| blob | 0937241a0584bf6691f96701dc58d13e559fa7a1 |
1 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3 PR lto/84805
4 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5 incomplete types.
7 2018-03-13 Martin Liska <mliska@suse.cz>
9 PR ipa/84658.
10 * (sem_item_optimizer::sem_item_optimizer): Initialize new
11 vector.
12 (sem_item_optimizer::~sem_item_optimizer): Release it.
13 (sem_item_optimizer::merge_classes): Register variable aliases.
14 (sem_item_optimizer::fixup_pt_set): New function.
15 (sem_item_optimizer::fixup_points_to_sets): Likewise.
16 * ipa-icf.h: Declare new variables and functions.
18 2018-03-13 Jakub Jelinek <jakub@redhat.com>
20 PR middle-end/84834
21 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
22 integer_pow2p@2 and test integer_pow2p in condition.
23 (A < 0 ? C : 0): Similarly for @1.
25 PR middle-end/84831
26 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
27 characters starting at p contain '\0' character, don't look beyond
28 that.
30 PR target/84827
31 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
32 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
34 PR target/84828
35 * reg-stack.c (change_stack): Change update_end var from int to
36 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
37 also call set_block_for_insn on the newly added insns and rescan.
39 PR target/84786
40 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
41 on the last operand.
43 PR c++/84704
44 * tree.c (stabilize_reference_1): Return save_expr (e) for
45 STATEMENT_LIST even if it doesn't have side-effects.
47 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
49 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
51 2018-03-12 Renlin Li <renlin.li@arm.com>
53 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
54 aarch64_output_scalar_simd_mov_immediate.
56 2018-03-12 Martin Sebor <msebor@redhat.com>
58 PR tree-optimization/83456
59 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
60 for perfectly overlapping calls to memcpy.
61 (gimple_fold_builtin_memory_chk): Same.
62 (gimple_fold_builtin_strcpy): Handle no-warning.
63 (gimple_fold_builtin_stxcpy_chk): Same.
64 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
66 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
68 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
69 parameter. Use it for SFmode.
70 (rs6000_function_arg_advance_1): Adjust.
71 (rs6000_function_arg): Adjust.
72 (rs6000_gimplify_va_arg): Pass false for that new parameter.
74 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
76 PR rtl-optimization/84169
77 PR rtl-optimization/84780
78 * combine.c (can_combine_p): Check for a 2-insn combination whether
79 the destination register is used between the two insns, too.
81 2018-03-12 Richard Biener <rguenther@suse.de>
83 PR tree-optimization/84803
84 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
85 for refs DR analysis didn't process.
87 2018-03-12 Richard Biener <rguenther@suse.de>
89 PR tree-optimization/84777
90 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
91 force-vectorize loops ignore whether we are optimizing for size.
93 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
95 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
96 (TARGET_MD_ASM_ADJUST): Define.
98 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
99 Kito Cheng <kito.cheng@gmail.com>
100 Chung-Ju Wu <jasonwucj@gmail.com>
102 * config/nds32/nds32.c (nds32_compute_stack_frame,
103 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
104 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
105 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
106 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
107 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
108 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
109 * config/nds32/nds32.md (prologue, epilogue): Use macro
110 NDS32_V3PUSH_AVAILABLE_P to do checking.
112 2018-03-11 Jakub Jelinek <jakub@redhat.com>
114 PR debug/58150
115 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
116 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
117 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
118 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
119 addition of most attributes on !orig_type_die or the attribute not
120 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
122 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
123 Chung-Ju Wu <jasonwucj@gmail.com>
125 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
126 __NDS32_VH__ macro.
127 * config/nds32/nds32.opt (mvh): New option.
129 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
130 Chung-Ju Wu <jasonwucj@gmail.com>
132 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
133 function.
134 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
135 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
136 definition.
138 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
139 Chung-Ju Wu <jasonwucj@gmail.com>
141 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
142 function.
143 * config/nds32/nds32-multiple.md (strlensi): New pattern.
144 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
146 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
147 Kito Cheng <kito.cheng@gmail.com>
148 Chung-Ju Wu <jasonwucj@gmail.com>
150 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
151 UNSPEC_FFMISM and UNSPEC_FLMISM.
152 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
153 for ffb, ffmism and flmism.
154 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
155 (unspec_ffmism): Ditto.
156 (unspec_flmism): Ditto.
157 (nds32_expand_builtin_impl): Check if string extension is available.
158 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
159 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
161 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
163 Reverting patch:
164 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
166 PR target/83712
167 * lra-assigns.c (assign_by_spills): Return a flag of reload
168 assignment failure. Do not process the reload assignment
169 failures. Do not spill other reload pseudos if they has the same
170 reg class.
171 (lra_assign): Add a return arg. Set up from the result of
172 assign_by_spills call.
173 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
174 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
175 usage_insns if it is not NULL.
176 (spill_hard_reg_in_range): New function.
177 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
178 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
179 function prototypes.
180 (lra_assign): Change prototype.
181 * lra.c (lra): Add code to deal with fails by splitting hard reg
182 live ranges.
184 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
186 PR target/84807
187 * config/i386/i386.opt: Replace Enforcment with Enforcement.
189 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
191 PR debug/84620
192 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
193 (dw_val_node): Add val_symbolic_view.
194 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
195 (symview_upper_bound): New.
196 (new_line_info_table): Initialize symviews_since_reset.
197 (dwarf2out_source_line): Count symviews_since_reset and set
198 symview_upper_bound.
199 (dw_val_equal_p): Handle symview.
200 (add_AT_symview): New.
201 (print_dw_val): Handle symview.
202 (attr_checksum, attr_checksum_ordered): Likewise.
203 (same_dw_val_p, size_of_die): Likewise.
204 (value_format, output_die): Likewise.
205 (add_high_low_attributes): Use add_AT_symview for entry_view.
206 (dwarf2out_finish): Reset symview_upper_bound, clear
207 zero_view_p.
209 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
211 PR target/83969
212 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
213 Add strict argument and use it.
214 (rs6000_split_multireg_move): Update for new strict argument.
215 (mem_operand_gpr): Disallow all non-offsettable addresses.
216 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
218 2018-03-09 Jakub Jelinek <jakub@redhat.com>
220 PR target/84772
221 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
222 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
223 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
225 PR c++/84767
226 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
227 decl, use remap_type if we want to use the type.
229 2018-03-09 Martin Sebor <msebor@redhat.com>
231 PR tree-optimization/84526
232 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
233 Remove dead code.
234 (builtin_access::generic_overlap): Be prepared to handle non-array
235 base objects.
237 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
239 PR rtl-optimization/84682
240 * lra-constraints.c (process_address_1): Check is_address flag
241 for address constraints.
242 (process_alt_operands): Likewise.
243 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
244 preprocess_constraints.
245 * recog.h (preprocess_constraints): Add oploc parameter.
246 Adjust callers.
247 * recog.c (preprocess_constraints): Test address_operand for
248 CT_ADDRESS constraints.
250 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
252 PR target/83712
253 * lra-assigns.c (assign_by_spills): Return a flag of reload
254 assignment failure. Do not process the reload assignment
255 failures. Do not spill other reload pseudos if they has the same
256 reg class.
257 (lra_assign): Add a return arg. Set up from the result of
258 assign_by_spills call.
259 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
260 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
261 usage_insns if it is not NULL.
262 (spill_hard_reg_in_range): New function.
263 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
264 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
265 function prototypes.
266 (lra_assign): Change prototype.
267 * lra.c (lra): Add code to deal with fails by splitting hard reg
268 live ranges.
270 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
272 PR target/83193
273 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
274 Accept complain bool parameter. Only emit errors if it is true.
275 (arm_parse_cpu_option_name): Likewise.
276 (arm_target_thumb_only): Adjust callers of the above.
277 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
278 prototype to take a default true bool parameter.
279 (arm_parse_arch_option_name): Likewise.
281 2018-03-09 David Malcolm <dmalcolm@redhat.com>
282 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
284 PR jit/64089
285 PR jit/84288
286 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
287 * configure: Regenerate.
288 * configure.ac ("linker --version-script option"): New.
289 ("linker soname option"): New.
291 2018-03-09 Richard Biener <rguenther@suse.de>
293 PR tree-optimization/84775
294 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
295 immediate uses of predicate stmts and mark them modified.
297 Revert
298 PR tree-optimization/84178
299 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
300 to caller.
301 (version_loop_for_if_conversion): Delay update_ssa call.
302 (tree_if_conversion): Delay update_ssa until after predicate
303 insertion.
305 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
307 PR target/84763
308 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
309 when the function accesses prior frames.
311 2018-03-08 Jakub Jelinek <jakub@redhat.com>
313 PR debug/84456
314 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
315 gen_llsym, otherwise call maybe_gen_llsym.
317 PR inline-asm/84742
318 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
319 has ',' character inside of it.
321 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
323 PR target/84748
324 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
325 as clobbering CC_REGNUM.
327 2018-03-08 Richard Biener <rguenther@suse.de>
329 PR middle-end/84552
330 * tree-scalar-evolution.c: Include tree-into-ssa.h.
331 (follow_copies_to_constant): Do not follow SSA names registered
332 for update.
334 2018-03-08 Richard Biener <rguenther@suse.de>
336 PR tree-optimization/84178
337 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
338 to caller.
339 (version_loop_for_if_conversion): Delay update_ssa call.
340 (tree_if_conversion): Delay update_ssa until after predicate
341 insertion.
343 2018-03-08 David Malcolm <dmalcolm@redhat.com>
345 PR tree-optimization/84178
346 * tree-if-conv.c (release_bb_predicate): Remove the
347 the assertion that the stmts have NULL use_ops.
348 Discard the statements, asserting that they haven't
349 yet been added to a BB.
351 2018-03-08 Richard Biener <rguenther@suse.de>
353 PR tree-optimization/84746
354 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
355 (phi_translate): Pass in destination ANTIC_OUT set.
356 (phi_translate_1): Likewise. For a simplified result lookup
357 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
358 (phi_translate_set): Adjust.
359 (do_pre_regular_insertion): Likewise.
360 (do_pre_partial_partial_insertion): Likewise.
362 2018-03-08 Martin Liska <mliska@suse.cz>
364 PR gcov-profile/84735
365 * doc/gcov.texi: Document usage of profile files.
366 * gcov-io.h: Document changes in the format.
368 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
370 PR debug/84404
371 PR debug/84408
372 * dwarf2out.c (struct dw_line_info_table): Update comments for
373 view == -1.
374 (FORCE_RESET_NEXT_VIEW): New.
375 (FORCE_RESETTING_VIEW_P): New.
376 (RESETTING_VIEW_P): Check for -1 too.
377 (ZERO_VIEW_P): Likewise.
378 (new_line_info_table): Force-reset next view.
379 (dwarf2out_begin_function): Likewise.
380 (dwarf2out_source_line): Simplify zero_view_p initialization.
381 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
382 view directly. Omit view when omitting .loc at line 0.
384 2018-03-08 Jakub Jelinek <jakub@redhat.com>
386 PR tree-optimization/84740
387 * tree-switch-conversion.c (process_switch): Call build_constructors
388 only if info.phi_count is non-zero.
390 PR tree-optimization/84739
391 * tree-tailcall.c (find_tail_calls): Check call arguments against
392 DECL_ARGUMENTS (current_function_decl) rather than
393 DECL_ARGUMENTS (func) when checking for tail recursion.
395 2018-03-07 Jakub Jelinek <jakub@redhat.com>
397 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
398 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
399 Volker Reichelt's entry and add entries for people that perform
400 GCC fuzzy testing and report numerous bugs.
402 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
404 PR target/82411
405 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
406 readonly data in sdata, if that is disabled.
407 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
408 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
409 -mreadonly-in-sdata option.
