond_umax@var{mode}): Document.
4981 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4982 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4983 (cond_umin_optab, cond_umax_optab): New optabs.
4984 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4985 (COND_IOR, COND_XOR): New internal functions.
4986 * internal-fn.h (get_conditional_internal_fn): Declare.
4987 * internal-fn.c (cond_binary_direct): New macro.
4988 (expand_cond_binary_optab_fn): Likewise.
4989 (direct_cond_binary_optab_supported_p): Likewise.
4990 (get_conditional_internal_fn): New function.
4991 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4992 Cope with reduction statements that are vectorized as calls rather
4994 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4995 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4996 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4997 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4998 (UNSPEC_COND_EOR): New unspecs.
4999 (optab): Add mappings for them.
5000 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5001 (sve_int_op, sve_fp_op): New int attributes.
5003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5004 Alan Hayward <alan.hayward@arm.com>
5005 David Sherwood <david.sherwood@arm.com>
5007 * optabs.def (while_ult_optab): New optab.
5008 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5009 * internal-fn.def (WHILE_ULT): New internal function.
5010 * internal-fn.h (direct_internal_fn_supported_p): New override
5011 that takes two types as argument.
5012 * internal-fn.c (while_direct): New macro.
5013 (expand_while_optab_fn): New function.
5014 (convert_optab_supported_p): Likewise.
5015 (direct_while_optab_supported_p): New macro.
5016 * wide-int.h (wi::udiv_ceil): New function.
5017 * tree-vectorizer.h (rgroup_masks): New structure.
5018 (vec_loop_masks): New typedef.
5019 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5021 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5022 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5023 (vect_max_vf): New function.
5024 (slpeel_make_loop_iterate_ntimes): Delete.
5025 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5026 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5027 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5028 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5029 internal-fn.h, stor-layout.h and optabs-query.h.
5030 (vect_set_loop_mask): New function.
5031 (add_preheader_seq): Likewise.
5032 (add_header_seq): Likewise.
5033 (interleave_supported_p): Likewise.
5034 (vect_maybe_permute_loop_masks): Likewise.
5035 (vect_set_loop_masks_directly): Likewise.
5036 (vect_set_loop_condition_masked): Likewise.
5037 (vect_set_loop_condition_unmasked): New function, split out from
5038 slpeel_make_loop_iterate_ntimes.
5039 (slpeel_make_loop_iterate_ntimes): Rename to..
5040 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5041 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5042 (vect_do_peeling): Update call accordingly.
5043 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5045 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5046 mask_compare_type, can_fully_mask_p and fully_masked_p.
5047 (release_vec_loop_masks): New function.
5048 (_loop_vec_info): Use it to free the loop masks.
5049 (can_produce_all_loop_masks_p): New function.
5050 (vect_get_max_nscalars_per_iter): Likewise.
5051 (vect_verify_full_masking): Likewise.
5052 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5053 retries, and free the mask rgroups before retrying. Check loop-wide
5054 reasons for disallowing fully-masked loops. Make the final decision
5055 about whether use a fully-masked loop or not.
5056 (vect_estimate_min_profitable_iters): Do not assume that peeling
5057 for the number of iterations will be needed for fully-masked loops.
5058 (vectorizable_reduction): Disable fully-masked loops.
5059 (vectorizable_live_operation): Likewise.
5060 (vect_halve_mask_nunits): New function.
5061 (vect_double_mask_nunits): Likewise.
5062 (vect_record_loop_mask): Likewise.
5063 (vect_get_loop_mask): Likewise.
5064 (vect_transform_loop): Handle the case in which the final loop
5065 iteration might handle a partial vector. Call vect_set_loop_condition
5066 instead of slpeel_make_loop_iterate_ntimes.
5067 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5068 (check_load_store_masking): New function.
5069 (prepare_load_store_mask): Likewise.
5070 (vectorizable_store): Handle fully-masked loops.
5071 (vectorizable_load): Likewise.
5072 (supportable_widening_operation): Use vect_halve_mask_nunits for
5074 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5075 (vect_gen_while): New function.
5076 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5077 (aarch64_uqdec<mode>): New insn.
5079 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5080 Alan Hayward <alan.hayward@arm.com>
5081 David Sherwood <david.sherwood@arm.com>
5083 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5084 (reduc_xor_scal_optab): New optabs.
5085 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5086 (reduc_xor_scal_@var{m}): Document.
5087 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5088 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5090 * fold-const-call.c (fold_const_call): Handle them.
5091 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5092 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5093 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5094 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5095 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5096 (UNSPEC_XORV): New unspecs.
5097 (optab): Add entries for them.
5098 (BITWISEV): New int iterator.
5099 (bit_reduc_op): New int attributes.
5101 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5102 Alan Hayward <alan.hayward@arm.com>
5103 David Sherwood <david.sherwood@arm.com>
5105 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5106 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5107 * optabs.def (vec_shl_insert_optab): New optab.
5108 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5109 (duplicate_and_interleave): Likewise.
5110 * tree-vect-loop.c: Include internal-fn.h.
5111 (neutral_op_for_slp_reduction): New function, split out from
5112 get_initial_defs_for_reduction.
5113 (get_initial_def_for_reduction): Handle option 2 for variable-length
5114 vectors by loading the neutral value into a vector and then shifting
5115 the initial value into element 0.
5116 (get_initial_defs_for_reduction): Replace the code argument with
5117 the neutral value calculated by neutral_op_for_slp_reduction.
5118 Use gimple_build_vector for constant-length vectors.
5119 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5120 but the first group_size elements have a neutral value.
5121 Use duplicate_and_interleave otherwise.
5122 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5123 Update call to get_initial_defs_for_reduction. Handle SLP
5124 reductions for variable-length vectors by creating one vector
5125 result for each scalar result, with the elements associated
5126 with other scalar results stubbed out with the neutral value.
5127 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5128 Require IFN_VEC_SHL_INSERT for double reductions on
5129 variable-length vectors, or SLP reductions that have
5130 a neutral value. Require can_duplicate_and_interleave_p
5131 support for variable-length unchained SLP reductions if there
5132 is no neutral value, such as for MIN/MAX reductions. Also require
5133 the number of vector elements to be a multiple of the number of
5134 SLP statements when doing variable-length unchained SLP reductions.
5135 Update call to vect_create_epilog_for_reduction.
5136 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5137 and remove initial values.
5138 (duplicate_and_interleave): Make public.
5139 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5140 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5143 Alan Hayward <alan.hayward@arm.com>
5144 David Sherwood <david.sherwood@arm.com>
5146 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5147 (can_duplicate_and_interleave_p): New function.
5148 (vect_get_and_check_slp_defs): Take the vector of statements
5149 rather than just the current one. Remove excess parentheses.
5150 Restriction rejectinon of vect_constant_def and vect_external_def
5151 for variable-length vectors to boolean types, or types for which
5152 can_duplicate_and_interleave_p is false.
5153 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5154 (duplicate_and_interleave): New function.
5155 (vect_get_constant_vectors): Use gimple_build_vector for
5156 constant-length vectors and suitable variable-length constant
5157 vectors. Use duplicate_and_interleave for other variable-length
5158 vectors. Don't defer the update when inserting new statements.
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 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5165 min_profitable_iters doesn't go negative.
5167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5168 Alan Hayward <alan.hayward@arm.com>
5169 David Sherwood <david.sherwood@arm.com>
5171 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5172 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5173 * optabs.def (vec_mask_load_lanes_optab): New optab.
5174 (vec_mask_store_lanes_optab): Likewise.
5175 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5176 (MASK_STORE_LANES): Likewise.
5177 * internal-fn.c (mask_load_lanes_direct): New macro.
5178 (mask_store_lanes_direct): Likewise.
5179 (expand_mask_load_optab_fn): Handle masked operations.
5180 (expand_mask_load_lanes_optab_fn): New macro.
5181 (expand_mask_store_optab_fn): Handle masked operations.
5182 (expand_mask_store_lanes_optab_fn): New macro.
5183 (direct_mask_load_lanes_optab_supported_p): Likewise.
5184 (direct_mask_store_lanes_optab_supported_p): Likewise.
5185 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5187 (vect_load_lanes_supported): Likewise.
5188 * tree-vect-data-refs.c (strip_conversion): New function.
5189 (can_group_stmts_p): Likewise.
5190 (vect_analyze_data_ref_accesses): Use it instead of checking
5191 for a pair of assignments.
5192 (vect_store_lanes_supported): Take a masked_p parameter.
5193 (vect_load_lanes_supported): Likewise.
5194 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5195 vect_store_lanes_supported and vect_load_lanes_supported.
5196 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5197 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5198 parameter. Don't allow gaps for masked accesses.
5199 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5200 and vect_load_lanes_supported.
5201 (get_load_store_type): Take a masked_p parameter and update
5202 call to get_group_load_store_type.
5203 (vectorizable_store): Update call to get_load_store_type.
5204 Handle IFN_MASK_STORE_LANES.
5205 (vectorizable_load): Update call to get_load_store_type.
5206 Handle IFN_MASK_LOAD_LANES.
5208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5209 Alan Hayward <alan.hayward@arm.com>
5210 David Sherwood <david.sherwood@arm.com>
5212 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5214 * config/aarch64/aarch64-protos.h
5215 (aarch64_sve_struct_memory_operand_p): Declare.
5216 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5217 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5218 (VPRED, vpred): Handle SVE structure modes.
5219 * config/aarch64/constraints.md (Utx): New constraint.
5220 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5221 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5222 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5223 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5224 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5225 structure modes. Split into pieces after RA.
5226 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5227 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5229 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5230 SVE structure modes.
5231 (aarch64_classify_address): Likewise.
5232 (sizetochar): Move earlier in file.
5233 (aarch64_print_operand): Handle SVE register lists.
5234 (aarch64_array_mode): New function.
5235 (aarch64_sve_struct_memory_operand_p): Likewise.
5236 (TARGET_ARRAY_MODE): Redefine.
5238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5239 Alan Hayward <alan.hayward@arm.com>
5240 David Sherwood <david.sherwood@arm.com>
5242 * target.def (array_mode): New target hook.
5243 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5244 * doc/tm.texi: Regenerate.
5245 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5246 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5247 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5249 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5253 Alan Hayward <alan.hayward@arm.com>
5254 David Sherwood <david.sherwood@arm.com>
5256 * fold-const.c (fold_binary_loc): Check the argument types
5257 rather than the result type when testing for a vector operation.
5259 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5261 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5262 * doc/tm.texi: Regenerate.
5264 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5265 Alan Hayward <alan.hayward@arm.com>
5266 David Sherwood <david.sherwood@arm.com>
5268 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5269 (sve): Document new AArch64 extension.
5270 * doc/md.texi (w): Extend the description of the AArch64
5271 constraint to include SVE vectors.
5272 (Upl, Upa): Document new AArch64 predicate constraints.
5273 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5275 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5276 (msve-vector-bits=): New option.
5277 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5278 SVE when these are disabled.
5279 (sve): New extension.
5280 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5281 modes. Adjust their number of units based on aarch64_sve_vg.
5282 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5283 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5284 aarch64_addr_query_type.
5285 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5286 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5287 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5288 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5289 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5290 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5291 (aarch64_simd_imm_zero_p): Delete.
5292 (aarch64_check_zero_based_sve_index_immediate): Declare.
5293 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5294 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5295 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5296 (aarch64_sve_float_mul_immediate_p): Likewise.
5297 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5299 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5300 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5301 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5302 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5303 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5304 (aarch64_regmode_natural_size): Likewise.
5305 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5306 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5308 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5309 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5310 for VG and the SVE predicate registers.
5311 (V_ALIASES): Add a "z"-prefixed alias.
5312 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5313 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5314 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5315 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5316 (REG_CLASS_NAMES): Add entries for them.
5317 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5318 and the predicate registers.
5319 (aarch64_sve_vg): Declare.
5320 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5321 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5322 (REGMODE_NATURAL_SIZE): Define.
5323 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5325 * config/aarch64/aarch64.c: Include cfgrtl.h.
5326 (simd_immediate_info): Add a constructor for series vectors,
5327 and an associated step field.
5328 (aarch64_sve_vg): New variable.
5329 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5330 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5331 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5332 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5333 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5334 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5335 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5336 (aarch64_get_mask_mode): New functions.
5337 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5338 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5339 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5340 predicate modes and predicate registers. Explicitly restrict
5341 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5342 to store a vector mode if it is recognized by
5343 aarch64_classify_vector_mode.
5344 (aarch64_regmode_natural_size): New function.
5345 (aarch64_hard_regno_caller_save_mode): Return the original mode
5347 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5348 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5349 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5350 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5352 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5353 does not overlap dest if the function is frame-related. Handle
5355 (aarch64_split_add_offset): New function.
5356 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5357 them aarch64_add_offset.
5358 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5359 and update call to aarch64_sub_sp.
5360 (aarch64_add_cfa_expression): New function.
5361 (aarch64_expand_prologue): Pass extra temporary registers to the
5362 functions above. Handle the case in which we need to emit new
5363 DW_CFA_expressions for registers that were originally saved
5364 relative to the stack pointer, but now have to be expressed
5365 relative to the frame pointer.
5366 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5368 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5369 IP0 and IP1 values for SVE frames.
5370 (aarch64_expand_vec_series): New function.
5371 (aarch64_expand_sve_widened_duplicate): Likewise.
5372 (aarch64_expand_sve_const_vector): Likewise.
5373 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5374 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5375 into the register, rather than emitting a SET directly.
5376 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5377 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5378 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5379 (offset_9bit_signed_scaled_p): New functions.
5380 (aarch64_replicate_bitmask_imm): New function.
5381 (aarch64_bitmask_imm): Use it.
5382 (aarch64_cannot_force_const_mem): Reject expressions involving
5383 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5384 (aarch64_classify_index): Handle SVE indices, by requiring
5385 a plain register index with a scale that matches the element size.
5386 (aarch64_classify_address): Handle SVE addresses. Assert that
5387 the mode of the address is VOIDmode or an integer mode.
5388 Update call to aarch64_classify_symbol.
5389 (aarch64_classify_symbolic_expression): Update call to
5390 aarch64_classify_symbol.
5391 (aarch64_const_vec_all_in_range_p): New function.
5392 (aarch64_print_vector_float_operand): Likewise.
5393 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5394 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5395 and the FP immediates 1.0 and 0.5.
5396 (aarch64_print_address_internal): Handle SVE addresses.
5397 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5398 (aarch64_regno_regclass): Handle predicate registers.
5399 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5401 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5402 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5403 (aarch64_convert_sve_vector_bits): New function.
5404 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5405 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5407 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5408 Handle SVE vector and predicate modes. Accept VL-based constants
5409 that need only one temporary register, and VL offsets that require
5410 no temporary registers.
5411 (aarch64_conditional_register_usage): Mark the predicate registers
5412 as fixed if SVE isn't available.
5413 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5414 Return true for SVE vector and predicate modes.
5415 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5416 rather than an unsigned int. Handle SVE modes.
5417 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5419 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5421 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5422 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5423 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5424 (aarch64_sve_float_mul_immediate_p): New functions.
5425 (aarch64_sve_valid_immediate): New function.
5426 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5427 Explicitly reject structure modes. Check for INDEX constants.
5428 Handle PTRUE and PFALSE constants.
5429 (aarch64_check_zero_based_sve_index_immediate): New function.
5430 (aarch64_simd_imm_zero_p): Delete.
5431 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5432 vector modes. Accept constants in the range of CNT[BHWD].
5433 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5434 ask for an Advanced SIMD mode.
5435 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5436 (aarch64_simd_vector_alignment): Handle SVE predicates.
5437 (aarch64_vectorize_preferred_vector_alignment): New function.
5438 (aarch64_simd_vector_alignment_reachable): Use it instead of
5440 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5441 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5443 (MAX_VECT_LEN): Delete.
5444 (expand_vec_perm_d): Add a vec_flags field.
5445 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5446 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5447 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5449 (aarch64_evpc_rev): Rename to...
5450 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5451 (aarch64_evpc_rev_global): New function.
5452 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5453 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5455 (aarch64_evpc_sve_tbl): New function.
5456 (aarch64_expand_vec_perm_const_1): Update after rename of
5457 aarch64_evpc_rev. Handle SVE permutes too, trying
5458 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5459 than aarch64_evpc_tbl.
5460 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5461 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5462 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5463 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5464 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5465 (aarch64_expand_sve_vcond): New functions.
5466 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5467 of aarch64_vector_mode_p.
5468 (aarch64_dwarf_poly_indeterminate_value): New function.
5469 (aarch64_compute_pressure_classes): Likewise.
5470 (aarch64_can_change_mode_class): Likewise.
5471 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5472 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5473 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5474 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5475 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5476 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5477 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5478 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5480 (Dn, Dl, Dr): Accept const as well as const_vector.
5481 (Dz): Likewise. Compare against CONST0_RTX.
5482 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5483 of "vector" where appropriate.
5484 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5485 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5486 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5487 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5488 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5489 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5490 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5491 (v_int_equiv): Extend to SVE modes.
5492 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5494 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5495 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5496 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5497 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5498 (SVE_COND_FP_CMP): New int iterators.
5499 (perm_hilo): Handle the new unpack unspecs.
5500 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5502 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5503 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5504 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5505 (aarch64_equality_operator, aarch64_constant_vector_operand)
5506 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5507 (aarch64_sve_nonimmediate_operand): Likewise.
5508 (aarch64_sve_general_operand): Likewise.
5509 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5510 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5511 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5512 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5513 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5514 (aarch64_sve_float_arith_immediate): Likewise.
5515 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5516 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5517 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5518 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5519 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5520 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5521 (aarch64_sve_float_arith_operand): Likewise.
5522 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5523 (aarch64_sve_float_mul_operand): Likewise.
5524 (aarch64_sve_vec_perm_operand): Likewise.
5525 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5526 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5527 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5528 as well as const_vector.
5529 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5530 in file. Use CONST0_RTX and CONSTM1_RTX.
5531 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5532 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5533 Use aarch64_simd_imm_zero.
5534 * config/aarch64/aarch64-sve.md: New file.
5535 * config/aarch64/aarch64.md: Include it.
5536 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5537 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5538 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5539 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5540 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5541 (sve): New attribute.
5542 (enabled): Disable instructions with the sve attribute unless
5544 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5545 aarch64_expand_mov_immediate.
5546 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5547 CNT[BHSD] immediates.
5548 (movti): Split CONST_POLY_INT moves into two halves.
5549 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5550 Split additions that need a temporary here if the destination
5551 is the stack pointer.
5552 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5553 (*add<mode>3_poly_1): New instruction.
5554 (set_clobber_cc): New expander.
5556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5558 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5559 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5560 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5561 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5562 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5563 Change innermode from fixed_mode_size to machine_mode.
5564 (simplify_subreg): Update call accordingly. Handle a constant-sized
5565 subreg of a variable-length CONST_VECTOR.
5567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5568 Alan Hayward <alan.hayward@arm.com>
5569 David Sherwood <david.sherwood@arm.com>
5571 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5572 (add_offset_to_base): New function, split out from...
5573 (create_mem_ref): ...here. When handling a scale other than 1,
5574 check first whether the address is valid without the offset.
5575 Add it into the base if so, leaving the index and scale as-is.
5577 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5580 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5581 fold_for_warn before checking if arg2 is INTEGER_CST.
5583 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5585 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5586 (store_multiple_operation): Delete.
5587 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5588 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5589 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5590 guarded by TARGET_STRING.
5591 (rs6000_output_load_multiple): Delete.
5592 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5593 OPTION_MASK_STRING / TARGET_STRING handling.
5594 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5595 (const rs6000_opt_masks) <"string">: Change mask to 0.
5596 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5597 (MASK_STRING): Delete.
5598 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5600 (load_multiple): Delete.
5607 (store_multiple): Delete.
5614 (movmemsi_8reg): Delete.
5615 (corresponding unnamed define_insn): Delete.
5616 (movmemsi_6reg): Delete.
5617 (corresponding unnamed define_insn): Delete.
5618 (movmemsi_4reg): Delete.
5619 (corresponding unnamed define_insn): Delete.
5620 (movmemsi_2reg): Delete.
5621 (corresponding unnamed define_insn): Delete.
5622 (movmemsi_1reg): Delete.
5623 (corresponding unnamed define_insn): Delete.
5624 * config/rs6000/rs6000.opt (mno-string): New.
5625 (mstring): Replace by deprecation warning stub.
5626 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5628 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5630 * regrename.c (regrename_do_replace): If replacing the same
5631 reg multiple times, try to reuse last created gen_raw_REG.
5634 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5635 main to workaround a bug in GDB.
5637 2018-01-12 Tom de Vries <tom@codesourcery.com>
5640 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5642 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5644 PR rtl-optimization/80481
5645 * ira-color.c (get_cap_member): New function.
5646 (allocnos_conflict_by_live_ranges_p): Use it.
5647 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5648 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5650 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5653 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5654 (*saddl_se_1): Ditto.
5656 (*ssubl_se_1): Ditto.
5658 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5660 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5661 rather than wi::to_widest for DR_INITs.
5662 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5663 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5664 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5666 (vect_analyze_group_access_1): Note that here.
5668 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5670 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5671 polynomial type sizes.
5673 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5675 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5676 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5677 (gimple_add_tmp_var): Likewise.
5679 2018-01-12 Martin Liska <mliska@suse.cz>
5681 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5682 (gimple_alloc_sizes): Likewise.
5683 (dump_gimple_statistics): Use PRIu64 in printf format.
5684 * gimple.h: Change uint64_t to int.
5686 2018-01-12 Martin Liska <mliska@suse.cz>
5688 * tree-core.h: Use uint64_t instead of int.
5689 * tree.c (tree_node_counts): Likewise.
5690 (tree_node_sizes): Likewise.
5691 (dump_tree_statistics): Use PRIu64 in printf format.
5693 2018-01-12 Martin Liska <mliska@suse.cz>
5695 * Makefile.in: As qsort_chk is implemented in vec.c, add
5696 vec.o to linkage of gencfn-macros.
5697 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5698 passing the info to record_node_allocation_statistics.
5699 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5701 * ggc-common.c (struct ggc_usage): Add operator== and use
5702 it in operator< and compare function.
5703 * mem-stats.h (struct mem_usage): Likewise.
5704 * vec.c (struct vec_usage): Remove operator< and compare
5705 function. Can be simply inherited.
5707 2018-01-12 Martin Jambor <mjambor@suse.cz>
5710 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5711 * tree-ssa-math-opts.c: Include domwalk.h.
5712 (convert_mult_to_fma_1): New function.
5713 (fma_transformation_info): New type.
5714 (fma_deferring_state): Likewise.
5715 (cancel_fma_deferring): New function.
5716 (result_of_phi): Likewise.
5717 (last_fma_candidate_feeds_initial_phi): Likewise.
5718 (convert_mult_to_fma): Added deferring logic, split actual
5719 transformation to convert_mult_to_fma_1.
5720 (math_opts_dom_walker): New type.
5721 (math_opts_dom_walker::after_dom_children): New method, body moved
5722 here from pass_optimize_widening_mul::execute, added deferring logic
5724 (pass_optimize_widening_mul::execute): Moved most of code to
5725 math_opts_dom_walker::after_dom_children.
5726 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5727 * config/i386/i386.c (ix86_option_override_internal): Added
5728 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5730 2018-01-12 Richard Biener <rguenther@suse.de>
5733 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5734 inline instance vars.
5736 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5739 * config/rx/rx.c (rx_is_restricted_memory_address):
5742 2018-01-12 Richard Biener <rguenther@suse.de>
5744 PR tree-optimization/80846
5745 * target.def (split_reduction): New target hook.
5746 * targhooks.c (default_split_reduction): New function.
5747 * targhooks.h (default_split_reduction): Declare.
5748 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5749 target requests first reduce vectors by combining low and high
5751 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5752 (get_vectype_for_scalar_type_and_size): Export.
5753 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5754 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5755 * doc/tm.texi: Regenerate.
5756 * config/i386/i386.c (ix86_split_reduction): Implement
5757 TARGET_VECTORIZE_SPLIT_REDUCTION.
5759 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5762 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5763 in PIC mode except for TARGET_VXWORKS_RTP.
5764 * config/sparc/sparc.c: Include cfgrtl.h.
5765 (TARGET_INIT_PIC_REG): Define.
5766 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5767 (sparc_pic_register_p): New predicate.
5768 (sparc_legitimate_address_p): Use it.
5769 (sparc_legitimize_pic_address): Likewise.
5770 (sparc_delegitimize_address): Likewise.
5771 (sparc_mode_dependent_address_p): Likewise.
5772 (gen_load_pcrel_sym): Remove 4th parameter.
5773 (load_got_register): Adjust call to above. Remove obsolete stuff.
5774 (sparc_expand_prologue): Do not call load_got_register here.
5775 (sparc_flat_expand_prologue): Likewise.
5776 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5777 (sparc_use_pseudo_pic_reg): New function.
5778 (sparc_init_pic_reg): Likewise.
5779 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5780 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5782 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5784 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5785 Add item for branch_cost.
5787 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5789 PR rtl-optimization/83565
5790 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5791 not extend the result to a larger mode for rotate operations.
5792 (num_sign_bit_copies1): Likewise.
5794 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5797 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5799 Use values-Xc.o for -pedantic.
5800 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5802 2018-01-12 Martin Liska <mliska@suse.cz>
5805 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5807 (possible_polymorphic_call_targets): Use it.
5808 (ipa_devirt): Likewise.
5810 2018-01-12 Martin Liska <mliska@suse.cz>
5812 * profile-count.h (enum profile_quality): Use 0 as invalid
5813 enum value of profile_quality.
5815 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5817 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5818 -mext-string options.
5820 2018-01-12 Richard Biener <rguenther@suse.de>
5822 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5823 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5824 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5826 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5828 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5830 * configure.ac (--with-long-double-format): Add support for the
5831 configuration option to change the default long double format on
5833 * config.gcc (powerpc*-linux*-*): Likewise.
5834 * configure: Regenerate.
5835 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5836 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5837 used without modification.
5839 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5841 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5842 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5843 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5844 MISC_BUILTIN_SPEC_BARRIER.
5845 (rs6000_init_builtins): Likewise.
5846 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5848 (speculation_barrier): New define_insn.
5849 * doc/extend.texi: Document __builtin_speculation_barrier.
5851 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5854 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5855 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5856 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5858 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5859 integral modes instead of "ss" and "sd".
5860 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5861 vectors with 32-bit and 64-bit elements.
5862 (vecdupssescalarmodesuffix): New mode attribute.
5863 (vec_dup<mode>): Use it.
5865 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5868 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5869 frame if argument is passed on stack.
5871 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5874 * ree.c (combine_reaching_defs): Optimize also
5875 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5876 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5878 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5881 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5883 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5886 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5887 after they are computed.