411 2018-03-07 Martin Sebor <msebor@redhat.com>
413 PR tree-optimization/84468
414 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
415 basic block when looking for nul assignment.
417 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
419 PR target/84277
420 * except.h (output_function_exception_table): Adjust prototype.
421 * except.c (output_function_exception_table): Remove FNNAME parameter
422 and add SECTION parameter. Ouput one part of the table at a time.
423 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
424 the first part of the exception table and emit unwind directives.
425 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
426 (i386_pe_seh_cold_init): Likewise.
427 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
428 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
429 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
430 (ix86_output_call_insn): Emit a nop in one more case for SEH.
431 * config/i386/winnt.c: Include except.h.
432 (struct seh_frame_state): Add reg_offset, after_prologue and
433 in_cold_section fields.
434 (i386_pe_seh_end_prologue): Set seh->after_prologue.
435 (i386_pe_seh_cold_init): New function.
436 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
437 to seh->in_cold_section.
438 (seh_emit_push): Record the offset of the push.
439 (seh_emit_save): Record the offet of the save.
440 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
441 Test seh->after_prologue to disregard the epilogue.
442 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
443 (i386_pe_end_cold_function): New function.
445 2018-03-07 Jakub Jelinek <jakub@redhat.com>
447 PR fortran/84565
448 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
449 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
451 PR c++/84704
452 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
453 on tmp_var.
454 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
455 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
457 PR middle-end/84723
458 * multiple_target.c: Include tree-inline.h and intl.h.
459 (expand_target_clones): Diagnose and fail if node->definition and
460 !tree_versionable_function_p (node->decl).
462 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
464 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
465 sprint_ul.
466 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
467 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
468 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
470 2018-03-06 Jakub Jelinek <jakub@redhat.com>
472 PR target/84710
473 * combine.c (try_combine): Use reg_or_subregno instead of handling
474 just paradoxical SUBREGs and REGs.
476 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
478 * config/arc/arc.c (arc_finalize_pic): Remove function.
479 (arc_must_save_register): We use single base PIC register, remove
480 checks to save/restore the PIC register.
481 (arc_expand_prologue): Likewise.
482 * config/arc/arc-protos.h (arc_set_default_type_attributes):
483 Remove.
484 (arc_verify_short): Likewise.
485 (arc_attr_type): Likewise.
486 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
487 (walk_stores): Likewise.
488 (arc_address_cost): Make it static.
489 (arc_verify_short): Likewise.
490 (branch_dest): Likewise.
491 (arc_attr_type): Likewise.
492 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
493 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
494 (arc_final_prescan_insn): Remove inserting the nops due to
495 hardware hazards. It is done in reorg step.
496 (insn_length_variant_t): Remove.
497 (insn_length_parameters_t): Likewise.
498 (arc_insn_length_parameters): Likewise.
499 (arc_get_insn_variants): Likewise.
500 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
502 2018-03-06 Jakub Jelinek <jakub@redhat.com>
504 PR inline-asm/84683
505 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
506 assertion failure.
508 PR tree-optimization/84687
509 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
510 on new_node->decl.
511 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
513 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
515 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
516 Rename to ppc_speculation_barrier.
517 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
518 __builtin_ppc_speculation_barrier.
520 2018-03-05 Jakub Jelinek <jakub@redhat.com>
522 PR target/84700
523 * combine.c (combine_simplify_rtx): Don't try to simplify if
524 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
525 are equal to x.
527 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
529 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
530 to 32 bytes when compiling for POWER9.
532 2018-03-05 Jakub Jelinek <jakub@redhat.com>
534 PR target/84564
535 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
536 regparm >= 3 with no arg reg available also for calls with
537 flag_force_indirect_call. Pass decl to ix86_function_regparm.
539 PR target/84524
540 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
541 orig,vex.
542 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
544 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
546 PR target/84264
547 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
549 2018-03-05 Richard Biener <rguenther@suse.de>
551 PR tree-optimization/84486
552 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
553 When inserting a __builtin_assume_aligned call set the LHS
554 SSA name alignment info accordingly.
556 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
558 PR tree-optimization/84114
559 * config/aarch64/aarch64.c (aarch64_reassociation_width)
560 Avoid reassociation of FLOAT_MODE addition.
562 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
564 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
565 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
566 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
567 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
568 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
569 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
570 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
571 and -mwbnoinvd.
572 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
573 __builtin_ia32_wbinvd): New builtins.
574 (SPECIAL_ARGS2): New.
575 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
576 (SPECIAL_ARGS2): New.
577 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
578 (ix86_valid_target_attribute_inner_p): Ditto.
579 (ix86_init_mmx_sse_builtins): Add special_args2.
580 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
581 TARGET_WBNOINVD_P): New.
582 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
583 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
584 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
585 * config/i386/immintrin.h (_wbinvd): New intrinsic.
586 * config/i386/pconfigintrin.h: New file.
587 * config/i386/wbnoinvdintrin.h: Ditto.
588 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
589 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
591 2018-03-05 Richard Biener <rguenther@suse.de>
593 PR tree-optimization/84670
594 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
595 member.
596 (BB_VISITED_WITH_VISITED_SUCCS): New define.
597 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
598 (compute_antic_aux): Only assert the number of values in ANTIC_IN
599 doesn't grow if all successors (recursively) were visited at least
600 once.
602 2018-03-05 Richard Biener <rguenther@suse.de>
604 PR tree-optimization/84650
605 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
606 if executed in the loop pipeline.
608 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
610 * doc/configfiles.texi (Configuration Files): Move info about
611 conditionalizing $target-protos.h to...
612 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
613 differs from $target-protos.h.
615 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
616 Chung-Ju Wu <jasonwucj@gmail.com>
618 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
619 * config/nds32/nds32-multiple.md (setmemsi): Define.
620 * config/nds32/nds32-memory-manipulation.c
621 (nds32_gen_dup_4_byte_to_word_value): New.
622 (emit_setmem_word_loop): New.
623 (emit_setmem_byte_loop): New.
624 (nds32_expand_setmem_loop): New.
625 (nds32_expand_setmem_loop_v3m): New.
626 (nds32_expand_setmem_unroll): New.
627 (nds32_expand_setmem): New.
629 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
630 Chung-Ju Wu <jasonwucj@gmail.com>
632 * config/nds32/nds32-memory-manipulation.c
633 (nds32_emit_load_store): New.
634 (nds32_emit_post_inc_load_store): New.
635 (nds32_emit_mem_move): New.
636 (nds32_emit_mem_move_block): New.
637 (nds32_expand_movmemsi_loop_unknown_size): New.
638 (nds32_expand_movmemsi_loop_known_size): New.
639 (nds32_expand_movmemsi_loop): New.
640 (nds32_expand_movmemsi_unroll): New.
641 (nds32_expand_movmemqi): Rename ...
642 (nds32_expand_movmemsi): ... to this.
643 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
644 (movmemsi): ... to this.
645 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
646 (nds32_expand_movmemsi): ... to this.
648 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
649 Monk Chiang <sh.chiang04@gmail.com>
650 Chung-Ju Wu <jasonwucj@gmail.com>
652 * config/nds32/nds32-protos.h
653 (nds32_expand_load_multiple): New arguments.
654 (nds32_expand_store_multiple): Ditto.
655 (nds32_valid_multiple_load_store): Rename ...
656 (nds32_valid_multiple_load_store_p): ... to this.
657 * config/nds32/nds32-memory-manipulation.c
658 (nds32_expand_load_multiple): Refine implementation.
659 (nds32_expand_store_multiple): Ditto.
660 * config/nds32/nds32-multiple.md
661 (load_multiple): Update nds32_expand_load_multiple interface.
662 (store_multiple): Update nds32_expand_store_multiple interface.
663 * config/nds32/nds32-predicates.c
664 (nds32_valid_multiple_load_store): Rename ...
665 (nds32_valid_multiple_load_store_p): ... to this and refine
666 implementation.
667 * config/nds32/predicates.md
668 (nds32_load_multiple_and_update_address_operation): New predicate.
669 (nds32_store_multiple_and_update_address_operation): New predicate.
671 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
672 Chung-Ju Wu <jasonwucj@gmail.com>
674 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
675 (combo): New attribute.
676 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
678 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
680 * config/nds32/nds32.opt: Change -mcmodel= default value.
682 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
683 Monk Chiang <sh.chiang04@gmail.com>
684 Chung-Ju Wu <jasonwucj@gmail.com>
686 * config/nds32/constants.md (unspec_element): New enum.
687 * config/nds32/constraints.md (Umw): New constraint.
688 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
689 * config/nds32/nds32-intrinsic.md: Likewise.
690 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
691 (nds32_valid_smw_lwm_base_p): New.
692 (nds32_output_smw_single_word): New.
693 (nds32_output_lmw_single_word): New.
694 (nds32_expand_unaligned_load): New.
695 (nds32_expand_unaligned_store): New.
696 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
697 (nds32_output_smw_single_word): Declare.
698 (nds32_output_lmw_single_word): Declare.
699 (nds32_expand_unaligned_load): Declare.
700 (nds32_expand_unaligned_store): Declare.
701 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
702 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
703 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
704 NDS32_BUILTIN_UASTORE_DW.
705 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
706 predicate.
708 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
709 Kito Cheng <kito.cheng@gmail.com>
710 Chung-Ju Wu <jasonwucj@gmail.com>
712 * config/nds32/nds32-intrinsic.c
713 (nds32_expand_builtin_null_ftype_reg): Delete.
714 (nds32_expand_builtin_reg_ftype_imm): Ditto.
715 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
716 (nds32_read_argument): New.
717 (nds32_legitimize_target): Ditto.
718 (nds32_legitimize_argument): Ditto.
719 (nds32_check_constant_argument): Ditto.
720 (nds32_expand_unop_builtin): Ditto.
721 (nds32_expand_unopimm_builtin): Ditto.
722 (nds32_expand_binop_builtin): Ditto.
723 (nds32_builtin_decl_impl): Ditto.
724 (builtin_description): Ditto.
725 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
726 (nds32_init_builtins_impl): Ditto.
727 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
728 (nds32_builtin_decl): New.
729 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
730 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
732 2018-03-02 Jeff Law <law@redhat.com>
734 * reorg.c (stop_search_p): Handle DEBUG_INSN.
735 (redundant_insn, fill_simple_delay_slots): Likewise.
736 (fill_slots_from_thread): Likewise.
737 * resource.c (mark_referenced_resources): Likewise.
738 (mark_set_resources, find_dead_or_set_registers): Likewise.
740 2018-03-02 Jakub Jelinek <jakub@redhat.com>
742 * substring-locations.h (format_warning_va): Formatting fix for
743 ATTRIBUTE_GCC_DIAG.
744 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
745 argument.
746 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
747 * substring-locations.c: Include intl.h.
748 (format_warning_va): Turned into small wrapper around
749 format_warning_n_va, renamed to ...
750 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
751 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
752 use ngettext.
753 (format_warning_at_substring_n): New function.
754 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
755 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
756 format_warning_at_substring with just a shorter name instead of
757 const function pointer.
758 (fmtwarn_n): New function.
759 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
760 appropriate, get rid of all the fmtstr temporaries, move conditionals
761 with G_() wrapped string literals directly into fmtwarn arguments,
762 cast dir.len to (int), formatting fixes.
764 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
766 * doc/invoke.texi: Remove "Cilk Plus" references.
768 2018-03-02 Jakub Jelinek <jakub@redhat.com>
769 Richard Biener <rguenther@suse.de>
771 PR ipa/84628
772 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
773 for error or warning attributes if CALL_FROM_THUNK_P is set.
774 Formatting fixes.
776 2018-03-02 Jakub Jelinek <jakub@redhat.com>
778 PR target/56540
779 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
780 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
782 PR target/56540
783 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
784 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
786 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
787 instead of -1U in last predictors element's probability member.