5889 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5891 PR tree-optimization/83695
5892 * gimple-loop-linterchange.cc
5893 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5894 reset cached scev information after interchange.
5895 (pass_linterchange::execute): Remove call to scev_reset_htab.
5897 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5899 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5900 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5901 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5902 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5903 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5904 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5905 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5906 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5907 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5908 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5909 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5910 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5911 (V_lane_reg): Likewise.
5912 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5914 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5915 (vfmal_lane_low<mode>_intrinsic,
5916 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5917 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5918 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5919 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5920 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5921 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5923 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5925 * config/arm/arm-cpus.in (fp16fml): New feature.
5926 (ALL_SIMD): Add fp16fml.
5927 (armv8.2-a): Add fp16fml as an option.
5928 (armv8.3-a): Likewise.
5929 (armv8.4-a): Add fp16fml as part of fp16.
5930 * config/arm/arm.h (TARGET_FP16FML): Define.
5931 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5933 * config/arm/arm-modes.def (V2HF): Define.
5934 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5935 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5936 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5937 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5938 vfmsl_low, vfmsl_high): New set of builtins.
5939 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5940 (vfml_op): New code attribute.
5941 (VFMLHALVES): New int iterator.
5942 (VFML, VFMLSEL): New mode attributes.
5943 (V_reg): Define mapping for V2HF.
5944 (V_hi, V_lo): New mode attributes.
5945 (VF_constraint): Likewise.
5946 (vfml_half, vfml_half_selector): New int attributes.
5947 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5949 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5950 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5952 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5953 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5954 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5955 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5957 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5958 Document new effective target and option set.
5960 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5962 * config/arm/arm-cpus.in (armv8_4): New feature.
5963 (ARMv8_4a): New fgroup.
5964 (armv8.4-a): New arch.
5965 * config/arm/arm-tables.opt: Regenerate.
5966 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5967 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5968 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5969 Add matching rules for -march=armv8.4-a and extensions.
5970 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5972 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5975 * config/rx/rx.md (BW): New mode attribute.
5976 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5978 2018-01-11 Richard Biener <rguenther@suse.de>
5980 PR tree-optimization/83435
5981 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5982 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5983 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5985 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5986 Alan Hayward <alan.hayward@arm.com>
5987 David Sherwood <david.sherwood@arm.com>
5989 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5991 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5992 (aarch64_print_address_internal): Use it to check for a zero offset.
5994 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5995 Alan Hayward <alan.hayward@arm.com>
5996 David Sherwood <david.sherwood@arm.com>
5998 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5999 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6000 Return a poly_int64 rather than a HOST_WIDE_INT.
6001 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6002 rather than a HOST_WIDE_INT.
6003 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6004 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6005 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6006 final_offset from HOST_WIDE_INT to poly_int64.
6007 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6008 to_constant when getting the number of units in an Advanced SIMD
6010 (aarch64_builtin_vectorized_function): Check for a constant number
6012 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6014 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6015 attribute instead of GET_MODE_NUNITS.
6016 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6017 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6018 GET_MODE_SIZE for fixed-size registers.
6019 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6020 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6021 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6022 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6023 (aarch64_print_operand, aarch64_print_address_internal)
6024 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6025 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6026 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6027 Handle polynomial GET_MODE_SIZE.
6028 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6029 wider than SImode without modification.
6030 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6031 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6032 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6033 passing and returning SVE modes.
6034 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6035 rather than GEN_INT.
6036 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6037 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6038 (aarch64_allocate_and_probe_stack_space): Likewise.
6039 (aarch64_layout_frame): Cope with polynomial offsets.
6040 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6041 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6043 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6044 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6045 poly_int64 rather than a HOST_WIDE_INT.
6046 (aarch64_get_separate_components, aarch64_process_components)
6047 (aarch64_expand_prologue, aarch64_expand_epilogue)
6048 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6049 (aarch64_anchor_offset): New function, split out from...
6050 (aarch64_legitimize_address): ...here.
6051 (aarch64_builtin_vectorization_cost): Handle polynomial
6052 TYPE_VECTOR_SUBPARTS.
6053 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6055 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6056 number of elements from the PARALLEL rather than the mode.
6057 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6058 rather than GET_MODE_BITSIZE.
6059 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6060 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6061 (aarch64_expand_vec_perm_const_1): Handle polynomial
6062 d->perm.length () and d->perm elements.
6063 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6064 Apply to_constant to d->perm elements.
6065 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6066 polynomial CONST_VECTOR_NUNITS.
6067 (aarch64_move_pointer): Take amount as a poly_int64 rather
6069 (aarch64_progress_pointer): Avoid temporary variable.
6070 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6071 the mode attribute instead of GET_MODE.
6073 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6074 Alan Hayward <alan.hayward@arm.com>
6075 David Sherwood <david.sherwood@arm.com>
6077 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6078 x exists before using it.
6079 (aarch64_add_constant_internal): Rename to...
6080 (aarch64_add_offset_1): ...this. Replace regnum with separate
6081 src and dest rtxes. Handle the case in which they're different,
6082 including when the offset is zero. Replace scratchreg with an rtx.
6083 Use 2 additions if there is no spare register into which we can
6084 move a 16-bit constant.
6085 (aarch64_add_constant): Delete.
6086 (aarch64_add_offset): Replace reg with separate src and dest
6087 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6088 Use aarch64_add_offset_1.
6089 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6090 an rtx rather than an int. Take the delta as a poly_int64
6091 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6092 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6093 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6094 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6095 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6097 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6098 aarch64_add_constant.
6100 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6102 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6103 Use scalar_float_mode.
6105 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6107 * config/aarch64/aarch64-simd.md
6108 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6109 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6110 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6111 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6112 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6113 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6114 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6115 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6116 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6117 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6119 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6122 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6123 targ_options->x_arm_arch_string is non NULL.
6125 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6127 * config/aarch64/aarch64.h
6128 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6130 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6133 * expmed.c (emit_store_flag_force): Swap if const op0
6134 and change VOIDmode to mode of op0.
6136 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6138 PR rtl-optimization/83761
6139 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6140 than bytes to mode_for_size.
6142 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6145 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6146 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6149 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6152 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6153 when in layout mode.
6154 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6155 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6158 2018-01-10 Michael Collison <michael.collison@arm.com>
6160 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6161 * config/aarch64/aarch64-option-extension.def: Add
6162 AARCH64_OPT_EXTENSION of 'fp16fml'.
6163 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6164 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6165 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6166 * config/aarch64/constraints.md (Ui7): New constraint.
6167 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6168 (VFMLA_SEL_W): Ditto.
6171 (VFMLA16_LOW): New int iterator.
6172 (VFMLA16_HIGH): Ditto.
6173 (UNSPEC_FMLAL): New unspec.
6174 (UNSPEC_FMLSL): Ditto.
6175 (UNSPEC_FMLAL2): Ditto.
6176 (UNSPEC_FMLSL2): Ditto.
6177 (f16mac): New code attribute.
6178 * config/aarch64/aarch64-simd-builtins.def
6179 (aarch64_fmlal_lowv2sf): Ditto.
6180 (aarch64_fmlsl_lowv2sf): Ditto.
6181 (aarch64_fmlalq_lowv4sf): Ditto.
6182 (aarch64_fmlslq_lowv4sf): Ditto.
6183 (aarch64_fmlal_highv2sf): Ditto.
6184 (aarch64_fmlsl_highv2sf): Ditto.
6185 (aarch64_fmlalq_highv4sf): Ditto.
6186 (aarch64_fmlslq_highv4sf): Ditto.
6187 (aarch64_fmlal_lane_lowv2sf): Ditto.
6188 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6189 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6190 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6191 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6192 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6193 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6194 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6195 (aarch64_fmlal_lane_highv2sf): Ditto.
6196 (aarch64_fmlsl_lane_highv2sf): Ditto.
6197 (aarch64_fmlal_laneq_highv2sf): Ditto.
6198 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6199 (aarch64_fmlalq_lane_highv4sf): Ditto.
6200 (aarch64_fmlsl_lane_highv4sf): Ditto.
6201 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6202 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6203 * config/aarch64/aarch64-simd.md:
6204 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6205 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6206 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6207 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6208 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6209 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6210 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6211 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6212 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6213 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6214 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6215 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6216 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6217 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6218 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6219 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6220 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6221 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6222 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6223 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6224 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6225 (vfmlsl_low_u32): Ditto.
6226 (vfmlalq_low_u32): Ditto.
6227 (vfmlslq_low_u32): Ditto.
6228 (vfmlal_high_u32): Ditto.
6229 (vfmlsl_high_u32): Ditto.
6230 (vfmlalq_high_u32): Ditto.
6231 (vfmlslq_high_u32): Ditto.
6232 (vfmlal_lane_low_u32): Ditto.
6233 (vfmlsl_lane_low_u32): Ditto.
6234 (vfmlal_laneq_low_u32): Ditto.
6235 (vfmlsl_laneq_low_u32): Ditto.
6236 (vfmlalq_lane_low_u32): Ditto.
6237 (vfmlslq_lane_low_u32): Ditto.
6238 (vfmlalq_laneq_low_u32): Ditto.
6239 (vfmlslq_laneq_low_u32): Ditto.
6240 (vfmlal_lane_high_u32): Ditto.
6241 (vfmlsl_lane_high_u32): Ditto.
6242 (vfmlal_laneq_high_u32): Ditto.
6243 (vfmlsl_laneq_high_u32): Ditto.
6244 (vfmlalq_lane_high_u32): Ditto.
6245 (vfmlslq_lane_high_u32): Ditto.
6246 (vfmlalq_laneq_high_u32): Ditto.
6247 (vfmlslq_laneq_high_u32): Ditto.
6248 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6249 (AARCH64_FL_FOR_ARCH8_4): New.
6250 (AARCH64_ISA_F16FML): New ISA flag.
6251 (TARGET_F16FML): New feature flag for fp16fml.
6252 (doc/invoke.texi): Document new fp16fml option.
6254 2018-01-10 Michael Collison <michael.collison@arm.com>
6256 * config/aarch64/aarch64-builtins.c:
6257 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6258 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6259 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6260 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6261 (AARCH64_ISA_SHA3): New ISA flag.
6262 (TARGET_SHA3): New feature flag for sha3.
6263 * config/aarch64/iterators.md (sha512_op): New int attribute.
6264 (CRYPTO_SHA512): New int iterator.
6265 (UNSPEC_SHA512H): New unspec.
6266 (UNSPEC_SHA512H2): Ditto.
6267 (UNSPEC_SHA512SU0): Ditto.
6268 (UNSPEC_SHA512SU1): Ditto.
6269 * config/aarch64/aarch64-simd-builtins.def
6270 (aarch64_crypto_sha512hqv2di): New builtin.
6271 (aarch64_crypto_sha512h2qv2di): Ditto.
6272 (aarch64_crypto_sha512su0qv2di): Ditto.
6273 (aarch64_crypto_sha512su1qv2di): Ditto.
6274 (aarch64_eor3qv8hi): Ditto.
6275 (aarch64_rax1qv2di): Ditto.
6276 (aarch64_xarqv2di): Ditto.
6277 (aarch64_bcaxqv8hi): Ditto.
6278 * config/aarch64/aarch64-simd.md:
6279 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6280 (aarch64_crypto_sha512su0qv2di): Ditto.
6281 (aarch64_crypto_sha512su1qv2di): Ditto.
6282 (aarch64_eor3qv8hi): Ditto.
6283 (aarch64_rax1qv2di): Ditto.
6284 (aarch64_xarqv2di): Ditto.
6285 (aarch64_bcaxqv8hi): Ditto.
6286 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6287 (vsha512h2q_u64): Ditto.
6288 (vsha512su0q_u64): Ditto.
6289 (vsha512su1q_u64): Ditto.
6290 (veor3q_u16): Ditto.
6291 (vrax1q_u64): Ditto.
6293 (vbcaxq_u16): Ditto.
6294 * config/arm/types.md (crypto_sha512): New type attribute.
6295 (crypto_sha3): Ditto.
6296 (doc/invoke.texi): Document new sha3 option.
6298 2018-01-10 Michael Collison <michael.collison@arm.com>
6300 * config/aarch64/aarch64-builtins.c:
6301 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6302 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6303 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6304 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6305 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6306 (AARCH64_ISA_SM4): New ISA flag.
6307 (TARGET_SM4): New feature flag for sm4.
6308 * config/aarch64/aarch64-simd-builtins.def
6309 (aarch64_sm3ss1qv4si): Ditto.
6310 (aarch64_sm3tt1aq4si): Ditto.
6311 (aarch64_sm3tt1bq4si): Ditto.
6312 (aarch64_sm3tt2aq4si): Ditto.
6313 (aarch64_sm3tt2bq4si): Ditto.
6314 (aarch64_sm3partw1qv4si): Ditto.
6315 (aarch64_sm3partw2qv4si): Ditto.
6316 (aarch64_sm4eqv4si): Ditto.
6317 (aarch64_sm4ekeyqv4si): Ditto.
6318 * config/aarch64/aarch64-simd.md:
6319 (aarch64_sm3ss1qv4si): Ditto.
6320 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6321 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6322 (aarch64_sm4eqv4si): Ditto.
6323 (aarch64_sm4ekeyqv4si): Ditto.
6324 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6325 (sm3part_op): Ditto.
6326 (CRYPTO_SM3TT): Ditto.
6327 (CRYPTO_SM3PART): Ditto.
6328 (UNSPEC_SM3SS1): New unspec.
6329 (UNSPEC_SM3TT1A): Ditto.
6330 (UNSPEC_SM3TT1B): Ditto.
6331 (UNSPEC_SM3TT2A): Ditto.
6332 (UNSPEC_SM3TT2B): Ditto.
6333 (UNSPEC_SM3PARTW1): Ditto.
6334 (UNSPEC_SM3PARTW2): Ditto.
6335 (UNSPEC_SM4E): Ditto.
6336 (UNSPEC_SM4EKEY): Ditto.
6337 * config/aarch64/constraints.md (Ui2): New constraint.
6338 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6339 * config/arm/types.md (crypto_sm3): New type attribute.
6340 (crypto_sm4): Ditto.
6341 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6342 (vsm3tt1aq_u32): Ditto.
6343 (vsm3tt1bq_u32): Ditto.
6344 (vsm3tt2aq_u32): Ditto.
6345 (vsm3tt2bq_u32): Ditto.
6346 (vsm3partw1q_u32): Ditto.
6347 (vsm3partw2q_u32): Ditto.
6348 (vsm4eq_u32): Ditto.
6349 (vsm4ekeyq_u32): Ditto.
6350 (doc/invoke.texi): Document new sm4 option.
6352 2018-01-10 Michael Collison <michael.collison@arm.com>
6354 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6355 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6356 (AARCH64_FL_FOR_ARCH8_4): New.
6357 (AARCH64_FL_V8_4): New flag.
6358 (doc/invoke.texi): Document new armv8.4-a option.
6360 2018-01-10 Michael Collison <michael.collison@arm.com>
6362 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6363 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6364 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6365 * config/aarch64/aarch64-option-extension.def: Add
6366 AARCH64_OPT_EXTENSION of 'sha2'.
6367 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6368 (crypto): Disable sha2 and aes if crypto disabled.
6369 (crypto): Enable aes and sha2 if enabled.
6370 (simd): Disable sha2 and aes if simd disabled.
6371 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6373 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6374 (TARGET_SHA2): New feature flag for sha2.
6375 (TARGET_AES): New feature flag for aes.
6376 * config/aarch64/aarch64-simd.md:
6377 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6378 conditional on TARGET_AES.
6379 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6380 (aarch64_crypto_sha1hsi): Make pattern conditional
6382 (aarch64_crypto_sha1hv4si): Ditto.
6383 (aarch64_be_crypto_sha1hv4si): Ditto.
6384 (aarch64_crypto_sha1su1v4si): Ditto.
6385 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6386 (aarch64_crypto_sha1su0v4si): Ditto.
6387 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6388 (aarch64_crypto_sha256su0v4si): Ditto.
6389 (aarch64_crypto_sha256su1v4si): Ditto.
6390 (doc/invoke.texi): Document new aes and sha2 options.
6392 2018-01-10 Martin Sebor <msebor@redhat.com>
6394 PR tree-optimization/83781
6395 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6398 2018-01-11 Martin Sebor <msebor@gmail.com>
6399 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6401 PR tree-optimization/83501
6402 PR tree-optimization/81703
6404 * tree-ssa-strlen.c (get_string_cst): Rename...
6405 (get_string_len): ...to this. Handle global constants.
6406 (handle_char_store): Adjust.
6408 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6409 Jim Wilson <jimw@sifive.com>
6411 * config/riscv/riscv-protos.h (riscv_output_return): New.
6412 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6413 (riscv_attribute_table, riscv_output_return),
6414 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6415 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6416 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6418 (riscv_expand_prologue): Add early return for naked function.
6419 (riscv_expand_epilogue): Likewise.
6420 (riscv_function_ok_for_sibcall): Return false for naked function.
6421 (riscv_set_current_function): New.
6422 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6423 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6424 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6425 * doc/extend.texi (RISC-V Function Attributes): New.
6427 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6429 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6430 check for 128-bit long double before checking TCmode.
6431 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6432 128-bit long doubles before checking TFmode or TCmode.
6433 (FLOAT128_IBM_P): Likewise.
6435 2018-01-10 Martin Sebor <msebor@redhat.com>
6437 PR tree-optimization/83671
6438 * builtins.c (c_strlen): Unconditionally return zero for the empty
6440 Use -Warray-bounds for warnings.
6441 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6442 for non-constant array indices with COMPONENT_REF, arrays of
6443 arrays, and pointers to arrays.
6444 (gimple_fold_builtin_strlen): Determine and set length range for
6445 non-constant character arrays.
6447 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6450 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6453 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6455 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6457 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6460 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6461 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6462 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6463 indexed_or_indirect_operand predicate.
6464 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6465 (*vsx_le_perm_load_v8hi): Likewise.
6466 (*vsx_le_perm_load_v16qi): Likewise.
6467 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6468 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6469 (*vsx_le_perm_store_v8hi): Likewise.
6470 (*vsx_le_perm_store_v16qi): Likewise.
6471 (eight unnamed splitters): Likewise.
6473 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6475 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6476 * config/rs6000/emmintrin.h: Likewise.
6477 * config/rs6000/mmintrin.h: Likewise.
6478 * config/rs6000/xmmintrin.h: Likewise.
6480 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6483 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6485 * tree.c (tree_nop_conversion): Return true for location wrapper
6487 (maybe_wrap_with_location): New function.
6488 (selftest::check_strip_nops): New function.
6489 (selftest::test_location_wrappers): New function.
6490 (selftest::tree_c_tests): Call it.
6491 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6492 (maybe_wrap_with_location): New decl.
6493 (EXPR_LOCATION_WRAPPER_P): New macro.
6494 (location_wrapper_p): New inline function.
6495 (tree_strip_any_location_wrapper): New inline function.
6497 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6500 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6501 stack_realign_offset for the largest alignment of stack slot
6503 (ix86_find_max_used_stack_alignment): New function.
6504 (ix86_finalize_stack_frame_flags): Use it. Set
6505 max_used_stack_alignment if we don't realign stack.
6506 * config/i386/i386.h (machine_function): Add
6507 max_used_stack_alignment.
6509 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6511 * config/arm/arm.opt (-mbranch-cost): New option.
6512 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6515 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6518 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6519 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6521 2018-01-10 Richard Biener <rguenther@suse.de>
6524 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6525 early out so it also covers the case where we have a non-NULL
6528 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6530 PR tree-optimization/83753
6531 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6532 for non-strided grouped accesses if the number of elements is 1.
6534 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6537 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6538 * i386.h (TARGET_USE_GATHER): Define.
6539 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6541 2018-01-10 Martin Liska <mliska@suse.cz>
6544 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6545 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6547 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6548 CLEANUP_NO_PARTITIONING is not set.
6550 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6552 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6553 for vectors, as a partial revert of r254296.
6554 * rtl.h (const_vec_p): Delete.
6555 (const_vec_duplicate_p): Don't test for vector CONSTs.
6556 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6557 * expmed.c (make_tree): Likewise.
6560 * common.md (E, F): Use CONSTANT_P instead of checking for
6562 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6563 checking for CONST_VECTOR.
6565 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6568 * predict.c (force_edge_cold): Handle in more sane way edges
6571 2018-01-09 Carl Love <cel@us.ibm.com>
6573 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6575 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6576 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6577 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6578 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6579 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6580 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6581 * config/rs6000/rs6000-protos.h: Add extern defition for
6582 rs6000_generate_float2_double_code.
6583 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6585 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6586 (float2_v2df): Add define_expand.
6588 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6591 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6592 op_mode in the force_to_mode call.
6594 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6596 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6597 instead of checking each element individually.
6598 (aarch64_evpc_uzp): Likewise.
6599 (aarch64_evpc_zip): Likewise.
6600 (aarch64_evpc_ext): Likewise.
6601 (aarch64_evpc_rev): Likewise.
6602 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6603 instead of checking each element individually. Return true without
6605 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6606 whether all selected elements come from the same input, instead of
6607 checking each element individually. Remove calls to gen_rtx_REG,
6608 start_sequence and end_sequence and instead assert that no rtl is
6611 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6613 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6614 order of HIGH and CONST checks.
6616 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6618 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6619 if the destination isn't an SSA_NAME.
6621 2018-01-09 Richard Biener <rguenther@suse.de>
6623 PR tree-optimization/83668
6624 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6626 (canonicalize_loop_form): ... here, renamed from ...
6627 (canonicalize_loop_closed_ssa_form): ... this and amended to
6628 swap successor edges for loop exit blocks to make us use
6629 the RPO order we need for initial schedule generation.
6631 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6633 PR tree-optimization/64811
6634 * match.pd: When optimizing comparisons with Inf, avoid
6635 introducing or losing exceptions from comparisons with NaN.
6637 2018-01-09 Martin Liska <mliska@suse.cz>
6640 * asan.c (shadow_mem_size): Add gcc_assert.
6642 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6644 Don't save registers in main().
6647 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6648 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6649 * config/avr/avr.c (avr_set_current_function): Don't error if
6650 naked, OS_task or OS_main are specified at the same time.
6651 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6653 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6655 * common/config/avr/avr-common.c (avr_option_optimization_table):
6656 Switch on -mmain-is-OS_task for optimizing compilations.
6658 2018-01-09 Richard Biener <rguenther@suse.de>
6660 PR tree-optimization/83572
6661 * graphite.c: Include cfganal.h.
6662 (graphite_transform_loops): Connect infinite loops to exit
6663 and remove fake edges at the end.
6665 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6667 * ipa-inline.c (edge_badness): Revert accidental checkin.
6669 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6672 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6673 symbols; not inline clones.
6675 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6678 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6679 hard registers. Formatting fixes.
6681 PR preprocessor/83722
6682 * gcc.c (try_generate_repro): Pass
6683 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6684 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6687 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6688 Kito Cheng <kito.cheng@gmail.com>
6690 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6691 (riscv_leaf_function_p): Delete.
6692 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6694 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6696 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6698 (do_ifelse): New function.
6699 (do_isel): New function.
6700 (do_sub3): New function.
6701 (do_add3): New function.
6702 (do_load_mask_compare): New function.
6703 (do_overlap_load_compare): New function.
6704 (expand_compare_loop): New function.
6705 (expand_block_compare): Call expand_compare_loop() when appropriate.
6706 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6708 (-mblock-compare-inline-loop-limit): New option.
6710 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6713 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6714 Reverse order of second and third operands in first alternative.
6715 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6716 of first and second elements in UNSPEC_VPERMR vector.
6717 (altivec_expand_vec_perm_le): Likewise.
6719 2018-01-08 Jeff Law <law@redhat.com>
6721 PR rtl-optimizatin/81308
6722 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6723 (process_switch): If group_case_labels makes a change, then set
6725 (pass_convert_switch::execute): If a switch is converted, then
6726 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6728 PR rtl-optimization/81308
6729 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6732 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6734 PR target/83663 - Revert r255946
6735 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6736 generation for cases where splatting a value is not useful.
6737 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6738 across a vec_duplicate and a paradoxical subreg forming a vector
6739 mode to a vec_concat.
6741 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6743 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6744 -march=armv8.3-a variants.
6745 * config/arm/t-multilib: Likewise.
6746 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6748 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6750 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6752 (cceq_ior_compare_complement): Give it a name so I can use it, and
6753 change boolean_or_operator predicate to boolean_operator so it can
6754 be used to generate a crand.
6755 (eqne): New code iterator.
6756 (bd/bd_neg): New code_attrs.
6757 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6758 a single define_insn.
6759 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6760 decrement (bdnzt/bdnzf/bdzt/bdzf).
6761 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6762 with the new names of the branch decrement patterns, and added the
6763 names of the branch decrement conditional patterns.
6765 2018-01-08 Richard Biener <rguenther@suse.de>
6767 PR tree-optimization/83563
6768 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6771 2018-01-08 Richard Biener <rguenther@suse.de>
6774 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6776 2018-01-08 Richard Biener <rguenther@suse.de>
6778 PR tree-optimization/83685
6779 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6780 references to abnormals.
6782 2018-01-08 Richard Biener <rguenther@suse.de>
6785 * dwarf2out.c (output_indirect_strings): Handle empty
6786 skeleton_debug_str_hash.
6787 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6789 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6791 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6792 (emit_store_direct): Likewise.
6793 (arc_trampoline_adjust_address): Likewise.
6794 (arc_asm_trampoline_template): New function.
6795 (arc_initialize_trampoline): Use asm_trampoline_template.
6796 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6797 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6798 * config/arc/arc.md (flush_icache): Delete pattern.
6800 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6802 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6803 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6806 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6809 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6810 by not USED_FOR_TARGET.
6811 (make_pass_resolve_sw_modes): Likewise.
6813 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6815 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6818 2018-01-08 Richard Biener <rguenther@suse.de>
6821 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6823 2018-01-08 Richard Biener <rguenther@suse.de>
6826 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6828 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6831 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6832 basic blocks with a small number of successors.
6833 (convert_control_dep_chain_into_preds): Improve handling of
6835 (dump_predicates): Split apart into...
6836 (dump_pred_chain): ...here...
6837 (dump_pred_info): ...and here.
6838 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6839 (can_chain_union_be_invalidated_p): Improve check for invalidation
6841 (uninit_uses_cannot_happen): Avoid unnecessary if
6842 convert_control_dep_chain_into_preds yielded nothing.
6844 2018-01-06 Martin Sebor <msebor@redhat.com>
6846 PR tree-optimization/83640
6847 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6848 subtracting negative offset from size.
6849 (builtin_access::overlap): Adjust offset bounds of the access to fall
6850 within the size of the object if possible.
6852 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6854 PR rtl-optimization/83699
6855 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6856 extract_bit_field_as_subreg to cases in which the extracted
6857 value is also a vector.