789 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
791 PR ipa/83983
792 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
793 arguments if they are comparable.
795 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
797 PR tree-optimization/84634
798 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
799 masks and masked_loop_p with a single loop_masks, making sure it's
800 null for bb vectorization.
802 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
804 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
805 (vect_analyze_data_ref_access): Use loop->safe_len rather than
806 loop->force_vectorize to check whether there is no alias.
808 2018-03-02 Jakub Jelinek <jakub@redhat.com>
810 PR target/84614
811 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
812 prototypes.
813 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
814 comments.
815 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
816 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
817 instead of a loop around prev_real_insn.
818 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
819 prev_real_insn.
821 PR inline-asm/84625
822 * config/i386/i386.c (ix86_print_operand): Use conditional
823 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
824 zero vector.
826 2018-03-02 Richard Biener <rguenther@suse.de>
828 PR tree-optimization/84427
829 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
830 (bitmap_set_subtract_values): Rewrite to handle multiple
831 exprs per value.
832 (clean): Likewise.
833 (prune_clobbered_mems): Likewise.
834 (phi_translate): Take edge instead of pred/phiblock.
835 (phi_translate_1): Likewise.
836 (phi_translate_set): Likewise. Insert all translated
837 exprs for a value into the set, keeping possibly multiple
838 expressions per value.
839 (compute_antic_aux): Adjust for phi_translate changes.
840 When intersecting union the expressions and prune those
841 not in the final value set, keeping possibly multiple
842 expressions per value. Do not use value-insertion
843 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
844 all expressions. Add verification that the value-sets
845 only shrink during iteration.
846 (compute_partial_antic_aux): Adjust for the phi_translate changes.
847 (do_pre_regular_insertion): Likewise.
848 (do_pre_partial_partial_insertion): Likewise.
850 2018-03-02 Richard Biener <rguenther@suse.de>
852 PR target/82005
853 * config/darwin.c (saved_debug_info_level): New static global.
854 (darwin_asm_lto_start): Disable debug info generation for LTO out.
855 (darwin_asm_lto_end): Restore debug info generation settings.
857 2018-03-01 Martin Liska <mliska@suse.cz>
859 PR sanitizer/82484
860 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
861 volatile arguments.
863 2018-03-01 Richard Biener <rguenther@suse.de>
865 PR debug/84645
866 * dwarf2out.c (gen_variable_die): Properly handle late VLA
867 type annotation with LTO when debug was disabled at compile-time.
869 2018-03-01 Matthew Fortune <mfortune@gmail.com>
871 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
872 XINT with INTVAL.
873 (mips_final_postscan_insn): Likewise.
875 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
877 PR rtl-optimization/84528
878 * alias.c (init_alias_target): Add commentary.
879 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
880 a unique base value if the frame pointer is not eliminated
881 to the stack pointer.
883 2018-03-01 Tom de Vries <tom@codesourcery.com>
885 PR rtl-optimization/83327
886 * lra-int.h (hard_regs_spilled_into): Declare.
887 * lra.c (hard_regs_spilled_into): Define.
888 (init_reg_info): Init hard_regs_spilled_into.
889 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
890 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
891 (process_bb_lives): Handle hard_regs_spilled_into.
892 (lra_create_live_ranges_1): Before doing liveness propagation, clear
893 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
895 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
897 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
898 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
899 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
900 * config/rs6000/aix72.h: New file.
902 2018-02-28 Jakub Jelinek <jakub@redhat.com>
904 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
905 instead of warning_at with conditional singular and plural messages
906 where possible.
908 PR target/52991
909 * stor-layout.c (update_alignment_for_field): For
910 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
911 && !DECL_PACKED (field), do the alignment update, just use
912 only desired_align instead of MAX (type_align, desired_align)
913 as the alignment.
914 (place_field): Don't do known_align < desired_align handling
915 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
916 is non-NULL, instead do it after rli->prev_field handling and
917 only if not within a bitfield word. For DECL_PACKED (field)
918 use type_align of BITS_PER_UNIT.
920 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
922 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
923 superfluous parentheses and trailing spaces.
925 2018-02-28 Richard Biener <rguenther@suse.de>
927 PR tree-optimization/84584
928 * graphite-scop-detection.c (scop_detection::add_scop): Discard
929 SCoPs with fake exit edge.
931 2018-02-28 Martin Liska <mliska@suse.cz>
933 PR testsuite/84597
934 * timevar.c (timer::print): Fix format to properly print 100%
935 values.
937 2018-02-28 Richard Biener <rguenther@suse.de>
939 PR middle-end/84607
940 * genmatch.c (capture_info::walk_match): Do not mark
941 captured expressions without operands as expr_p given
942 they act more like predicates and should be subject to
943 "lost tail" side-effect preserving.
945 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
947 PR rtl-optimization/81611
948 * auto-inc-dec.c (attempt_change): Move dead note from
949 mem_insn if it's the next use of regno
950 (find_address): Take address use of reg holding
951 non-incremented value. Add parm to limit search to the named
952 reg only.
953 (merge_in_block): Attempt to use a mem insn that is the next
954 use of the original regno.
956 2018-02-27 Martin Sebor <msebor@redhat.com>
958 PR c++/83871
959 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
960 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
962 2018-02-27 Martin Sebor <msebor@redhat.com>
964 PR translation/84207
965 * diagnostic-core.h (warning_n, error_n, inform_n): Change
966 n argument to unsigned HOST_WIDE_INT.
967 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
968 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
969 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
970 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
972 2018-02-27 Richard Biener <rguenther@suse.de>
974 PR tree-optimization/84512
975 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
976 Do not use the estimate returned from record_stmt_cost for
977 the scalar iteration cost but sum properly using add_stmt_cost.
979 2018-02-27 Richard Biener <rguenther@suse.de>
981 PR tree-optimization/84466
982 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
983 Adjust last change to less strictly validate use operands.
985 2018-02-27 Martin Liska <mliska@suse.cz>
987 PR gcov-profile/84548
988 * gcov.c (process_file): Allow partial overlap and consider it
989 also as group functions.
990 (output_lines): Properly calculate range of lines for a group.
992 2018-02-27 Martin Liska <mliska@suse.cz>
994 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
995 'ggc' suffixes. Change first column width.
996 (timer::print): Fix formatting of the column.
998 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1000 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1001 preserve inline entry blocks for the sake of debug inline
1002 entry point markers alone.
1003 (remove_unused_locals): Suggest in comments a better place to
1004 force the preservation of inline entry blocks that are
1005 otherwise unused, but do not preserve them.
1007 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1009 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1011 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1013 PR target/84039
1014 * config/i386/constraints.md (Bs): Replace
1015 ix86_indirect_branch_register with
1016 TARGET_INDIRECT_BRANCH_REGISTER.
1017 (Bw): Likewise.
1018 * config/i386/i386.md (indirect_jump): Likewise.
1019 (tablejump): Likewise.
1020 (*sibcall_memory): Likewise.
1021 (*sibcall_value_memory): Likewise.
1022 Peepholes of indirect call and jump via memory: Likewise.
1023 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1024 (*sibcall_value_GOT_32): Likewise.
1025 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1026 (GOT_memory_operand): Likewise.
1027 (call_insn_operand): Likewise.
1028 (sibcall_insn_operand): Likewise.
1029 (GOT32_symbol_operand): Likewise.
1030 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1032 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1034 PR rtl-optimization/83496
1035 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1036 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1037 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1038 redundant insn, if any.
1039 (relax_delay_slots): Likewise.
1040 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1042 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1044 PR tree-optimization/83965
1045 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1046 that grouped statements are part of a reduction chain. Return
1047 true if the statement is not marked as a reduction itself but
1048 is part of a group.
1049 (vect_recog_dot_prod_pattern): Don't check whether the statement
1050 is part of a group here.
1051 (vect_recog_sad_pattern): Likewise.
1052 (vect_recog_widen_sum_pattern): Likewise.
1054 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1056 PR debug/84545
1057 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1059 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1061 PR target/84530
1062 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1063 the bool argument.
1064 (ix86_output_indirect_function_return): New prototype.
1065 (ix86_split_simple_return_pop_internal): Likewise.
1066 * config/i386/i386.c (indirect_return_via_cx): New.
1067 (indirect_return_via_cx_bnd): Likewise.
1068 (indirect_thunk_name): Handle return va CX_REG.
1069 (output_indirect_thunk_function): Create alias for
1070 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1071 (ix86_output_indirect_jmp): Remove the bool argument.
1072 (ix86_output_indirect_function_return): New function.
1073 (ix86_split_simple_return_pop_internal): Likewise.
1074 * config/i386/i386.md (*indirect_jump): Don't pass false
1075 to ix86_output_indirect_jmp.
1076 (*tablejump_1): Likewise.
1077 (simple_return_pop_internal): Change it to define_insn_and_split.
1078 Call ix86_split_simple_return_pop_internal to split it for
1079 -mfunction-return=.
1080 (simple_return_indirect_internal): Call
1081 ix86_output_indirect_function_return instead of
1082 ix86_output_indirect_jmp.
1084 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1086 PR bootstrap/84405
1087 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1088 memset and value initialization afterwards.
1090 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1092 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1094 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1096 PR target/84521
1097 * common/config/aarch64/aarch64-common.c
1098 (aarch_option_optimization_table[]): Switch
1099 off fomit-frame-pointer
1101 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1102 Chung-Ju Wu <jasonwucj@gmail.com>
1104 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1105 volatile memory.
1106 (store_multiple): Ditto.
1108 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1110 * config.gcc: Add --with-cpu support for nds32 target.
1111 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1112 * config/nds32/nds32.opt: Add -mcpu= option.
1114 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1116 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1117 isel=yes): Warn for these deprecated options.
1119 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1121 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1122 ISA_2_5_MASKS_EMBEDDED.
1124 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1126 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1127 p->max as pointers rather than using iterative_hash_expr.
1129 2018-02-23 Carl Love <cel@us.ibm.com>
1131 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1132 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1133 BU_P8V_OVERLOAD_2.
1134 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1135 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1136 P8V_BUILTIN_VEC_VUNSIGNED2.
1138 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1140 PR target/81572
1141 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1142 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1143 LRA_UNKNOWN_ALT.
1144 * lra-constraints.c (curr_insn_transform): Set up
1145 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1146 LRA_UNKNOWN_ALT.
1147 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1148 * lra-eliminations.c (spill_pseudos): Ditto.
1149 (process_insn_for_elimination): Ditto.
1150 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1151 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1152 LRA_NON_CLOBBERED_ALT.
1154 2018-02-22 Martin Sebor <msebor@redhat.com>
1156 PR tree-optimization/84480
1157 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1158 to maybe_diag_stxncpy_trunc. Call it.
1159 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1160 from gimple_fold_builtin_strcpy. Print inlining stack.
1161 (handle_builtin_stxncpy): Print inlining stack.
1162 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1164 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1166 PR target/84176
1167 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1168 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1169 and -fcheck-pointer-bounds are used together.
1170 (indirect_thunk_prefix): New enum.
1171 (indirect_thunk_need_prefix): New function.
1172 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1173 "_nt" instead of "_bnd" for NOTRACK prefix.
1174 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1175 (output_indirect_thunk_function): Likewise.
1176 (): Likewise.
1177 (ix86_code_end): Update output_indirect_thunk_function calls.
1178 (ix86_output_indirect_branch_via_reg): Replace
1179 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1180 (ix86_output_indirect_branch_via_push): Likewise.
1181 (ix86_output_function_return): Likewise.
1182 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1183 incompatible with -fcf-protection=branch and
1184 -fcheck-pointer-bounds.
1186 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1188 PR target/83335
1189 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1190 Change gcc_assert call to output_operand_lossage.