6859 * lra-constraints.c (process_alt_operands): Test for the equivalence
6860 substitutions when detecting a possible reload cycle.
6862 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6865 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6866 by default if flag_selective_schedling{,2}. Formatting fixes.
6868 PR rtl-optimization/83682
6869 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6870 if it has non-VECTOR_MODE element mode.
6871 (vec_duplicate_p): Likewise.
6874 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6875 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6877 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6880 * config/i386/i386-builtin.def
6881 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6882 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6883 Require also OPTION_MASK_ISA_AVX512F in addition to
6884 OPTION_MASK_ISA_GFNI.
6885 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6886 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6887 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6888 to OPTION_MASK_ISA_GFNI.
6889 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6890 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6891 OPTION_MASK_ISA_AVX512BW.
6892 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6893 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6894 addition to OPTION_MASK_ISA_GFNI.
6895 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6896 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6897 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6898 to OPTION_MASK_ISA_GFNI.
6899 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6900 a requirement for all ISAs rather than any of them with a few
6902 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6904 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6905 bitmasks to be enabled with 3 exceptions, instead of requiring any
6906 enabled ISA with lots of exceptions.
6907 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6908 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6909 Change avx512bw in isa attribute to avx512f.
6910 * config/i386/sgxintrin.h: Add license boilerplate.
6911 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6912 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6913 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6914 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6916 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6917 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6918 temporarily sse2 rather than sse if not enabled already.
6921 * config/i386/sse.md (VI248_VLBW): Rename to ...
6922 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6923 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6924 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6925 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6926 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6927 mode iterator instead of VI248_VLBW.
6929 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6931 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6932 (record_modified): Skip clobbers; add debug output.
6933 (param_change_prob): Use sreal frequencies.
6935 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6937 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6938 punt for user-aligned variables.
6940 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6942 * tree-chrec.c (chrec_contains_symbols): Return true for
6945 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6948 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6949 of (x|y) == x for BICS pattern.
6951 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6953 PR tree-optimization/83605
6954 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6955 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6958 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6960 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6961 * config/epiphany/rtems.h: New file.
6963 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6964 Uros Bizjak <ubizjak@gmail.com>
6967 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6968 QIreg_operand instead of register_operand predicate.
6969 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6970 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6971 comments instead of -fmitigate[-_]rop.
6973 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6976 * cgraphunit.c (symbol_table::compile): Switch to text_section
6977 before calling assembly_start debug hook.
6978 * run-rtl-passes.c (run_rtl_passes): Likewise.
6981 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6983 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6984 range_int_cst_p rather than !symbolic_range_p before calling
6985 extract_range_from_multiplicative_op_1.
6987 2018-01-04 Jeff Law <law@redhat.com>
6989 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6990 redundant test in assertion.
6992 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6994 * doc/rtl.texi: Document machine_mode wrapper classes.
6996 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6998 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7001 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7003 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7004 the VEC_PERM_EXPR fold to fail.
7006 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7009 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7010 to switched_sections.
7012 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7015 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7018 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7021 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7022 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7024 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7027 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7028 is BLKmode and bitpos not zero or mode change is needed.
7030 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7033 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7036 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7039 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7040 instead of MULT rtx. Update all corresponding splitters.
7042 (*ssub<modesuffix>): Ditto.
7044 (*cmp_sadd_di): Update split patterns.
7045 (*cmp_sadd_si): Ditto.
7046 (*cmp_sadd_sidi): Ditto.
7047 (*cmp_ssub_di): Ditto.
7048 (*cmp_ssub_si): Ditto.
7049 (*cmp_ssub_sidi): Ditto.
7050 * config/alpha/predicates.md (const23_operand): New predicate.
7051 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7052 Look for ASHIFT, not MULT inner operand.
7053 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7055 2018-01-04 Martin Liska <mliska@suse.cz>
7057 PR gcov-profile/83669
7058 * gcov.c (output_intermediate_file): Add version to intermediate
7060 * doc/gcov.texi: Document new field 'version' in intermediate
7061 file format. Fix location of '-k' option of gcov command.
7063 2018-01-04 Martin Liska <mliska@suse.cz>
7066 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7068 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7070 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7072 2018-01-03 Martin Sebor <msebor@redhat.com>
7074 PR tree-optimization/83655
7075 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7076 checking calls with invalid arguments.
7078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7080 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7081 (vectorizable_mask_load_store): Delete.
7082 (vectorizable_call): Return false for masked loads and stores.
7083 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7084 instead of gimple_assign_rhs1.
7085 (vectorizable_load): Handle IFN_MASK_LOAD.
7086 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7090 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7092 (vectorizable_mask_load_store): ...here.
7093 (vectorizable_load): ...and here.
7095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7097 * tree-vect-stmts.c (vect_build_all_ones_mask)
7098 (vect_build_zero_merge_argument): New functions, split out from...
7099 (vectorizable_load): ...here.
7101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7103 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7105 (vectorizable_mask_load_store): ...here.
7106 (vectorizable_store): ...and here.
7108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7110 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7112 (vectorizable_mask_load_store): ...here.
7114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7116 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7117 (vect_model_store_cost): Take a vec_load_store_type instead of a
7119 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7120 (vect_model_store_cost): Take a vec_load_store_type instead of a
7122 (vectorizable_mask_load_store): Update accordingly.
7123 (vectorizable_store): Likewise.
7124 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7128 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7129 IFN_MASK_LOAD calls here rather than...
7130 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7133 Alan Hayward <alan.hayward@arm.com>
7134 David Sherwood <david.sherwood@arm.com>
7136 * expmed.c (extract_bit_field_1): For vector extracts,
7137 fall back to extract_bit_field_as_subreg if vec_extract
7140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7141 Alan Hayward <alan.hayward@arm.com>
7142 David Sherwood <david.sherwood@arm.com>
7144 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7145 they are variable or constant sized.
7146 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7147 slots for constant-sized data.
7149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7150 Alan Hayward <alan.hayward@arm.com>
7151 David Sherwood <david.sherwood@arm.com>
7153 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7154 handling COND_EXPRs with boolean comparisons, try to find a better
7155 basis for the mask type than the boolean itself.
7157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7159 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7160 is calculated and how it can be overridden.
7161 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7162 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7164 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7168 Alan Hayward <alan.hayward@arm.com>
7169 David Sherwood <david.sherwood@arm.com>
7171 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7172 Remove the mode argument.
7173 (aarch64_simd_valid_immediate): Remove the mode and inverse
7175 * config/aarch64/iterators.md (bitsize): New iterator.
7176 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7177 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7178 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7179 aarch64_simd_valid_immediate.
7180 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7181 (aarch64_reg_or_bic_imm): Likewise.
7182 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7183 with an insn_type enum and msl with a modifier_type enum.
7184 Replace element_width with a scalar_mode. Change the shift
7185 to unsigned int. Add constructors for scalar_float_mode and
7186 scalar_int_mode elements.
7187 (aarch64_vect_float_const_representable_p): Delete.
7188 (aarch64_can_const_movi_rtx_p)
7189 (aarch64_simd_scalar_immediate_valid_for_move)
7190 (aarch64_simd_make_constant): Update call to
7191 aarch64_simd_valid_immediate.
7192 (aarch64_advsimd_valid_immediate_hs): New function.
7193 (aarch64_advsimd_valid_immediate): Likewise.
7194 (aarch64_simd_valid_immediate): Remove mode and inverse
7195 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7196 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7197 and aarch64_float_const_representable_p on the result.
7198 (aarch64_output_simd_mov_immediate): Remove mode argument.
7199 Update call to aarch64_simd_valid_immediate and use of
7200 simd_immediate_info.
7201 (aarch64_output_scalar_simd_mov_immediate): Update call
7204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7205 Alan Hayward <alan.hayward@arm.com>
7206 David Sherwood <david.sherwood@arm.com>
7208 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7209 (mode_nunits): Likewise CONST_MODE_NUNITS.
7210 * machmode.def (ADJUST_NUNITS): Document.
7211 * genmodes.c (mode_data::need_nunits_adj): New field.
7212 (blank_mode): Update accordingly.
7213 (adj_nunits): New variable.
7214 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7216 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7217 listed in adj_nunits.
7218 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7219 listed in adj_nunits. Don't emit case statements for such modes.
7220 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7221 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7222 nothing if adj_nunits is nonnull.
7223 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7224 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7225 (emit_mode_fbit): Update use of print_maybe_const_decl.
7226 (emit_move_size): Likewise. Treat the array as non-const
7228 (emit_mode_adjustments): Handle adj_nunits.
7230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7232 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7233 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7234 (VECTOR_MODES): Use it.
7235 (make_vector_modes): Take the prefix as an argument.
7237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7238 Alan Hayward <alan.hayward@arm.com>
7239 David Sherwood <david.sherwood@arm.com>
7241 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7242 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7243 for MODE_VECTOR_BOOL.
7244 * machmode.def (VECTOR_BOOL_MODE): Document.
7245 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7246 (make_vector_bool_mode): New function.
7247 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7249 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7250 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7252 * stor-layout.c (int_mode_for_mode): Likewise.
7253 * tree.c (build_vector_type_for_mode): Likewise.
7254 * varasm.c (output_constant_pool_2): Likewise.
7255 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7256 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7257 for MODE_VECTOR_BOOL.
7258 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7259 of mode class checks.
7260 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7261 instead of a list of mode class checks.
7262 (expand_vector_scalar_condition): Likewise.
7263 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7266 Alan Hayward <alan.hayward@arm.com>
7267 David Sherwood <david.sherwood@arm.com>
7269 * machmode.h (mode_size): Change from unsigned short to
7271 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7272 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7273 or if measurement_type is not polynomial.
7274 (fixed_size_mode::includes_p): Check for constant-sized modes.
7275 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7276 return a poly_uint16 rather than an unsigned short.
7277 (emit_mode_size): Change the type of mode_size from unsigned short
7278 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7279 (emit_mode_adjustments): Cope with polynomial vector sizes.
7280 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7282 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7284 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7285 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7286 * caller-save.c (setup_save_areas): Likewise.
7287 (replace_reg_with_saved_mem): Likewise.
7288 * calls.c (emit_library_call_value_1): Likewise.
7289 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7290 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7291 (gen_lowpart_for_combine): Likewise.
7292 * convert.c (convert_to_integer_1): Likewise.
7293 * cse.c (equiv_constant, cse_insn): Likewise.
7294 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7295 (cselib_subst_to_values): Likewise.
7296 * dce.c (word_dce_process_block): Likewise.
7297 * df-problems.c (df_word_lr_mark_ref): Likewise.
7298 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7299 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7300 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7301 (rtl_for_decl_location): Likewise.
7302 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7303 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7304 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7305 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7306 (expand_expr_real_1): Likewise.
7307 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7308 (pad_below): Likewise.
7309 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7310 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7311 * ira.c (get_subreg_tracking_sizes): Likewise.
7312 * ira-build.c (ira_create_allocno_objects): Likewise.
7313 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7314 (ira_sort_regnos_for_alter_reg): Likewise.
7315 * ira-costs.c (record_operand_costs): Likewise.
7316 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7317 (resolve_simple_move): Likewise.
7318 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7319 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7320 (lra_constraints): Likewise.
7321 (CONST_POOL_OK_P): Reject variable-sized modes.
7322 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7323 (add_pseudo_to_slot, lra_spill): Likewise.
7324 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7325 * optabs-query.c (get_best_extraction_insn): Likewise.
7326 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7327 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7328 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7329 * recog.c (offsettable_address_addr_space_p): Likewise.
7330 * regcprop.c (maybe_mode_change): Likewise.
7331 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7332 * regrename.c (build_def_use): Likewise.
7333 * regstat.c (dump_reg_info): Likewise.
7334 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7335 (find_reloads, find_reloads_subreg_address): Likewise.
7336 * reload1.c (eliminate_regs_1): Likewise.
7337 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7338 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7339 (simplify_binary_operation_1, simplify_subreg): Likewise.
7340 * targhooks.c (default_function_arg_padding): Likewise.
7341 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7342 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7343 (verify_gimple_assign_ternary): Likewise.
7344 * tree-inline.c (estimate_move_cost): Likewise.
7345 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7346 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7347 (get_address_cost_ainc): Likewise.
7348 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7349 (vect_supportable_dr_alignment): Likewise.
7350 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7351 (vectorizable_reduction): Likewise.
7352 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7353 (vectorizable_operation, vectorizable_load): Likewise.
7354 * tree.c (build_same_sized_truth_vector_type): Likewise.
7355 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7356 * var-tracking.c (emit_note_insn_var_location): Likewise.
7357 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7358 (ADDR_VEC_ALIGN): Likewise.
7360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7361 Alan Hayward <alan.hayward@arm.com>
7362 David Sherwood <david.sherwood@arm.com>
7364 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7366 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7367 or if measurement_type is polynomial.
7368 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7369 * combine.c (make_extraction): Likewise.
7370 * dse.c (find_shift_sequence): Likewise.
7371 * dwarf2out.c (mem_loc_descriptor): Likewise.
7372 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7373 (extract_bit_field, extract_low_bits): Likewise.
7374 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7375 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7376 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7377 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7378 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7379 * reload.c (find_reloads): Likewise.
7380 * reload1.c (alter_reg): Likewise.
7381 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7382 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7383 * tree-if-conv.c (predicate_mem_writes): Likewise.
7384 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7385 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7386 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7387 * valtrack.c (dead_debug_insert_temp): Likewise.
7388 * varasm.c (mergeable_constant_section): Likewise.