1192 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1194 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1196 2018-02-22 DJ Delorie <dj@redhat.com>
1197 Sebastian Perta <sebastian.perta@renesas.com>
1198 Oleg Endo <olegendo@gcc.gnu.org>
1200 * config/rx/rx.c (rx_rtx_costs): New function.
1201 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1203 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1205 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1207 2018-02-22 Martin Liska <mliska@suse.cz>
1209 PR driver/83193
1210 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1211 Add "native" as a possible value.
1213 2018-02-22 Martin Liska <mliska@suse.cz>
1215 PR driver/83193
1216 * config/i386/i386.c (ix86_option_override_internal):
1217 Add "native" as a possible value for -march and -mtune.
1219 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1221 PR target/84502
1222 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1223 to all type variants.
1225 PR tree-optimization/84503
1226 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1227 width as info->bitpos + info->bitsize - start.
1228 (merged_store_group::merge_overlapping): Simplify width computation.
1229 (check_no_overlap): New function.
1230 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1231 start + width and last_order of the group, fail if check_no_overlap
1232 fails.
1233 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1234 to group if check_no_overlap fails.
1236 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1238 * config/rs6000/altivec.md: Delete contraint arguments to
1239 define_expand, define_split, and define_peephole2, and in
1240 define_insn_and_split if always unused.
1241 * config/rs6000/darwin.md: Ditto.
1242 * config/rs6000/dfp.md: Ditto.
1243 * config/rs6000/rs6000.md: Ditto.
1244 * config/rs6000/sync.md: Ditto.
1245 * config/rs6000/vector.md: Ditto.
1246 * config/rs6000/vsx.md: Ditto.
1248 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1250 * config/rs6000/altivec.md: Write output control strings as braced
1251 blocks instead of double-quoted strings.
1252 * config/rs6000/darwin.md: Ditto.
1253 * config/rs6000/rs6000.md: Ditto.
1254 * config/rs6000/vector.md: Ditto.
1255 * config/rs6000/vsx.md: Ditto.
1257 2018-02-21 Jason Merrill <jason@redhat.com>
1259 PR c++/84314 - ICE with templates and fastcall attribute.
1260 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1262 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1264 * ipa-cp.c (determine_versionability): Fix comment typos.
1266 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1268 PR c/84229
1269 * ipa-cp.c (determine_versionability): Do not version functions caling
1270 va_arg_pack.
1272 2018-02-21 Martin Liska <mliska@suse.cz>
1274 PR driver/83193
1275 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1276 Add "native" as a possible value.
1277 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1278 the macro when native cpu detection is available.
1280 2018-02-21 Martin Liska <mliska@suse.cz>
1282 PR driver/83193
1283 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1284 Add "native" as a possible value.
1285 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1286 when native cpu detection is available.
1288 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1289 Martin Sebor <msebor@redhat.com>
1291 PR tree-optimization/84478
1292 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1293 false.
1294 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1295 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1296 support which is conservatively correct, for 2 only stay conservative
1297 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1298 argument to the 2 argument get_range_strlen, adjust 6 arg
1299 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1300 false.
1301 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1302 (gimple_fold_builtin_strlen): Pass true as last argument to
1303 get_range_strlen.
1305 2018-02-20 Martin Sebor <msebor@redhat.com>
1307 PR middle-end/84095
1308 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1309 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1310 (builtin_memref::builtin_memref): Factor out parts into
1311 set_base_and_offset and call it.
1313 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1315 PR middle-end/84406
1316 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1317 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1318 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1319 search at the associated MODE_INT.
1321 2018-02-20 Jeff Law <law@redhat.com>
1323 PR middle-end/82123
1324 PR tree-optimization/81592
1325 PR middle-end/79257
1326 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1327 for range data rather than using global data.
1328 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1329 range data rather than using global data.
1330 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1331 pass it to children as needed.
1332 (struct directive::fmtresult): Similarly.
1333 (struct directive::set_width): Similarly.
1334 (struct directive::set_precision): Similarly.
1335 (format_integer, format_directive, parse_directive): Similarly.
1336 (format_none): Accept unnamed vr_values parameter.
1337 (format_percent, format_floating, format_character): Similarly.
1338 (format_string, format_plain): Similarly.
1339 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1340 the EVRP range analyzer for range data rather than using global data.
1341 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1342 gimple-ssa-evrp-analyze.h
1343 (class sprintf_dom_walker): Add after_dom_children member function.
1344 Add evrp_range_analyzer member.
1345 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1346 range analyzer as needed.
1347 (sprintf_dom_walker::after_dom_children): New member function.
1348 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1349 if not optimizing.
1350 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1351 (evrp_range_analyzer::pop_to_marker): Likewise.
1353 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1355 PR tree-optimization/84419
1356 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1357 with the required type if its current type is compatible but
1358 different.
1360 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1362 PR middle-end/82004
1363 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1364 after vectorization.
1366 2018-02-20 Martin Liska <mliska@suse.cz>
1368 PR driver/83193
1369 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1370 possible values if we don't have a hint.
1372 2018-02-20 Martin Liska <mliska@suse.cz>
1374 PR c/84310
1375 PR target/79747
1376 * final.c (shorten_branches): Build align_tab array with one
1377 more element.
1378 * opts.c (finish_options): Add alignment option limit check.
1379 (MAX_CODE_ALIGN): Likewise.
1380 (MAX_CODE_ALIGN_VALUE): Likewise.
1381 * doc/invoke.texi: Document maximum allowed option value for
1382 all -falign-* options.
1384 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1386 PR target/84146
1387 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1388 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1389 * var-tracking.c (emit_note_insn_var_location): Remove all references
1390 to NOTE_INSN_CALL_ARG_LOCATION.
1391 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1392 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1393 Use copy_rtx_if_shared.
1394 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1395 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1396 (dwarf2out_var_location): Remove handling of
1397 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1398 on call_insn.
1399 * final.c (final_scan_insn): Remove all references to
1400 NOTE_INSN_CALL_ARG_LOCATION.
1401 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1402 before dumping final insns.
1403 * except.c (emit_note_eh_region_end): Remove all references to
1404 NOTE_INSN_CALL_ARG_LOCATION.
1405 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1406 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1407 * config/arc/arc.c (hwloop_optimize): Likewise.
1408 * config/arm/arm.c (create_fix_barrier): Likewise.
1409 * config/s390/s390.c (s390_chunkify_start): Likewise.
1410 * config/sh/sh.c (find_barrier): Likewise.
1411 * config/i386/i386.c (rest_of_insert_endbranch,
1412 ix86_seh_fixup_eh_fallthru): Likewise.
1413 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1414 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1415 * config/frv/frv.c (frv_function_prologue): Likewise.
1416 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1417 reg note.
1418 (note_outside_basic_block_p): Remove all references to
1419 NOTE_INSN_CALL_ARG_LOCATION.
1420 * gengtype.c (adjust_field_rtx_def): Likewise.
1421 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1422 Likewise.
1423 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1424 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1426 PR c++/84444
1427 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1428 is ADDR_EXPR.
1430 PR tree-optimization/84452
1431 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1432 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1433 is NULL.
1435 2018-02-19 Martin Liska <mliska@suse.cz>
1437 PR sanitizer/82183
1438 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1440 2018-02-19 Martin Liska <mliska@suse.cz>
1441 Richard Sandiford <richard.sandiford@linaro.org>
1443 PR tree-optimization/82491
1444 * gimple-fold.c (get_base_constructor): Make earlier bail out
1445 to prevent ubsan.
1447 2018-02-19 Carl Love <cel@us.ibm.com>
1449 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1450 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1451 BU_P8V_OVERLOAD_1.
1452 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1453 P8V_BUILTIN_VEC_NEG.
1455 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1457 * config/rl78/rl78.md (movdf): New define expand.
1459 2018-02-19 Martin Liska <mliska@suse.cz>
1461 PR other/80589
1462 * doc/invoke.texi: Fix typo.
1463 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1465 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1467 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1468 handle rs6000_single_float and rs6000_double_float specially for
1469 e500 family CPUs.
1471 2018-02-16 Jeff Law <law@redhat.com>
1473 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1474 (pop_regs): Use it.
1476 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1478 PR ipa/84425
1479 * ipa-inline.c (inline_small_functions): Fix a typo.
1481 2018-02-16 Nathan Sidwell <nathan@acm.org>
1483 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1485 2018-02-16 Carl Love <cel@us.ibm.com>
1487 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1488 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1489 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1490 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1491 expansion to P8V_BUILTIN_VEC_FLOAT2.
1493 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1495 PR rtl-optimization/70023
1496 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1497 src_regno into account.
1499 2018-02-16 Carl Love <cel@us.ibm.com>
1501 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1502 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1503 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1504 * config/rs6000/rs6000.c: Remove case statements for
1505 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1506 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1507 and P9V_BUILTIN_VEC_VINSERT4B.
1508 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1509 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1510 * config/rs6000/vsx.md:
1511 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1512 vec_insert4b.
1514 2018-02-16 Carl Love <cel@us.ibm.com>
1516 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1517 vec_insert4b.
1518 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1519 definitions.
1520 * config/rs6000/rs6000-c.c: Add the definitions for
1521 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1522 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1523 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1524 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1525 definition for insert4b and define insn *insert3b_internal.
1526 * doc/extend.texi: Add documentation for vec_extract4b.
1528 2018-02-16 Nathan Sidwell <nathan@acm.org>
1530 * doc/extend.texi (Backwards Compatibility): Mention friend
1531 injection. Note for-scope is deprecated.
1532 * doc/invoke.texi (-ffriend-injection): Deprecate.
1534 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1536 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1537 that moved to I2, also allow destinations that are a paradoxical
1538 subreg (instead of a normal reg).
1540 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1542 PR target/83831
1543 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1544 to QImode.
1546 2018-02-16 Richard Biener <rguenther@suse.de>
1548 PR tree-optimization/84037
1549 PR tree-optimization/84016
1550 PR target/82862
1551 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1552 Adjust vec_construct for the fact we need additional higher latency
1553 128bit inserts for AVX256 and AVX512 vector builds.
1554 (ix86_add_stmt_cost): Scale vector construction cost for
1555 elementwise loads.
1557 2018-02-16 Richard Biener <rguenther@suse.de>
1559 PR tree-optimization/84417
1560 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1561 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1562 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1564 2018-02-16 Martin Liska <mliska@suse.cz>
1566 PR sanitizer/84307
1567 * internal-fn.def (ASAN_CHECK): Set proper flags.
1568 (ASAN_MARK): Likewise.
1570 2018-02-16 Julia Koval <julia.koval@intel.com>
1572 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1573 from PTA_CANNONLAKE.
1575 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1577 PR target/84272
1578 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1579 Use ++iter rather than iter++ for std::list iterators.
1580 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1581 defer deleting them until all nodes in the forest are processed. Do
1582 free even leaf nodes. Change to_process into auto_vec.
1584 PR bootstrap/84405
1585 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1586 * vec.h (vec_default_construct): Use memset instead of placement new
1587 if BROKEN_VALUE_INITIALIZATION is defined.
1588 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1589 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1590 is defined.
1592 PR rtl-optimization/83723
1593 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1594 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1595 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1596 recursive calls.
1597 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1598 callers.
1599 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1601 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1603 PR rtl-optimization/81443
1604 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1605 from inner REGs to paradoxical SUBREGs.
1607 2018-02-16 Richard Biener <rguenther@suse.de>
1609 PR tree-optimization/84399
1610 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1611 For operands we can analyze at their definition make sure we can
1612 analyze them at each use as well.
1614 2018-02-16 Richard Biener <rguenther@suse.de>
1616 PR tree-optimization/84190
1617 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1618 volatile accesses if the decl isn't volatile.