7389 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7392 Alan Hayward <alan.hayward@arm.com>
7393 David Sherwood <david.sherwood@arm.com>
7395 * expr.c (expand_assignment): Cope with polynomial mode sizes
7396 when assigning to a CONCAT.
7398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7399 Alan Hayward <alan.hayward@arm.com>
7400 David Sherwood <david.sherwood@arm.com>
7402 * machmode.h (mode_precision): Change from unsigned short to
7404 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7406 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7407 or if measurement_type is not polynomial.
7408 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7409 in which the mode is already known to be a scalar_int_mode.
7410 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7411 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7413 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7414 for GET_MODE_PRECISION.
7415 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7416 for GET_MODE_PRECISION.
7417 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7419 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7420 (expand_field_assignment, make_extraction): Likewise.
7421 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7422 (get_last_value): Likewise.
7423 * convert.c (convert_to_integer_1): Likewise.
7424 * cse.c (cse_insn): Likewise.
7425 * expr.c (expand_expr_real_1): Likewise.
7426 * lra-constraints.c (simplify_operand_subreg): Likewise.
7427 * optabs-query.c (can_atomic_load_p): Likewise.
7428 * optabs.c (expand_atomic_load): Likewise.
7429 (expand_atomic_store): Likewise.
7430 * ree.c (combine_reaching_defs): Likewise.
7431 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7432 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7433 * tree.h (type_has_mode_precision_p): Likewise.
7434 * ubsan.c (instrument_si_overflow): Likewise.
7436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7437 Alan Hayward <alan.hayward@arm.com>
7438 David Sherwood <david.sherwood@arm.com>
7440 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7441 polynomial numbers of units.
7442 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7443 (valid_vector_subparts_p): New function.
7444 (build_vector_type): Remove temporary shim and take the number
7445 of units as a poly_uint64 rather than an int.
7446 (build_opaque_vector_type): Take the number of units as a
7447 poly_uint64 rather than an int.
7448 * tree.c (build_vector_from_ctor): Handle polynomial
7449 TYPE_VECTOR_SUBPARTS.
7450 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7451 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7452 (build_vector_from_val): If the number of units is variable,
7453 use build_vec_duplicate_cst for constant operands and
7454 VEC_DUPLICATE_EXPR otherwise.
7455 (make_vector_type): Remove temporary is_constant ().
7456 (build_vector_type, build_opaque_vector_type): Take the number of
7457 units as a poly_uint64 rather than an int.
7458 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7460 * cfgexpand.c (expand_debug_expr): Likewise.
7461 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7462 (store_constructor, expand_expr_real_1): Likewise.
7463 (const_scalar_mask_from_tree): Likewise.
7464 * fold-const-call.c (fold_const_reduction): Likewise.
7465 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7466 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7467 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7468 (fold_relational_const): Likewise.
7469 (native_interpret_vector): Likewise. Change the size from an
7470 int to an unsigned int.
7471 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7472 TYPE_VECTOR_SUBPARTS.
7473 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7474 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7475 duplicating a non-constant operand into a variable-length vector.
7476 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7477 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7478 * ipa-icf.c (sem_variable::equals): Likewise.
7479 * match.pd: Likewise.
7480 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7481 * print-tree.c (print_node): Likewise.
7482 * stor-layout.c (layout_type): Likewise.
7483 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7484 * tree-cfg.c (verify_gimple_comparison): Likewise.
7485 (verify_gimple_assign_binary): Likewise.
7486 (verify_gimple_assign_ternary): Likewise.
7487 (verify_gimple_assign_single): Likewise.
7488 * tree-pretty-print.c (dump_generic_node): Likewise.
7489 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7490 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7491 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7492 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7493 (vect_shift_permute_load_chain): Likewise.
7494 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7495 (expand_vector_condition, optimize_vector_constructor): Likewise.
7496 (lower_vec_perm, get_compute_type): Likewise.
7497 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7498 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7499 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7500 (vect_recog_mask_conversion_pattern): Likewise.
7501 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7502 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7503 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7504 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7505 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7506 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7507 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7508 (supportable_widening_operation): Likewise.
7509 (supportable_narrowing_operation): Likewise.
7510 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7512 * varasm.c (output_constant): Likewise.
7514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7515 Alan Hayward <alan.hayward@arm.com>
7516 David Sherwood <david.sherwood@arm.com>
7518 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7519 so that both the length == 3 and length != 3 cases set up their
7520 own permute vectors. Add comments explaining why we know the
7521 number of elements is constant.
7522 (vect_permute_load_chain): Likewise.
7524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7525 Alan Hayward <alan.hayward@arm.com>
7526 David Sherwood <david.sherwood@arm.com>
7528 * machmode.h (mode_nunits): Change from unsigned char to
7530 (ONLY_FIXED_SIZE_MODES): New macro.
7531 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7532 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7533 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7535 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7536 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7537 or if measurement_type is not polynomial.
7538 * genmodes.c (ZERO_COEFFS): New macro.
7539 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7541 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7542 Use ZERO_COEFFS when emitting initializers.
7543 * data-streamer.h (bp_pack_poly_value): New function.
7544 (bp_unpack_poly_value): Likewise.
7545 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7546 for GET_MODE_NUNITS.
7547 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7548 for GET_MODE_NUNITS.
7549 * tree.c (make_vector_type): Remove temporary shim and make
7550 the real function take the number of units as a poly_uint64
7552 (build_vector_type_for_mode): Handle polynomial nunits.
7553 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7554 * emit-rtl.c (const_vec_series_p_1): Likewise.
7555 (gen_rtx_CONST_VECTOR): Likewise.
7556 * fold-const.c (test_vec_duplicate_folding): Likewise.
7557 * genrecog.c (validate_pattern): Likewise.
7558 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7559 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7560 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7561 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7562 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7563 * rtlanal.c (subreg_get_info): Likewise.
7564 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7565 (vect_grouped_load_supported): Likewise.
7566 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7567 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7568 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7569 (simplify_const_unary_operation, simplify_binary_operation_1)
7570 (simplify_const_binary_operation, simplify_ternary_operation)
7571 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7572 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7573 instead of CONST_VECTOR_NUNITS.
7574 * varasm.c (output_constant_pool_2): Likewise.
7575 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7576 explicit-encoded elements in the XVEC for variable-length vectors.
7578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7580 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7583 Alan Hayward <alan.hayward@arm.com>
7584 David Sherwood <david.sherwood@arm.com>
7586 * coretypes.h (fixed_size_mode): Declare.
7587 (fixed_size_mode_pod): New typedef.
7588 * builtins.h (target_builtins::x_apply_args_mode)
7589 (target_builtins::x_apply_result_mode): Change type to
7590 fixed_size_mode_pod.
7591 * builtins.c (apply_args_size, apply_result_size, result_vector)
7592 (expand_builtin_apply_args_1, expand_builtin_apply)
7593 (expand_builtin_return): Update accordingly.
7595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7597 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7598 * cselib.c (cselib_hash_rtx): Likewise.
7599 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7600 CONST_VECTOR encoding.
7602 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7603 Jeff Law <law@redhat.com>
7606 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7607 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7608 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7609 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7612 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7613 explicitly probe *sp in a noreturn function if there were any callee
7614 register saves or frame pointer is needed.
7616 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7619 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7620 BLKmode for ternary, binary or unary expressions.
7623 * var-tracking.c (delete_vta_debug_insn): New inline function.
7624 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7625 insns from get_insns () to NULL instead of each bb separately.
7626 Use delete_vta_debug_insn. No longer static.
7627 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7628 delete_vta_debug_insns callers.
7629 * rtl.h (delete_vta_debug_insns): Declare.
7630 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7631 instead of variable_tracking_main.
7633 2018-01-03 Martin Sebor <msebor@redhat.com>
7635 PR tree-optimization/83603
7636 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7637 arguments past the endof the argument list in functions declared
7638 without a prototype.
7639 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7640 Avoid checking when arguments are null.
7642 2018-01-03 Martin Sebor <msebor@redhat.com>
7645 * doc/extend.texi (attribute const): Fix a typo.
7646 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7647 issuing -Wsuggest-attribute for void functions.
7649 2018-01-03 Martin Sebor <msebor@redhat.com>
7651 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7652 offset_int::from instead of wide_int::to_shwi.
7653 (maybe_diag_overlap): Remove assertion.
7654 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7655 * gimple-ssa-sprintf.c (format_directive): Same.
7656 (parse_directive): Same.
7657 (sprintf_dom_walker::compute_format_length): Same.
7658 (try_substitute_return_value): Same.
7660 2018-01-03 Jeff Law <law@redhat.com>
7663 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7664 non-constant residual for zero at runtime and avoid probing in
7665 that case. Reorganize code for trailing problem to mirror handling
7668 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7670 PR tree-optimization/83501
7671 * tree-ssa-strlen.c (get_string_cst): New.
7672 (handle_char_store): Call get_string_cst.
7674 2018-01-03 Martin Liska <mliska@suse.cz>
7676 PR tree-optimization/83593
7677 * tree-ssa-strlen.c: Include tree-cfg.h.
7678 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7679 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7680 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7682 (strlen_dom_walker::before_dom_children): Call
7683 gimple_purge_dead_eh_edges. Dump tranformation with details
7685 (strlen_dom_walker::before_dom_children): Update call by adding
7686 new argument cleanup_eh.
7687 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7689 2018-01-03 Martin Liska <mliska@suse.cz>
7692 * cif-code.def (VARIADIC_THUNK): New enum value.
7693 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7696 2018-01-03 Jan Beulich <jbeulich@suse.com>
7698 * sse.md (mov<mode>_internal): Tighten condition for when to use
7699 vmovdqu<ssescalarsize> for TI and OI modes.
7701 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7703 Update copyright years.
7705 2018-01-03 Martin Liska <mliska@suse.cz>
7708 * ipa-visibility.c (function_and_variable_visibility): Skip
7709 functions with noipa attribure.
7711 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7713 * gcc.c (process_command): Update copyright notice dates.
7714 * gcov-dump.c (print_version): Ditto.
7715 * gcov.c (print_version): Ditto.
7716 * gcov-tool.c (print_version): Ditto.
7717 * gengtype.c (create_file): Ditto.
7718 * doc/cpp.texi: Bump @copying's copyright year.
7719 * doc/cppinternals.texi: Ditto.
7720 * doc/gcc.texi: Ditto.
7721 * doc/gccint.texi: Ditto.
7722 * doc/gcov.texi: Ditto.
7723 * doc/install.texi: Ditto.
7724 * doc/invoke.texi: Ditto.
7726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7728 * vector-builder.h (vector_builder::m_full_nelts): Change from
7729 unsigned int to poly_uint64.
7730 (vector_builder::full_nelts): Update prototype accordingly.
7731 (vector_builder::new_vector): Likewise.
7732 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7733 (vector_builder::operator ==): Likewise.
7734 (vector_builder::finalize): Likewise.
7735 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7736 Take the number of elements as a poly_uint64 rather than an
7738 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7739 from unsigned int to poly_uint64.
7740 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7741 (vec_perm_indices::new_vector): Likewise.
7742 (vec_perm_indices::length): Likewise.
7743 (vec_perm_indices::nelts_per_input): Likewise.
7744 (vec_perm_indices::input_nelts): Likewise.
7745 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7746 number of elements per input as a poly_uint64 rather than an
7747 unsigned int. Use the original encoding for variable-length
7748 vectors, rather than clamping each individual element.
7749 For the second and subsequent elements in each pattern,
7750 clamp the step and base before clamping their sum.
7751 (vec_perm_indices::series_p): Handle polynomial element counts.
7752 (vec_perm_indices::all_in_range_p): Likewise.
7753 (vec_perm_indices_to_tree): Likewise.
7754 (vec_perm_indices_to_rtx): Likewise.
7755 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7756 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7757 (tree_vector_builder::new_binary_operation): Handle polynomial
7758 element counts. Return false if we need to know the number
7759 of elements at compile time.
7760 * fold-const.c (fold_vec_perm): Punt if the number of elements
7761 isn't known at compile time.
7763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7765 * vec-perm-indices.h (vec_perm_builder): Change element type
7766 from HOST_WIDE_INT to poly_int64.
7767 (vec_perm_indices::element_type): Update accordingly.
7768 (vec_perm_indices::clamp): Handle polynomial element_types.
7769 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7770 (vec_perm_indices::all_in_range_p): Likewise.
7771 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7773 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7774 polynomial vec_perm_indices element types.
7775 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7776 * fold-const.c (fold_vec_perm): Likewise.
7777 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7778 * tree-vect-generic.c (lower_vec_perm): Likewise.
7779 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7780 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7781 element type to HOST_WIDE_INT.
7783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7784 Alan Hayward <alan.hayward@arm.com>
7785 David Sherwood <david.sherwood@arm.com>
7787 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7788 rather than an int. Use plus_constant.
7789 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7790 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7793 Alan Hayward <alan.hayward@arm.com>
7794 David Sherwood <david.sherwood@arm.com>
7796 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7797 a HOST_WIDE_INT to a poly_int64.
7799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7800 Alan Hayward <alan.hayward@arm.com>
7801 David Sherwood <david.sherwood@arm.com>
7803 * calls.c (load_register_parameters): Cope with polynomial
7804 mode sizes. Require a constant size for BLKmode parameters
7805 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7806 forces a parameter to be padded at the lsb end in order to
7807 fill a complete number of words, require the parameter size
7808 to be ordered wrt UNITS_PER_WORD.
7810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7811 Alan Hayward <alan.hayward@arm.com>
7812 David Sherwood <david.sherwood@arm.com>
7814 * reload1.c (spill_stack_slot_width): Change element type
7815 from unsigned int to poly_uint64_pod.