1620 2018-02-15 Jason Merrill <jason@redhat.com>
1622 PR c++/84314 - ICE with templates and fastcall attribute.
1623 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1624 TYPE_CANONICAL on an existing type.
1626 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1628 PR tree-optimization/84383
1629 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1630 dstoff nor call operand_equal_p if dstbase is NULL.
1632 PR tree-optimization/84334
1633 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1634 also a CONSTANT_CLASS_P, punt.
1636 2018-02-14 Jim Wilson <jimw@sifive.com>
1638 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1639 first SMALL_OPERAND check. New local min_second_step. Move assert
1640 to where locals are set. Add TARGET_RVC support.
1641 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1643 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1645 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1647 2018-02-14 Martin Sebor <msebor@redhat.com>
1649 PR tree-optimization/83698
1650 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1651 arrays constrain the offset range to their bounds.
1652 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1653 (builtin_access::overlap): Avoid setting the size of overlap if it's
1654 already been set.
1655 (maybe_diag_overlap): Also consider arrays when deciding what values
1656 of offsets to include in diagnostics.
1658 2018-02-14 Martin Sebor <msebor@redhat.com>
1660 PR c/84108
1661 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1662 that correspond to the kind of a declaration.
1664 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1666 PR target/83984
1667 * config/pa/pa.md: Load address of PIC label using the linkage table
1668 if the label is nonlocal.
1670 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1672 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1673 warning message if user requests -maltivec=be.
1674 * doc/invoke.texi: Document deprecation of -maltivec=be.
1676 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1678 PR target/84220
1679 * config/rs6000/rs6000-c.c: Update definitions for
1680 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1681 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1683 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1685 PR target/84239
1686 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1687 add _get_ssp intrinsics. Remove argument from
1688 __builtin_ia32_rdssp[d|q].
1689 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1690 * config/i386/i386-builtin.def: Remove argument from
1691 __builtin_ia32_rdssp[d|q].
1692 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1693 ix86_expand_special_args_builtin for _rdssp[d|q].
1694 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1695 Clear register before usage.
1696 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1697 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1699 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1701 PR tree-optimization/84357
1702 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1703 operand 1 of an ARRAY_REF too.
1705 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1707 PR target/83831
1708 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1709 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1710 declarations.
1711 (set_of_reg): New struct.
1712 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1713 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1714 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1715 functions.
1716 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1717 Split into bitclr, bitset, bitinvert patterns if appropriate.
1718 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1719 use rx_fuse_in_memory_bitop.
1720 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1721 to named insn, correct maximum insn length.
1723 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1725 PR target/79242
1726 * machmode.def: Define a complex mode for PARTIAL_INT.
1727 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1728 MODE_PARTIAL_INT.
1729 * doc/rtl.texi: Document CSPImode.
1730 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1731 handling.
1732 (msp430_hard_regno_nregs_with_padding): Likewise.
1734 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1736 PR target/84279
1737 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1739 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1741 PR rtl-optimization/84169
1742 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1743 we generated a parallel as new i3 and we split that to new i2 and i3
1744 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1745 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1746 those to i2, not i1. Partially rewrite this scan code.
1748 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1750 PR c/82210
1751 * stor-layout.c (place_field): For variable length fields, adjust
1752 offset_align afterwards not just based on the field's alignment,
1753 but also on the size.
1755 PR middle-end/84309
1756 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1757 of exps and logs in the use_exp2 case.
1759 2018-02-13 Jeff Law <law@redhat.com>
1761 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1762 entry for "vector".
1764 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1765 ARGS as unused.
1767 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1769 PR debug/84342
1770 PR debug/84319
1771 * common.opt (gas-loc-support, gas-locview-support): New.
1772 (ginline-points, ginternal-reset-location-views): New.
1773 * doc/invoke.texi: Document them. Use @itemx where intended.
1774 (gvariable-location-views): Adjust.
1775 * target.def (reset_location_view): New.
1776 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1777 (TARGET_RESET_LOCATION_VIEW): New.
1778 * doc/tm.texi: Rebuilt.
1779 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1780 (dwarf2out_default_as_locview_support): New.
1781 (output_asm_line_debug_info): Use option variables.
1782 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1783 (output_loc_list): Likewise.
1784 (add_high_low_attributes): Check option variables.
1785 Don't output entry view attribute in strict mode.
1786 (gen_inlined_subroutine_die): Check option variables.
1787 (dwarf2out_inline_entry): Likewise.
1788 (init_sections_and_labels): Likewise.
1789 (dwarf2out_early_finish): Likewise.
1790 (maybe_reset_location_view): New, from...
1791 (dwarf2out_var_location): ... here. Call it.
1792 * debug.h (dwarf2out_default_as_loc_support): Declare.
1793 (dwarf2out_default_as_locview_support): Declare.
1794 * hooks.c (hook_int_rtx_insn_0): New.
1795 * hooks.h (hook_int_rtx_insn_0): Declare.
1796 * toplev.c (process_options): Take -gas-loc-support and
1797 -gas-locview-support from dwarf2out. Enable
1798 -gvariable-location-views by default only with locview
1799 assembler support. Enable -ginternal-reset-location-views by
1800 default only if the target defines the corresponding hook.
1801 Enable -ginline-points by default if location views are
1802 enabled; force it disabled if statement frontiers are
1803 disabled.
1804 * tree-inline.c (expand_call_inline): Check option variables.
1805 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1807 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1809 PR tree-optimization/84321
1810 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1811 handling. Also check whether the anti-range contains any values
1812 that satisfy the mask; switch to a VR_RANGE if not.
1814 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1816 PR sanitizer/84340
1817 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1819 2018-02-13 Martin Jambor <mjambor@suse.cz>
1821 PR c++/83990
1822 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1823 of call statements, also set location of a load to a temporary.
1825 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1827 * config/rl78/rl78.c (add_vector_labels): New function.
1828 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1829 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1830 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1831 which checks that no arguments are passed.
1832 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1833 * doc/extend.texi: Documentation for the new attribute.
1835 2018-02-13 Andreas Schwab <schwab@suse.de>
1837 * config/riscv/linux.h (CPP_SPEC): Define.
1839 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1841 PR target/84335
1842 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1843 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1844 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1845 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1846 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1847 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1848 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1849 temporarily for AES and PCLMUL builtins.
1851 PR tree-optimization/84339
1852 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1853 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1854 Formatting fixes.
1856 PR middle-end/84309
1857 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1858 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1859 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1860 inline function.
1861 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1862 inline function.
1863 * omp-simd-clone.h: New file.
1864 * omp-simd-clone.c: Include omp-simd-clone.h.
1865 (expand_simd_clones): No longer static.
1866 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1867 cgraph.h and omp-simd-clone.h.
1868 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1869 (vect_recog_widen_shift_pattern): Formatting fix.
1870 (vect_pattern_recog_1): Don't check optab for calls.
1872 PR target/84336
1873 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1874 operands[2] into a REG before using gen_lowpart on it.
1876 2018-02-12 Jeff Law <law@redhat.com>
1878 PR target/83760
1879 * config/sh/sh.c (find_barrier): Consider a sibling call
1880 a barrier as well.
1882 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1883 successfully back substituting a reg.
1885 2018-02-12 Richard Biener <rguenther@suse.de>
1887 PR tree-optimization/84037
1888 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1889 parameter, move visited init to caller.
1890 (vect_slp_analyze_operations): Separate cost from validity
1891 check, initialize visited once for all instances.
1892 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1893 for all instances.
1894 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1895 out an assert.
1896 (vect_model_promotion_demotion_cost): Likewise.
1897 (vectorizable_bswap): Guard cost modeling with !slp_node
1898 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1899 SLP stmts.
1900 (vectorizable_call): Likewise.
1901 (vectorizable_conversion): Likewise.
1902 (vectorizable_assignment): Likewise.
1903 (vectorizable_shift): Likewise.
1904 (vectorizable_operation): Likewise.
1905 (vectorizable_store): Likewise.
1906 (vectorizable_load): Likewise.
1907 (vectorizable_condition): Likewise.
1908 (vectorizable_comparison): Likewise.
1910 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1912 PR sanitizer/84307
1913 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1914 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1915 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1917 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1919 PR middle-end/83665
1920 * params.def (inline-min-speedup): Increase from 8 to 15.
1921 (max-inline-insns-auto): Decrease from 40 to 30.
1922 * ipa-split.c (consider_split): Add some buffer for function to
1923 be considered inlining candidate.
1924 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1925 default values.
1927 2018-02-12 Richard Biener <rguenther@suse.de>
1929 PR tree-optimization/84037
1930 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1931 matched stmts if we cannot swap the non-matched ones.
1933 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1935 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1936 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1937 _mm_maskz_scalef_round_ss): New intrinsics.
1938 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1939 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1940 __builtin_ia32_scalefss_round): Remove.
1941 (__builtin_ia32_scalefsd_mask_round,
1942 __builtin_ia32_scalefss_mask_round): New intrinsics.
1943 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1944 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1945 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1946 "<round_constraint>")): Changed to ...
1947 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1948 "<round_scalar_constraint>")): ... this.
1949 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1950 %0, %1, %2<round_op3>}"): Changed to ...
1951 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1952 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1953 %2<round_scalar_mask_op3>}"): ... this.
1954 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1956 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1958 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1959 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1960 (_mm_maskz_sqrt_round_ss): New intrinsics.
1961 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1962 (__builtin_ia32_sqrtsd_mask_round)
1963 (__builtin_ia32_sqrtss_mask_round): New builtins.
1964 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1965 (__builtin_ia32_sqrtss_round): Remove.
1966 (__builtin_ia32_sqrtsd_mask_round)
1967 (__builtin_ia32_sqrtss_mask_round): New builtins.
1968 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1969 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1970 ((match_operand:VF_128 1 "vector_operand"
1971 "xBm,<round_constraint>")): Changed to ...
1972 ((match_operand:VF_128 1 "vector_operand"
1973 "xBm,<round_scalar_constraint>")): ... this.
1974 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1975 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1976 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1977 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1978 %<iptr>1<round_scalar_mask_op3>}): ... this.
1979 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1980 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1982 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1984 PR target/84266
1985 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1986 Cast vec_cmpeq result to correct type.
1987 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1988 Cast vec_cmpgt result to correct type.
1990 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1992 * final.c (final_scan_insn_1): Renamed from...
1993 (final_scan_insn): ... this. New wrapper, to recover
1994 seen from the outermost call in recursive ones.
1995 * config/sparc/sparc.c (output_return): Drop seen from call.
1996 (output_sibcall): Likewise.
1997 * config/visium/visium.c (output_branch): Likewise.
1999 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2001 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2002 function label.
2004 2018-02-10 Alan Modra <amodra@gmail.com>
2006 PR target/84300
2007 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2008 Specify LR as an input.
2010 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2012 PR sanitizer/83987
2013 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2014 remove_member_access_dummy_vars): New functions.
2015 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2016 lower_omp_1, execute_lower_omp): Use them.
2018 PR rtl-optimization/84308
2019 * shrink-wrap.c (spread_components): Release todo vector.
2021 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2023 PR rtl-optimization/57193
2024 * ira-color.c (struct allocno_color_data): Add member
2025 conflict_allocno_hard_prefs.
2026 (update_conflict_allocno_hard_prefs): New.
2027 (bucket_allocno_compare_func): Add a preference based on
2028 conflict_allocno_hard_prefs.
2029 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2030 (color_allocnos): Remove a dead code. Initiate
2031 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2033 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2035 PR target/84226
2036 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2037 constraint from =wa to wa. Avoid a subreg on the output operand,
2038 instead use a pseudo and subreg it in a move.
2039 (p9_xxbrd_<mode>): Changed to ...
2040 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2041 (p9_xxbrd_v2df): New expander.