7816 (alter_reg): Treat mode sizes as polynomial.
7818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7819 Alan Hayward <alan.hayward@arm.com>
7820 David Sherwood <david.sherwood@arm.com>
7822 * reload.c (complex_word_subreg_p): New function.
7823 (reload_inner_reg_of_subreg, push_reload): Use it.
7825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7826 Alan Hayward <alan.hayward@arm.com>
7827 David Sherwood <david.sherwood@arm.com>
7829 * lra-constraints.c (process_alt_operands): Reject matched
7830 operands whose sizes aren't ordered.
7831 (match_reload): Refer to this check here.
7833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7834 Alan Hayward <alan.hayward@arm.com>
7835 David Sherwood <david.sherwood@arm.com>
7837 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7838 that the mode size is in the set {1, 2, 4, 8, 16}.
7840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7841 Alan Hayward <alan.hayward@arm.com>
7842 David Sherwood <david.sherwood@arm.com>
7844 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7845 Use plus_constant instead of gen_rtx_PLUS.
7847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7848 Alan Hayward <alan.hayward@arm.com>
7849 David Sherwood <david.sherwood@arm.com>
7851 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7852 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7853 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7854 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7855 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7856 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7857 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7858 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7859 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7860 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7862 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7863 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7864 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7865 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7866 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7867 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7868 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7869 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7870 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7871 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7873 * expr.c (emit_move_resolve_push): Treat the input and result
7874 of PUSH_ROUNDING as a poly_int64.
7875 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7876 (emit_push_insn): Likewise.
7877 * lra-eliminations.c (mark_not_eliminable): Likewise.
7878 * recog.c (push_operand): Likewise.
7879 * reload1.c (elimination_effects): Likewise.
7880 * rtlanal.c (nonzero_bits1): Likewise.
7881 * calls.c (store_one_arg): Likewise. Require the padding to be
7882 known at compile time.
7884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7885 Alan Hayward <alan.hayward@arm.com>
7886 David Sherwood <david.sherwood@arm.com>
7888 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7889 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7899 Alan Hayward <alan.hayward@arm.com>
7900 David Sherwood <david.sherwood@arm.com>
7902 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7903 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7904 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
7911 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7912 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7913 as a poly_uint64 too.
7915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7916 Alan Hayward <alan.hayward@arm.com>
7917 David Sherwood <david.sherwood@arm.com>
7919 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
7926 REGMODE_NATURAL_SIZE.
7927 * expmed.c (store_bit_field_1): Likewise.
7928 * expr.c (store_constructor): Likewise.
7929 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7930 and polynomial REGMODE_NATURAL_SIZE.
7931 (gen_lowpart_common): Likewise.
7932 * reginfo.c (record_subregs_of_mode): Likewise.
7933 * rtlanal.c (read_modify_subreg_p): Likewise.
7935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7936 Alan Hayward <alan.hayward@arm.com>
7937 David Sherwood <david.sherwood@arm.com>
7939 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7940 numbers of elements.
7942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7943 Alan Hayward <alan.hayward@arm.com>
7944 David Sherwood <david.sherwood@arm.com>
7946 * match.pd: Cope with polynomial numbers of vector elements.
7948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7949 Alan Hayward <alan.hayward@arm.com>
7950 David Sherwood <david.sherwood@arm.com>
7952 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7953 in a POINTER_PLUS_EXPR.
7955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7956 Alan Hayward <alan.hayward@arm.com>
7957 David Sherwood <david.sherwood@arm.com>
7959 * omp-simd-clone.c (simd_clone_subparts): New function.
7960 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7961 (ipa_simd_modify_function_body): Likewise.
7963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7964 Alan Hayward <alan.hayward@arm.com>
7965 David Sherwood <david.sherwood@arm.com>
7967 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7968 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7969 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7970 (expand_vector_condition, vector_element): Likewise.
7971 (subparts_gt): New function.
7972 (get_compute_type): Use subparts_gt.
7973 (count_type_subparts): Delete.
7974 (expand_vector_operations_1): Use subparts_gt instead of
7975 count_type_subparts.
7977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7978 Alan Hayward <alan.hayward@arm.com>
7979 David Sherwood <david.sherwood@arm.com>
7981 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7982 (vect_compile_time_alias): ...this new function. Do the calculation
7983 on poly_ints rather than trees.
7984 (vect_prune_runtime_alias_test_list): Update call accordingly.
7986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7987 Alan Hayward <alan.hayward@arm.com>
7988 David Sherwood <david.sherwood@arm.com>
7990 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7992 (vect_schedule_slp_instance): Likewise.
7994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7995 Alan Hayward <alan.hayward@arm.com>
7996 David Sherwood <david.sherwood@arm.com>
7998 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7999 constant and extern definitions for variable-length vectors.
8000 (vect_get_constant_vectors): Note that the number of units
8001 is known to be constant.
8003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8004 Alan Hayward <alan.hayward@arm.com>
8005 David Sherwood <david.sherwood@arm.com>
8007 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8008 of units as polynomial. Choose between WIDE and NARROW based
8011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8012 Alan Hayward <alan.hayward@arm.com>
8013 David Sherwood <david.sherwood@arm.com>
8015 * tree-vect-stmts.c (simd_clone_subparts): New function.
8016 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8019 Alan Hayward <alan.hayward@arm.com>
8020 David Sherwood <david.sherwood@arm.com>
8022 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8023 vectors as polynomial. Use build_index_vector for
8026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8027 Alan Hayward <alan.hayward@arm.com>
8028 David Sherwood <david.sherwood@arm.com>
8030 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8031 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8032 for variable-length vectors.
8033 (vectorizable_mask_load_store): Treat the number of units as
8034 polynomial, asserting that it is constant if the condition has
8035 already been enforced.
8036 (vectorizable_store, vectorizable_load): Likewise.
8038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8039 Alan Hayward <alan.hayward@arm.com>
8040 David Sherwood <david.sherwood@arm.com>
8042 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8043 of units as polynomial. Punt if we can't tell at compile time
8044 which vector contains the final result.
8046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8047 Alan Hayward <alan.hayward@arm.com>
8048 David Sherwood <david.sherwood@arm.com>
8050 * tree-vect-loop.c (vectorizable_induction): Treat the number
8051 of units as polynomial. Punt on SLP inductions. Use an integer
8052 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8053 cast of such a series for variable-length floating-point
8056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8057 Alan Hayward <alan.hayward@arm.com>
8058 David Sherwood <david.sherwood@arm.com>
8060 * tree.h (build_index_vector): Declare.
8061 * tree.c (build_index_vector): New function.
8062 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8063 of units as polynomial, forcibly converting it to a constant if
8064 vectorizable_reduction has already enforced the condition.
8065 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8066 to create a {1,2,3,...} vector.
8067 (vectorizable_reduction): Treat the number of units as polynomial.
8068 Choose vectype_in based on the largest scalar element size rather
8069 than the smallest number of units. Enforce the restrictions
8072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8073 Alan Hayward <alan.hayward@arm.com>
8074 David Sherwood <david.sherwood@arm.com>
8076 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8077 number of units as polynomial.
8079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8080 Alan Hayward <alan.hayward@arm.com>
8081 David Sherwood <david.sherwood@arm.com>
8083 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8084 * target.def (autovectorize_vector_sizes): Return the vector sizes
8085 by pointer, using vector_sizes rather than a bitmask.
8086 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8087 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8088 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8090 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8091 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8092 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8093 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8094 * omp-general.c (omp_max_vf): Likewise.
8095 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8096 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8097 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8098 * tree-vect-slp.c (vect_slp_bb): Likewise.
8099 * doc/tm.texi: Regenerate.
8100 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8102 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8103 the vector size as a poly_uint64 rather than an unsigned int.
8104 (current_vector_size): Change from an unsigned int to a poly_uint64.
8105 (get_vectype_for_scalar_type): Update accordingly.
8106 * tree.h (build_truth_vector_type): Take the size and number of
8107 units as a poly_uint64 rather than an unsigned int.
8108 (build_vector_type): Add a temporary overload that takes
8109 the number of units as a poly_uint64 rather than an unsigned int.
8110 * tree.c (make_vector_type): Likewise.
8111 (build_truth_vector_type): Take the number of units as a poly_uint64
8112 rather than an unsigned int.
8114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8115 Alan Hayward <alan.hayward@arm.com>
8116 David Sherwood <david.sherwood@arm.com>
8118 * target.def (get_mask_mode): Take the number of units and length
8119 as poly_uint64s rather than unsigned ints.
8120 * targhooks.h (default_get_mask_mode): Update accordingly.
8121 * targhooks.c (default_get_mask_mode): Likewise.
8122 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8123 * doc/tm.texi: Regenerate.
8125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8126 Alan Hayward <alan.hayward@arm.com>
8127 David Sherwood <david.sherwood@arm.com>
8129 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8130 * omp-general.c (omp_max_vf): Likewise.
8131 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8132 (expand_omp_simd): Handle polynomial safelen.
8133 * omp-low.c (omplow_simd_context): Add a default constructor.
8134 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8135 (lower_rec_simd_input_clauses): Update accordingly.
8136 (lower_rec_input_clauses): Likewise.
8138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8139 Alan Hayward <alan.hayward@arm.com>
8140 David Sherwood <david.sherwood@arm.com>
8142 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8143 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8144 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8145 (vect_analyze_slp_cost): Likewise.
8146 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8147 (vect_model_load_cost): Likewise.
8149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8150 Alan Hayward <alan.hayward@arm.com>
8151 David Sherwood <david.sherwood@arm.com>
8153 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8154 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8155 from an unsigned int * to a poly_uint64_pod *.
8156 (calculate_unrolling_factor): New function.
8157 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8160 Alan Hayward <alan.hayward@arm.com>
8161 David Sherwood <david.sherwood@arm.com>
8163 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8164 from an unsigned int to a poly_uint64.
8165 (_loop_vec_info::slp_unrolling_factor): Likewise.
8166 (_loop_vec_info::vectorization_factor): Change from an int
8168 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8169 (vect_get_num_vectors): New function.
8170 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8171 (vect_get_num_copies): Use vect_get_num_vectors.
8172 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8173 to an unsigned int *.
8174 (vect_analyze_data_refs): Change min_vf from an int * to a
8176 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8177 than an unsigned HOST_WIDE_INT.
8178 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8179 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8180 to an unsigned int *.
8181 (vect_analyze_data_ref_dependences): Likewise.
8182 (vect_compute_data_ref_alignment): Handle polynomial vf.
8183 (vect_enhance_data_refs_alignment): Likewise.
8184 (vect_prune_runtime_alias_test_list): Likewise.
8185 (vect_shift_permute_load_chain): Likewise.
8186 (vect_supportable_dr_alignment): Likewise.
8187 (dependence_distance_ge_vf): Take the vectorization factor as a
8188 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8189 (vect_analyze_data_refs): Change min_vf from an int * to a
8191 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8192 vfm1 as a poly_uint64 rather than an int. Make the same change
8193 for the returned bound_scalar.
8194 (vect_gen_vector_loop_niters): Handle polynomial vf.
8195 (vect_do_peeling): Likewise. Update call to
8196 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8197 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8199 * tree-vect-loop.c (vect_determine_vectorization_factor)
8200 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8201 (vect_get_known_peeling_cost): Likewise.
8202 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8203 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8204 (vect_transform_loop): Likewise. Use the lowest possible VF when
8205 updating the upper bounds of the loop.
8206 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8208 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8209 polynomial unroll factors.
8210 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8211 (vect_make_slp_decision): Likewise.
8212 (vect_supported_load_permutation_p): Likewise, and polynomial
8214 (vect_analyze_slp_cost): Handle polynomial vf.
8215 (vect_slp_analyze_node_operations): Likewise.
8216 (vect_slp_analyze_bb_1): Likewise.
8217 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8218 than an unsigned HOST_WIDE_INT.
8219 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8220 (vectorizable_load): Handle polynomial vf.
8221 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8223 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8226 Alan Hayward <alan.hayward@arm.com>
8227 David Sherwood <david.sherwood@arm.com>
8229 * match.pd: Handle bit operations involving three constants
8230 and try to fold one pair.
8232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8234 * tree-vect-loop-manip.c: Include gimple-fold.h.
8235 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8236 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8237 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8238 Add a path that uses a step of VF instead of 1, but disable it
8240 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8241 and niters_no_overflow parameters. Update calls to
8242 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8243 Create a new SSA name if the latter choses to use a ste other
8244 than zero, and return it via niters_vector_mult_vf_var.
8245 * tree-vect-loop.c (vect_transform_loop): Update calls to
8246 vect_do_peeling, vect_gen_vector_loop_niters and
8247 slpeel_make_loop_iterate_ntimes.
8248 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8249 (vect_gen_vector_loop_niters): Update declarations after above changes.
8251 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8253 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8254 128-bit round to integer instructions.
8255 (ceil<mode>2): Likewise.
8256 (btrunc<mode>2): Likewise.
8257 (round<mode>2): Likewise.
8259 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8261 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8262 unaligned VSX load/store on P8/P9.
8263 (expand_block_clear): Allow the use of unaligned VSX
8264 load/store on P8/P9.
8266 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8268 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8270 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8271 swap associated with both a load and a store.
8273 2018-01-02 Andrew Waterman <andrew@sifive.com>
8275 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8276 * config/riscv/riscv.md (clear_cache): Use it.
8278 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8280 * web.c: Remove out-of-date comment.