2042 (p9_xxbrw_<mode>): Changed to ...
2043 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2044 (p9_xxbrw_v4sf): New expander.
2046 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2048 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2050 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2052 PR target/83926
2053 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2054 multiply in 32-bit mode.
2055 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2056 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2057 mode.
2059 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2061 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2062 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2063 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2064 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2066 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2068 PR lto/84213
2069 * dwarf2out.c (is_trivial_indirect_ref): New function.
2070 (dwarf2out_late_global_decl): Do not generate a location
2071 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2072 and that are not defined in the current unit.
2074 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2076 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2077 instead of a libcall for UNORDERED.
2079 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2081 PR target/82641
2082 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2083 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2085 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2087 PR target/PR84295
2088 * config/s390/s390.c (s390_set_current_function): Invoke
2089 s390_indirect_branch_settings also if fndecl didn't change.
2091 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2093 * config/rs6000/rs6000.md (blockage): Set length to zero.
2095 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2097 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2099 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2101 PR sanitizer/84285
2102 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2103 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2104 -static-lib*san.
2106 PR debug/84252
2107 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2108 PARALLEL incoming that failed vt_get_decl_and_offset check.
2110 PR middle-end/84237
2111 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2112 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2113 TREE_READONLY bit.
2114 (get_variable_section): For decls in named .bss* sections pass true as
2115 second argument to bss_initializer_p.
2117 2018-02-09 Marek Polacek <polacek@redhat.com>
2118 Jakub Jelinek <jakub@redhat.com>
2120 PR c++/83659
2121 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2122 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2123 Sync some changes from cxx_fold_indirect_ref.
2125 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2127 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2128 markers.
2129 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2130 (BLOCK_INLINE_ENTRY_LABEL): New.
2131 (dwarf2out_var_location): Disregard inline entry markers.
2132 (inline_entry_data): New struct.
2133 (inline_entry_data_hasher): New hashtable type.
2134 (inline_entry_data_hasher::hash): New.
2135 (inline_entry_data_hasher::equal): New.
2136 (inline_entry_data_table): New variable.
2137 (add_high_low_attributes): Add DW_AT_entry_pc and
2138 DW_AT_GNU_entry_view attributes if a pending entry is found
2139 in inline_entry_data_table. Add old entry_pc attribute only
2140 if debug nonbinding markers are disabled.
2141 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2142 markers are enabled.
2143 (block_within_block_p, dwarf2out_inline_entry): New.
2144 (dwarf2out_finish): Check that no entries remained in
2145 inline_entry_data_table.
2146 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2147 (final_scan_insn, notice_source_line): Likewise.
2148 (rest_of_clean_state): Skip inline entry markers.
2149 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2150 markers.
2151 * gimple.c (gimple_build_debug_inline_entry): New.
2152 * gimple.h (enum gimple_debug_subcode): Add
2153 GIMPLE_DEBUG_INLINE_ENTRY.
2154 (gimple_build_debug_inline_entry): Declare.
2155 (gimple_debug_inline_entry_p): New.
2156 (gimple_debug_nonbind_marker_p): Adjust.
2157 * insn-notes.def (INLINE_ENTRY): New.
2158 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2159 inline entry marker notes.
2160 (print_insn): Likewise.
2161 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2162 (INSN_DEBUG_MARKER_KIND): Likewise.
2163 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2164 * tree-inline.c (expand_call_inline): Build and insert
2165 debug_inline_entry stmt.
2166 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2167 inline entry blocks early, if nonbind markers are enabled.
2168 (dump_scope_block): Dump fragment info.
2169 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2170 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2171 (gimple_build_debug_inline_entry): New.
2172 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2173 Enable/disable inline entry points too.
2174 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2175 (DEBUG_INSN): Describe inline entry markers.
2177 * common.opt (gvariable-location-views): New.
2178 (gvariable-location-views=incompat5): New.
2179 * config.in: Rebuilt.
2180 * configure: Rebuilt.
2181 * configure.ac: Test assembler for view support.
2182 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2183 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2184 * dwarf2out.c (var_loc_view): New typedef.
2185 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2186 (dwarf2out_locviews_in_attribute): New.
2187 (dwarf2out_locviews_in_loclist): New.
2188 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2189 (enum dw_line_info_opcode): Add LI_adv_address.
2190 (struct dw_line_info_table): Add view.
2191 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2192 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2193 (zero_view_p): New variable.
2194 (ZERO_VIEW_P): New macro.
2195 (output_asm_line_debug_info): New.
2196 (struct var_loc_node): Add view.
2197 (add_AT_view_list, AT_loc_list): New.
2198 (add_var_loc_to_decl): Add view param. Test it against last.
2199 (new_loc_list): Add view params. Record them.
2200 (AT_loc_list_ptr): Handle loc and view lists.
2201 (view_list_to_loc_list_val_node): New.
2202 (print_dw_val): Handle dw_val_class_view_list.
2203 (size_of_die): Likewise.
2204 (value_format): Likewise.
2205 (loc_list_has_views): New.
2206 (gen_llsym): Set vl_symbol too.
2207 (maybe_gen_llsym, skip_loc_list_entry): New.
2208 (dwarf2out_maybe_output_loclist_view_pair): New.
2209 (output_loc_list): Output view list or entries too.
2210 (output_view_list_offset): New.
2211 (output_die): Handle dw_val_class_view_list.
2212 (output_dwarf_version): New.
2213 (output_compilation_unit_header): Use it.
2214 (output_skeleton_debug_sections): Likewise.
2215 (output_rnglists, output_line_info): Likewise.
2216 (output_pubnames, output_aranges): Update version comments.
2217 (output_one_line_info_table): Output view numbers in asm comments.
2218 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2219 Call maybe_gen_llsym.
2220 (loc_list_from_tree_1): Adjust.
2221 (add_AT_location_description): Create view list attribute if
2222 needed, check it's absent otherwise.
2223 (convert_cfa_to_fb_loc_list): Adjust.
2224 (maybe_emit_file): Call output_asm_line_debug_info for test.
2225 (dwarf2out_var_location): Reset views as needed. Precompute
2226 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2227 attribute. Set view.
2228 (new_line_info_table): Reset next view.
2229 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2230 (dwarf2out_source_line): Likewise. Output view resets and labels to
2231 the assembler, or select appropriate line info opcodes.
2232 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2233 (optimize_string_length): Catch it. Adjust.
2234 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2235 dw_val_class_view_list, and remove it if no longer needed.
2236 (hash_loc_list): Hash view numbers.
2237 (loc_list_hasher::equal): Compare them.
2238 (optimize_location_lists): Check whether a view list symbol is
2239 needed, and whether the locview attribute is present, and
2240 whether they match. Remove the locview attribute if no longer
2241 needed.
2242 (index_location_lists): Call skip_loc_list_entry for test.
2243 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2244 Use output_dwarf_version.
2245 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2246 (struct dw_val_node): Add val_view_list.
2247 * final.c (SEEN_NEXT_VIEW): New.
2248 (set_next_view_needed): New.
2249 (clear_next_view_needed): New.
2250 (maybe_output_next_view): New.
2251 (final_start_function): Rename to...
2252 (final_start_function_1): ... this. Take pointer to FIRST,
2253 add SEEN parameter. Emit param bindings in the initial view.
2254 (final_start_function): Reintroduce SEEN-less interface.
2255 (final): Rename to...
2256 (final_1): ... this. Take SEEN parameter. Output final pending
2257 next view at the end.
2258 (final): Reintroduce seen-less interface.
2259 (final_scan_insn): Output pending next view before switching
2260 sections or ending a block. Mark the next view as needed when
2261 outputting variable locations. Notify debug backend of section
2262 changes, and of location view changes.
2263 (rest_of_handle_final): Adjust.
2264 * toplev.c (process_options): Autodetect value for debug variable
2265 location views option. Warn on incompat5 without -gdwarf-5.
2266 * doc/invoke.texi (gvariable-location-views): New.
2267 (gvariable-location-views=incompat5): New.
2268 (gno-variable-location-views): New.
2270 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2272 PR tree-optimization/84136
2273 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2274 that the result of find_edge is non-NULL.
2276 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2278 PR target/83008
2279 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2280 storing integer register in SImode. Fix cost of 256 and 512
2281 byte aligned SSE register store.
2283 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2285 * config/i386/i386.c (ix86_multiplication_cost): Fix
2286 multiplication cost for TARGET_AVX512DQ.
2288 2018-02-08 Marek Polacek <polacek@redhat.com>
2290 PR tree-optimization/84238
2291 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2292 get_range_strlen.
2294 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2296 PR tree-optimization/84265
2297 * tree-vect-stmts.c (vectorizable_store): Don't treat
2298 VMAT_CONTIGUOUS accesses as grouped.
2299 (vectorizable_load): Likewise.
2301 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2303 PR tree-optimization/81635
2304 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2305 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2306 (test_round_for_mask): New functions.
2307 (wide_int_cc_tests): Call test_round_for_mask.
2308 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2309 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2310 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2311 range returned by get_range_info.
2313 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2315 PR ipa/81360
2316 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2317 * symtab.c: Include builtins.h
2318 (symtab_node::output_to_lto_symbol_table_p): Move here
2319 from lto-streamer-out.c:output_symbol_p.
2320 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2321 (output_symbol_p): Move all logic to symtab.c
2322 (produce_symtab): Update.
2324 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2326 * config/s390/s390-opts.h (enum indirect_branch): Define.
2327 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2328 (s390_indirect_branch_via_thunk)
2329 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2330 (enum s390_indirect_branch_type): Define.
2331 * config/s390/s390.c (struct s390_frame_layout, struct
2332 machine_function): Remove.
2333 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2334 (indirect_branch_table_label_no, indirect_branch_table_name):
2335 Define variables.
2336 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2337 (enum s390_indirect_branch_option): Define.
2338 (s390_return_addr_from_memory): New function.
2339 (s390_handle_string_attribute): New function.
2340 (s390_attribute_table): Add new attribute handler.
2341 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2342 (s390_indirect_branch_via_thunk): New function.
2343 (s390_indirect_branch_via_inline_thunk): New function.
2344 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2345 sibling call optimization for non z10 compiles.
2346 (s390_emit_call): Force indirect branch target to be a single
2347 register. Add r1 clobber for non-z10 compiles.
2348 (s390_emit_epilogue): Emit return jump via return_use expander.
2349 (s390_reorg): Handle JUMP_INSNs as execute targets.
2350 (s390_option_override_internal): Perform validity checks for the
2351 new command line options.
2352 (s390_indirect_branch_attrvalue): New function.
2353 (s390_indirect_branch_settings): New function.
2354 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2355 (s390_output_indirect_thunk_function): New function.
2356 (s390_code_end): Implement target hook.
2357 (s390_case_values_threshold): Implement target hook.
2358 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2359 macros.
2360 * config/s390/s390.h (struct s390_frame_layout)
2361 (struct machine_function): Move here from s390.c.
2362 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2363 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2364 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2365 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2366 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2367 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2368 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2369 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2370 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2371 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2372 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2373 (mnemonic attribute): Add values which aren't recognized
2374 automatically.
2375 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2376 pattern for branch conversion. Fix mnemonic attribute.
2377 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2378 indirect branch via thunk if requested.
2379 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2380 ("*indirect_jump"): Disable for branch conversion using out of
2381 line thunks.
2382 ("indirect_jump_via_thunk<mode>_z10")
2383 ("indirect_jump_via_thunk<mode>")
2384 ("indirect_jump_via_inlinethunk<mode>_z10")
2385 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2386 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2387 ("casesi_jump_via_inlinethunk<mode>_z10")
2388 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2389 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2390 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2391 ("*indirect2_jump"): Disable for branch conversion.
2392 ("casesi_jump"): Turn into expander and expand patterns for branch
2393 conversion.
2394 ("return_use"): New expander.
2395 ("*return"): Emit return via thunk and rename it to ...
2396 ("*return<mode>"): ... this one.
2397 * config/s390/s390.opt: Add new options and and enum for the
2398 option values.
2400 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2402 * lra-constraints.c (match_reload): Unconditionally use
2403 gen_lowpart_SUBREG, rather than selecting between that
2404 and equivalent gen_rtx_SUBREG code.
2406 2018-02-08 Richard Biener <rguenther@suse.de>
2408 PR tree-optimization/84233
2409 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2410 changed flag instead of boguously re-using phi_inserted.
2412 2018-02-08 Martin Jambor <mjambor@suse.cz>
2414 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2415 static local variables.
2417 2018-02-08 Richard Biener <rguenther@suse.de>
2419 PR tree-optimization/84278
2420 * tree-vect-stmts.c (vectorizable_store): When looking for
2421 smaller vector types to perform grouped strided loads/stores
2422 make sure the mode is supported by the target.
2423 (vectorizable_load): Likewise.
2425 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2427 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2428 Increase LDP/STP opportunities by adding adjacent callee-saves.
2430 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2432 PR rtl-optimization/84068
2433 PR rtl-optimization/83459
2434 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2436 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2438 PR tree-optimization/84224
2439 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2440 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2441 non-zero arguments.
2443 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2445 PR target/84113
2446 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2447 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2448 count, remove one USE.
2450 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2452 * doc/install.texi (Configuration): Document the
2453 --with-long-double-format={ibm,ieee} PowerPC configuration
2454 options.
2456 PR target/84154
2457 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2458 Convert from define_expand to be define_insn_and_split. Rework
2459 float/double/_Float128 conversions to QI/HI/SImode to work with
2460 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2461 conversions to QI/HImode types did a store and then a load to
2462 truncate the value. For conversions to VSX registers, don't split
2463 the insn, instead emit the code directly. Use the code iterator
2464 any_fix to combine signed and unsigned conversions.
2465 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2466 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2467 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2468 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2469 (fix_<mode>di2_hw): Likewise.
2470 (fixuns_<mode>di2_hw): Likewise.
2471 (fix_<mode>si2_hw): Likewise.
2472 (fixuns_<mode>si2_hw): Likewise.
2473 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2474 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2475 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2476 fix<uns>_trunc<SFDF:mode>si2_p8.
2477 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2478 used.
2479 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2480 (fix<uns>_<mode>_mem): Likewise.
2481 (fctiw<u>z_<mode>_mem): Likewise.
2482 (fix<uns>_<mode>_mem): Likewise.
2483 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2484 the register allocator from doing a direct move to the GPRs to do
2485 a store, and instead use the ISA 3.0 store byte/half-word from
2486 vector register instruction. For IEEE 128-bit floating point,
2487 also optimize stores of 32-bit ints.
2488 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2490 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2492 * genextract.c (push_pathstr_operand): New function to support
2493 [a-zA-Z].
2494 (walk_rtx): Call push_pathstr_operand.
2495 (print_path): Support [a-zA-Z].
2497 2018-02-07 Richard Biener <rguenther@suse.de>
2499 PR tree-optimization/84037
2500 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2501 (cse_and_gimplify_to_preheader): Declare.
2502 (vect_get_place_in_interleaving_chain): Likewise.
2503 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2504 ivexpr_map.
2505 (_loop_vec_info::~_loop_vec_info): Delete it.
2506 (cse_and_gimplify_to_preheader): New function.
2507 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2508 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2509 (vectorizable_load): Likewise. For grouped stores always base
2510 the IV on the first element.
2511 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2512 condition before gimplifying.
2514 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2516 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2517 *DIV_EXPR and *MOD_EXPR.
2519 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2521 PR target/84248
2522 * config/i386/i386.c (ix86_option_override_internal): Mask out
2523 the CF_SET bit when checking -fcf-protection.
2525 2018-02-07 Tom de Vries <tom@codesourcery.com>
2527 PR libgomp/84217
2528 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2529 enough.
2531 2018-02-07 Richard Biener <rguenther@suse.de>
2533 PR tree-optimization/84204
2534 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2535 this place.
2537 PR tree-optimization/84205
2538 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2539 special-case isl_ast_op_zdiv_r.
2541 PR tree-optimization/84223
2542 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2543 Only add conditions from within the region.
2544 (gather_bbs::after_dom_children): Adjust.
2546 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2548 PR target/84209
2549 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2550 * config/avr/avr.md: Only post-reload split REG-REG moves if
2551 either register is GENERAL_REG_P.
2553 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2555 PR tree-optimization/84235
2556 * tree-ssa-scopedtables.c
2557 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2558 if the subtraction is performed in floating point type where NaNs are
2559 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2560 build 1. Formatting fix.
2562 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2564 PR target/84146
2565 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2566 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2567 and skip it regardless of bb boundaries. Use CALL_P macro,
2568 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2569 unnecessarily, formatting fix.
2571 2018-02-06 Michael Collison <michael.collison@arm.com>
2573 * config/arm/thumb2.md:
2574 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2575 (*thumb_mov_notscc): Ditto.
2577 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2579 PR target/84154
2580 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2581 unsigned_fix, not "s".
2583 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2585 * configure.ac (gcc_fn_eh_frame_ro): New function.
2586 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2587 correct .eh_frame permissions.
2588 * configure: Regenerate.
2590 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2592 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2593 irrelevant options.
2595 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2597 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2598 Display warning message for -mno-speculate-indirect-jumps.
2600 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2602 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2603 Undocumented.
2604 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2606 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2608 PR tree-optimization/84225
2609 * tree-eh.c (find_trapping_overflow): Only call
2610 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2612 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2614 PR target/84145
2615 * config/i386/i386.c: Reimplement the check of possible options
2616 -mibt/-mshstk conbination. Change error messages.
2617 * doc/invoke.texi: Fix a typo: remove extra '='.
2619 2018-02-06 Marek Polacek <polacek@redhat.com>
2621 PR tree-optimization/84228
2622 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2624 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2626 PR target/82641
2627 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2628 emitted arch directives.
2629 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2630 __ARM_FEATURE_COPROC before changing architectures.
2632 2018-02-06 Richard Biener <rguenther@suse.de>
2634 * config/i386/i386.c (print_reg): Fix typo.
2635 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2637 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2639 * configure: Regenerate.
2641 2018-02-05 Martin Sebor <msebor@redhat.com>
2643 PR tree-optimization/83369
2644 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2645 inlining context.
2647 2018-02-05 Martin Liska <mliska@suse.cz>
2649 * doc/invoke.texi: Cherry-pick upstream r323995.
2651 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2653 * ira.c (ira_init_register_move_cost): Adjust comment.
2655 2018-02-05 Martin Liska <mliska@suse.cz>
2657 PR gcov-profile/84137
2658 * doc/gcov.texi: Fix typo in documentation.
2660 2018-02-05 Martin Liska <mliska@suse.cz>
2662 PR gcov-profile/83879
2663 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2664 using dlopen functionality.
2666 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2668 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2669 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2670 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2671 _mm_maskz_range_round_ss): New intrinsics.
2672 (__builtin_ia32_rangesd128_round)
2673 (__builtin_ia32_rangess128_round): Remove.
2674 (__builtin_ia32_rangesd128_mask_round,
2675 __builtin_ia32_rangess128_mask_round): New builtins.
2676 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2677 __builtin_ia32_rangess128_round): Remove.
2678 (__builtin_ia32_rangesd128_mask_round,
2679 __builtin_ia32_rangess128_mask_round): New builtins.
2680 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2681 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2682 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2683 "<round_saeonly_constraint>")): Changed to ...
2684 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2685 "<round_saeonly_scalar_constraint>")): ... this.
2686 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2687 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2688 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2689 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2690 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2692 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2694 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2695 options.
2696 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2697 Remove all values except native, 8540 and 8548.
2699 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2701 * config/i386/i386.c (ix86_output_function_return): Pass
2702 INVALID_REGNUM, instead of -1, as invalid register number to
2703 indirect_thunk_name and output_indirect_thunk.
2705 2018-02-02 Julia Koval <julia.koval@intel.com>
2707 * config.gcc: Add -march=icelake.
2708 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2709 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2710 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2711 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2712 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2713 (processor_target_table): Add icelake.
2714 (ix86_option_override_internal): Handle new PTAs.
2715 (get_builtin_code_for_version): Handle icelake.
2716 (M_INTEL_COREI7_ICELAKE): New.
2717 (fold_builtin_cpu): Handle icelake.
2718 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2719 * doc/invoke.texi: Add -march=icelake.
2721 2018-02-02 Julia Koval <julia.koval@intel.com>
2723 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2724 to wide_int_bitmask.
2725 * wide-int-bitmask.h: New.
2727 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2729 PR target/84066
2730 * config/i386/i386.md: Replace Pmode with word_mode in
2731 builtin_setjmp_setup and builtin_longjmp to support x32.
2733 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2735 PR target/56010
2736 PR target/83743
2737 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2738 #include "opts.h".
2739 (rs6000_supported_cpu_names): New static variable.
2740 (linux_cpu_translation_table): Likewise.
2741 (elf_platform) <cpu>: Define new static variable and use it.
2742 Translate kernel AT_PLATFORM name to canonical name if needed.
2743 Error if platform name is unknown.
2745 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2747 PR target/84089
2748 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2750 2018-02-01 Jeff Law <law@redhat.com>
2752 PR target/84128
2753 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2754 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2755 the scratch if RELEASE_VIA_POP is false.
2756 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2757 If we have to save a temporary register, decrement SIZE appropriately.
2758 Pass new arguments to release_scratch_register_on_entry.
2759 (ix86_adjust_stack_and_probe): Likewise.
2760 (ix86_emit_probe_stack_range): Pass new arguments to
2761 release_scratch_register_on_entry.
2763 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2765 PR rtl-optimization/84157
2766 * combine.c (change_zero_ext): Use REG_P predicate in
2767 front of HARD_REGISTER_P predicate.
2769 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2771 * config/avr/avr.c (avr_option_override): Move disabling of
2772 -fdelete-null-pointer-checks to...
2773 * common/config/avr/avr-common.c (avr_option_optimization_table):
2774 ...here.
2776 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2778 PR tree-optimization/81635
2779 * tree-data-ref.c (split_constant_offset_1): For types that
2780 wrap on overflow, try to use range info to prove that wrapping
2781 cannot occur.
2783 2018-02-01 Renlin Li <renlin.li@arm.com>
2785 PR target/83370
2786 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2787 TAILCALL_ADDR_REGS.
2788 (aarch64_register_move_cost): Likewise.
2789 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2790 TAILCALL_ADDR_REGS.
2791 (REG_CLASS_NAMES): Likewise.
2792 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2793 TAILCALL_ADDR_REGS. Remove IP registers.
2794 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2796 2018-02-01 Richard Biener <rguenther@suse.de>
2798 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2799 for specifying RPO order and allow NULL for that.
2800 * domwalk.c (dom_walker::dom_walker): Likewise.
2801 (dom_walker::walk): Handle NULL RPO order.
2802 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2803 in RPO order.
2804 (rewrite_update_dom_walker): Likewise.
2805 (mark_def_dom_walker): Likewise.
2807 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2809 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2810 (aarch64_maybe_expand_sve_subreg_move): Declare.
2811 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2812 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2813 predicate.
2814 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2815 that are semantically a reverse operation.
2816 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2817 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2818 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2819 functions.
2820 (aarch64_can_change_mode_class): For big-endian, forbid changes
2821 between two SVE modes if they have different element sizes.
2823 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2825 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2826 the TImode handling for big-endian targets.
2828 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2830 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2831 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2832 not just bytes.