8282 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8284 * expr.c (fixup_args_size_notes): Check that any existing
8285 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8286 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8287 (emit_single_push_insn): ...here.
8289 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8291 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8292 (const_vector_encoded_nelts): New function.
8293 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8294 (const_vector_int_elt, const_vector_elt): Declare.
8295 * emit-rtl.c (const_vector_int_elt_1): New function.
8296 (const_vector_elt): Likewise.
8297 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8298 of CONST_VECTOR_ELT.
8300 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8302 * expr.c: Include rtx-vector-builder.h.
8303 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8304 directly on the tree encoding.
8305 (const_vector_from_tree): Likewise.
8306 * optabs.c: Include rtx-vector-builder.h.
8307 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8308 sequence of "u" values.
8309 * vec-perm-indices.c: Include rtx-vector-builder.h.
8310 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8311 directly on the vec_perm_indices encoding.
8313 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8315 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8316 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8317 * rtx-vector-builder.h: New file.
8318 * rtx-vector-builder.c: Likewise.
8319 * rtl.h (rtx_def::u2): Add a const_vector field.
8320 (CONST_VECTOR_NPATTERNS): New macro.
8321 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8322 (CONST_VECTOR_DUPLICATE_P): Likewise.
8323 (CONST_VECTOR_STEPPED_P): Likewise.
8324 (CONST_VECTOR_ENCODED_ELT): Likewise.
8325 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8326 (unwrap_const_vec_duplicate): Likewise.
8327 (const_vec_series_p): Check for a non-duplicated vector encoding.
8328 Say that the function only returns true for integer vectors.
8329 * emit-rtl.c: Include rtx-vector-builder.h.
8330 (gen_const_vec_duplicate_1): Delete.
8331 (gen_const_vector): Call gen_const_vec_duplicate instead of
8332 gen_const_vec_duplicate_1.
8333 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8334 (gen_const_vec_duplicate): Use rtx_vector_builder.
8335 (gen_const_vec_series): Likewise.
8336 (gen_rtx_CONST_VECTOR): Likewise.
8337 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8338 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8339 Build a new vector rather than modifying a CONST_VECTOR in-place.
8340 (handle_special_swappables): Update call accordingly.
8341 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8342 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8343 Build a new vector rather than modifying a CONST_VECTOR in-place.
8344 (handle_special_swappables): Update call accordingly.
8346 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8348 * simplify-rtx.c (simplify_const_binary_operation): Use
8349 CONST_VECTOR_ELT instead of XVECEXP.
8351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8353 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8354 the selector elements to be different from the data elements
8355 if the selector is a VECTOR_CST.
8356 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8357 ssizetype for the selector.
8359 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8361 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8362 before testing each element individually.
8363 * tree-vect-generic.c (lower_vec_perm): Likewise.
8365 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8367 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8368 * selftest-run-tests.c (selftest::run_tests): Call it.
8369 * vector-builder.h (vector_builder::operator ==): New function.
8370 (vector_builder::operator !=): Likewise.
8371 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8372 (vec_perm_indices::all_from_input_p): New function.
8373 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8374 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8375 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8376 instead of reading the VECTOR_CST directly. Detect whether both
8377 vector inputs are the same before constructing the vec_perm_indices,
8378 and update the number of inputs argument accordingly. Use the
8379 utility functions added above. Only construct sel2 if we need to.
8381 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8383 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8384 the broadcast of the low byte.
8385 (expand_mult_highpart): Use an explicit encoding for the permutes.
8386 * optabs-query.c (can_mult_highpart_p): Likewise.
8387 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8388 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8389 (vectorizable_bswap): Likewise.
8390 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8391 explicit encoding for the power-of-2 permutes.
8392 (vect_permute_store_chain): Likewise.
8393 (vect_grouped_load_supported): Likewise.
8394 (vect_permute_load_chain): Likewise.
8396 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8398 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8399 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8400 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8401 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8402 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8403 (vect_gen_perm_mask_any): Likewise.
8405 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8407 * int-vector-builder.h: New file.
8408 * vec-perm-indices.h: Include int-vector-builder.h.
8409 (vec_perm_indices): Redefine as an int_vector_builder.
8410 (auto_vec_perm_indices): Delete.
8411 (vec_perm_builder): Redefine as a stand-alone class.
8412 (vec_perm_indices::vec_perm_indices): New function.
8413 (vec_perm_indices::clamp): Likewise.
8414 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8415 (vec_perm_indices::new_vector): New function.
8416 (vec_perm_indices::new_expanded_vector): Update for new
8417 vec_perm_indices class.
8418 (vec_perm_indices::rotate_inputs): New function.
8419 (vec_perm_indices::all_in_range_p): Operate directly on the
8420 encoded form, without computing elided elements.
8421 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8422 encoding. Update for new vec_perm_indices class.
8423 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8424 the given vec_perm_builder.
8425 (expand_vec_perm_var): Update vec_perm_builder constructor.
8426 (expand_mult_highpart): Use vec_perm_builder instead of
8427 auto_vec_perm_indices.
8428 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8429 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8430 or double series encoding as appropriate.
8431 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8432 vec_perm_indices instead of auto_vec_perm_indices.
8433 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8434 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8435 (vect_permute_store_chain): Likewise.
8436 (vect_grouped_load_supported): Likewise.
8437 (vect_permute_load_chain): Likewise.
8438 (vect_shift_permute_load_chain): Likewise.
8439 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8440 (vect_transform_slp_perm_load): Likewise.
8441 (vect_schedule_slp_instance): Likewise.
8442 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8443 (vectorizable_mask_load_store): Likewise.
8444 (vectorizable_bswap): Likewise.
8445 (vectorizable_store): Likewise.
8446 (vectorizable_load): Likewise.
8447 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8448 vec_perm_indices instead of auto_vec_perm_indices. Use
8449 tree_to_vec_perm_builder to read the vector from a tree.
8450 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8451 vec_perm_builder instead of a vec_perm_indices.
8452 (have_whole_vector_shift): Use vec_perm_builder and
8453 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8454 truncation to calc_vec_perm_mask_for_shift.
8455 (vect_create_epilog_for_reduction): Likewise.
8456 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8457 from auto_vec_perm_indices to vec_perm_indices.
8458 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8459 instead of changing individual elements.
8460 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8461 the vector in d.perm.
8462 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8463 from auto_vec_perm_indices to vec_perm_indices.
8464 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8465 instead of changing individual elements.
8466 (arm_vectorize_vec_perm_const): Use new_vector to install
8467 the vector in d.perm.
8468 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8469 Update vec_perm_builder constructor.
8470 (rs6000_expand_interleave): Likewise.
8471 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8472 (rs6000_expand_interleave): Likewise.
8474 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8476 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8477 to qimode could truncate the indices.
8478 * optabs.c (expand_vec_perm_var): Likewise.
8480 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8482 * Makefile.in (OBJS): Add vec-perm-indices.o.
8483 * vec-perm-indices.h: New file.
8484 * vec-perm-indices.c: Likewise.
8485 * target.h (vec_perm_indices): Replace with a forward class
8487 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8488 * optabs.h: Include vec-perm-indices.h.
8489 (expand_vec_perm): Delete.
8490 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8491 (expand_vec_perm_const): Declare.
8492 * target.def (vec_perm_const_ok): Replace with...
8493 (vec_perm_const): ...this new hook.
8494 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8495 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8496 * doc/tm.texi: Regenerate.
8497 * optabs.def (vec_perm_const): Delete.
8498 * doc/md.texi (vec_perm_const): Likewise.
8499 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8500 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8501 expand_vec_perm for constant permutation vectors. Assert that
8502 the mode of variable permutation vectors is the integer equivalent
8503 of the mode that is being permuted.
8504 * optabs-query.h (selector_fits_mode_p): Declare.
8505 * optabs-query.c: Include vec-perm-indices.h.
8506 (selector_fits_mode_p): New function.
8507 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8508 is defined, instead of checking whether the vec_perm_const_optab
8509 exists. Use targetm.vectorize.vec_perm_const instead of
8510 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8511 fit in the vector mode before using a variable permute.
8512 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8513 vec_perm_indices instead of an rtx.
8514 (expand_vec_perm): Replace with...
8515 (expand_vec_perm_const): ...this new function. Take the selector
8516 as a vec_perm_indices rather than an rtx. Also take the mode of
8517 the selector. Update call to shift_amt_for_vec_perm_mask.
8518 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8519 Use vec_perm_indices::new_expanded_vector to expand the original
8520 selector into bytes. Check whether the indices fit in the vector
8521 mode before using a variable permute.
8522 (expand_vec_perm_var): Make global.
8523 (expand_mult_highpart): Use expand_vec_perm_const.
8524 * fold-const.c: Includes vec-perm-indices.h.
8525 * tree-ssa-forwprop.c: Likewise.
8526 * tree-vect-data-refs.c: Likewise.
8527 * tree-vect-generic.c: Likewise.
8528 * tree-vect-loop.c: Likewise.
8529 * tree-vect-slp.c: Likewise.
8530 * tree-vect-stmts.c: Likewise.
8531 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8533 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8534 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8535 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8536 (aarch64_vectorize_vec_perm_const): ...this new function.
8537 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8538 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8539 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8540 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8541 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8542 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8543 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8545 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8546 check for NEON modes.
8547 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8548 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8549 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8550 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8552 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8553 the old VEC_PERM_CONST conditions.
8554 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8555 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8556 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8557 (ia64_vectorize_vec_perm_const_ok): Merge into...
8558 (ia64_vectorize_vec_perm_const): ...this new function.
8559 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8560 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8561 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8562 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8563 * config/mips/mips.c (mips_expand_vec_perm_const)
8564 (mips_vectorize_vec_perm_const_ok): Merge into...
8565 (mips_vectorize_vec_perm_const): ...this new function.
8566 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8567 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8568 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8569 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8570 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8571 (rs6000_expand_vec_perm_const): Delete.
8572 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8574 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8575 (altivec_expand_vec_perm_const_le): Take each operand individually.
8576 Operate on constant selectors rather than rtxes.
8577 (altivec_expand_vec_perm_const): Likewise. Update call to
8578 altivec_expand_vec_perm_const_le.
8579 (rs6000_expand_vec_perm_const): Delete.
8580 (rs6000_vectorize_vec_perm_const_ok): Delete.
8581 (rs6000_vectorize_vec_perm_const): New function.
8582 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8583 an element count and rtx array.
8584 (rs6000_expand_extract_even): Update call accordingly.
8585 (rs6000_expand_interleave): Likewise.
8586 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8587 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8588 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8589 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8590 (rs6000_expand_vec_perm_const): Delete.
8591 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8592 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8593 (altivec_expand_vec_perm_const_le): Take each operand individually.
8594 Operate on constant selectors rather than rtxes.
8595 (altivec_expand_vec_perm_const): Likewise. Update call to
8596 altivec_expand_vec_perm_const_le.
8597 (rs6000_expand_vec_perm_const): Delete.
8598 (rs6000_vectorize_vec_perm_const_ok): Delete.
8599 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8600 reference to the SPE evmerge intructions.
8601 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8602 an element count and rtx array.
8603 (rs6000_expand_extract_even): Update call accordingly.
8604 (rs6000_expand_interleave): Likewise.
8605 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8606 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8608 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8610 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8612 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8613 vector mode and that that mode matches the mode of the data
8615 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8616 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8617 directly using expand_vec_perm_1 when forcing selectors into
8619 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8621 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8623 * optabs-query.h (can_vec_perm_p): Delete.
8624 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8625 * optabs-query.c (can_vec_perm_p): Split into...
8626 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8627 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8628 particular selector is valid.
8629 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8630 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8631 (vect_grouped_load_supported): Likewise.
8632 (vect_shift_permute_load_chain): Likewise.
8633 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8634 (vect_transform_slp_perm_load): Likewise.
8635 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8636 (vectorizable_bswap): Likewise.
8637 (vect_gen_perm_mask_checked): Likewise.
8638 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8639 implementations of variable permutation vectors into account
8640 when deciding which selector to use.
8641 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8642 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8643 with a false third argument.
8644 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8645 to test whether the constant selector is valid and can_vec_perm_var_p
8646 to test whether a variable selector is valid.
8648 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8650 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8651 * optabs-query.c (can_vec_perm_p): Likewise.
8652 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8653 instead of vec_perm_indices.
8654 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8655 (vect_gen_perm_mask_checked): Likewise,
8656 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8657 (vect_gen_perm_mask_checked): Likewise,
8659 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8661 * optabs-query.h (qimode_for_vec_perm): Declare.
8662 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8663 (qimode_for_vec_perm): ...this new function.
8664 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8666 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8668 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8669 does not have a conditional at the top.
8671 2018-01-02 Richard Biener <rguenther@suse.de>
8673 * ipa-inline.c (big_speedup_p): Fix expression.
8675 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8678 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8681 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8685 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8686 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8687 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8688 cond_taken_branch_cost 3->4.
8690 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8692 PR tree-optimization/83581
8693 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8694 TODO_cleanup_cfg if any changes have been made.
8697 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8698 convert_modes if target mode has the right side, but different mode
8702 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8703 last argument when extracting from CONCAT. If either from_real or
8704 from_imag is NULL, use expansion through memory. If result is not
8705 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8706 the parts directly to inner mode, if even that fails, use expansion
8710 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8711 check for bswap in mode rather than HImode and use that in expand_unop
8714 Copyright (C) 2018 Free Software Foundation, Inc.
8716 Copying and distribution of this file, with or without modification,
8717 are permitted in any medium without royalty provided the copyright
8718 notice and this notice are preserved.