2833 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2834 Remove BSWAP handing for big-endian targets and use the form of
2835 LD1RQ appropariate for the mode.
2837 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2839 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2840 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2841 duplicated element.
2843 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2845 PR tearget/83845
2846 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2847 check for operands that need to go through aarch64_sve_reload_be.
2849 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2851 PR tree-optimization/81661
2852 PR tree-optimization/84117
2853 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2854 * tree-eh.c: Include gimplify.h.
2855 (find_trapping_overflow, replace_trapping_overflow,
2856 rewrite_to_non_trapping_overflow): New functions.
2857 * tree-vect-loop.c: Include tree-eh.h.
2858 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2859 * tree-data-ref.c: Include tree-eh.h.
2860 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2862 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2864 PR rtl-optimization/84123
2865 * combine.c (change_zero_ext): Check if hard register satisfies
2866 can_change_dest_mode before calling gen_lowpart_SUBREG.
2868 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2870 PR target/82444
2871 * ira.c (ira_init_register_move_cost): Remove assert.
2873 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2875 PR rtl-optimization/84071
2876 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2877 * doc/tm.texi: Regenerate.
2879 2018-01-31 Richard Biener <rguenther@suse.de>
2881 PR tree-optimization/84132
2882 * tree-data-ref.c (analyze_miv_subscript): Properly
2883 check whether evolution_function_is_affine_multivariate_p
2884 before calling gcd_of_steps_may_divide_p.
2886 2018-01-31 Julia Koval <julia.koval@intel.com>
2888 PR target/83618
2889 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2890 * config/i386/i386.md (rdpid_rex64) New.
2891 (rdpid): Make 32bit only.
2893 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2895 PR lto/84105
2896 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2897 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2899 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2901 Revert
2902 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2904 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2906 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2908 PR rtl-optimization/84071
2909 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2910 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2912 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2914 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2915 (arc_attribute_table): Add 'aux' attribute.
2916 (arc_in_small_data_p): Consider aux like variables.
2917 (arc_is_aux_reg_p): New function.
2918 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2919 (arc_get_aux_arg): New function.
2920 (prepare_move_operands): Handle aux-register access.
2921 (arc_handle_aux_attribute): New function.
2922 * doc/extend.texi (ARC Variable attributes): Add subsection.
2924 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2926 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2927 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2928 (arc_attribute_table): Add 'uncached' attribute.
2929 (arc_print_operand): Print '.di' flag for uncached memory
2930 accesses.
2931 (arc_in_small_data_p): Do not consider for small data the uncached
2932 types.
2933 (arc_is_uncached_mem_p): New function.
2934 * config/arc/predicates.md (compact_store_memory_operand): Check
2935 for uncached memory accesses.
2936 (nonvol_nonimm_operand): Likewise.
2937 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2939 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2941 PR c/84100
2942 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2943 falign-loops=): Add Optimization flag.
2945 2018-01-30 Jeff Law <law@redhat.com>
2947 PR target/84064
2948 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2949 INT_REGISTERS_SAVED. Check it prior to calling
2950 get_scratch_register_on_entry.
2951 (ix86_adjust_stack_and_probe): Similarly.
2952 (ix86_emit_probe_stack_range): Similarly.
2953 (ix86_expand_prologue): Corresponding changes.
2955 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2957 PR target/40411
2958 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2959 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2961 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2963 PR target/84112
2964 * lra-constraints.c (curr_insn_transform): Process AND in the
2965 address.
2967 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2969 PR rtl-optimization/83986
2970 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2971 dependence against last_pending_memory_flush in addition to
2972 pending_jump_insns.
2974 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2976 PR tree-optimization/81611
2977 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2978 copies.
2980 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2982 PR target/83758
2983 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2984 a reg rtx.
2986 2018-01-30 Richard Biener <rguenther@suse.de>
2987 Jakub Jelinek <jakub@redhat.com>
2989 PR tree-optimization/84111
2990 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2991 inner loops added during recursion, as they don't have up-to-date
2992 SSA form.
2994 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2996 PR ipa/81360
2997 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2998 (can_inline_edge_by_limits_p): ... here.
2999 (can_early_inline_edge_p, check_callers,
3000 update_caller_keys, update_callee_keys, recursive_inlining,
3001 add_new_edges_to_heap, speculation_useful_p,
3002 inline_small_functions,
3003 inline_small_functions, flatten_function,
3004 inline_to_all_callers_1): Update.
3006 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3008 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3009 zeros correctly.
3011 2018-01-30 Richard Biener <rguenther@suse.de>
3013 PR tree-optimization/83008
3014 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3015 invariant and constant vector uses in stmts when they need
3016 more than one stmt.
3018 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3020 PR bootstrap/84017
3021 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3022 * configure: Regenerate.
3024 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3026 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3027 pattern.
3028 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3029 Use gen_rtx_REG rather than gen_lowpart.
3031 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3033 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3034 rather than 0 when creating partial subregs.
3036 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3038 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3039 of usage.
3041 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3043 PR target/81550
3044 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3045 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3046 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3047 flags. This restores the settings used before the 2017-07-24.
3048 Turning off pre increment/decrement/modify allows IVOPTS to
3049 optimize DF/SF loops where the index is an int.
3051 2018-01-29 Richard Biener <rguenther@suse.de>
3052 Kelvin Nilsen <kelvin@gcc.gnu.org>
3054 PR bootstrap/80867
3055 * tree-vect-stmts.c (vectorizable_call): Don't call
3056 targetm.vectorize_builtin_md_vectorized_function if callee is
3057 NULL.
3059 2018-01-22 Carl Love <cel@us.ibm.com>
3061 * doc/extend.tex: Fix typo in second arg in
3062 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3064 2018-01-29 Richard Biener <rguenther@suse.de>
3066 PR tree-optimization/84086
3067 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3068 (flush_ssaname_freelist): When SSA names were released reset
3069 the SCEV hash table.
3071 2018-01-29 Richard Biener <rguenther@suse.de>
3073 PR tree-optimization/84057
3074 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3075 removed paths when removing edges.
3077 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3079 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3080 -mfunction-return=@var{choice}.
3082 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3084 PR diagnostic/84034
3085 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3086 Handle CR like TAB.
3087 (layout::print_source_line): Likewise.
3088 (test_get_line_width_without_trailing_whitespace): Add test cases.
3090 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3092 PR middle-end/84040
3093 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3094 debug insns.
3096 2018-01-26 Jim Wilson <jimw@sifive.com>
3098 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3100 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3101 specified.
3103 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3105 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3106 and CMP + SUB-immediate -> SUBS.
3108 2018-01-26 Martin Sebor <msebor@redhat.com>
3110 PR tree-optimization/83896
3111 * tree-ssa-strlen.c (get_string_len): Rename...
3112 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3113 Avoid assuming length is constant.
3114 (handle_char_store): Use HOST_WIDE_INT for string length.
3116 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3118 PR target/81763
3119 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3120 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3122 2018-01-26 Richard Biener <rguenther@suse.de>
3124 PR rtl-optimization/84003
3125 * dse.c (record_store): Only record redundant stores when
3126 the earlier store aliases at least all accesses the later one does.
3128 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3130 PR rtl-optimization/83985
3131 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3132 REG_CFA_RESTORE insns.
3133 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3134 REG_CFA_RESTORE insns here.
3136 PR c/83989
3137 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3138 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3140 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3142 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3143 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3144 (arc_init): Likewise.
3145 (arc_override_options): Likewise.
3146 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3147 value.
3148 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3149 support.
3150 * config/arc/arc.h (TARGET_DBNZ): Define.
3151 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3152 properly set the tune attribute.
3153 (dbnz): Use TARGET_DBNZ guard.
3154 * config/arc/arc.opt (mtune): Add core3 option.
3156 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3158 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3159 recognize new pic like addresses.
3160 (arc_delegitimize_address): Clean up.
3162 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3164 * config/arc/arc-arches.def: Option mrf16 valid for all
3165 architectures.
3166 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3167 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3168 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3169 * config/arc/arc-tables.opt: Regenerate.
3170 * config/arc/arc.c (arc_conditional_register_usage): Handle
3171 reduced register file case.
3172 (arc_file_start): Set must have build attributes.
3173 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3174 mrf16 option value.
3175 * config/arc/arc.opt (mrf16): Add new option.
3176 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3177 * config/arc/genmultilib.awk: Handle new mrf16 option.
3178 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3179 * config/arc/t-multilib: Regenerate.
3180 * doc/invoke.texi (ARC Options): Document mrf16 option.
3182 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3184 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3185 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3186 (arc_attribute_table): Add 'secure_call' attribute.
3187 (arc_print_operand): Print secure call operand.
3188 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3189 secure.
3190 (arc_is_secure_call_p): New function. * config/arc/arc.md
3191 (call_i): Add support for sjli instruction.
3192 (call_value_i): Likewise.
3193 * config/arc/constraints.md (Csc): New constraint.
3195 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3196 John Eric Martin <John.Martin@emmicro-us.com>
3198 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3199 * config/arc/arc.c (_arc_jli_section): New struct.
3200 (arc_jli_section): New type.
3201 (rc_jli_sections): New static variable.
3202 (arc_handle_jli_attribute): New function.
3203 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3204 (arc_file_end): New function.
3205 (TARGET_ASM_FILE_END): Define.
3206 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3207 (arc_add_jli_section): New function.
3208 (jli_call_scan): Likewise.
3209 (arc_reorg): Call jli_call_scan.
3210 (arc_output_addsi): Remove 'S' from printing asm operand.
3211 (arc_is_jli_call_p): New function.
3212 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3213 operand.
3214 (movhi_insn): Likewise.
3215 (movsi_insn): Likewise.
3216 (movsi_set_cc_insn): Likewise.
3217 (loadqi_update): Likewise.
3218 (load_zeroextendqisi_update): Likewise.
3219 (load_signextendqisi_update): Likewise.
3220 (loadhi_update): Likewise.
3221 (load_zeroextendhisi_update): Likewise.
3222 (load_signextendhisi_update): Likewise.
3223 (loadsi_update): Likewise.
3224 (loadsf_update): Likewise.
3225 (movsicc_insn): Likewise.
3226 (bset_insn): Likewise.
3227 (bxor_insn): Likewise.
3228 (bclr_insn): Likewise.
3229 (bmsk_insn): Likewise.
3230 (bicsi3_insn): Likewise.
3231 (cmpsi_cc_c_insn): Likewise.
3232 (movsi_ne): Likewise.
3233 (movsi_cond_exec): Likewise.
3234 (clrsbsi2): Likewise.
3235 (norm_f): Likewise.
3236 (normw): Likewise.
3237 (swap): Likewise.
3238 (divaw): Likewise.
3239 (flag): Likewise.
3240 (sr): Likewise.
3241 (kflag): Likewise.
3242 (ffs): Likewise.
3243 (ffs_f): Likewise.
3244 (fls): Likewise.
3245 (call_i): Remove 'S' asm letter, add jli instruction.
3246 (call_value_i): Likewise.
3247 * config/arc/arc.op (mjli-always): New option.
3248 * config/arc/constraints.md (Cji): New constraint.
3249 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3250 operand.
3251 (subsf3_fpx): Likewise.
3252 (mulsf3_fpx): Likewise.
3253 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3254 asm operand.
3255 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3256 function attrbutes.
3257 * doc/invoke.texi (ARC): Document mjli-always option.
3259 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3261 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3262 avoid addition with 0 and use incw and decw where possible.
3264 2018-01-26 Richard Biener <rguenther@suse.de>
3266 PR tree-optimization/81082
3267 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3268 association if it requires casting to unsigned.
3269 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3270 from fold_plusminus_mult_expr to catch important cases late when
3271 range info is available.
3273 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3275 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3276 * configure.ac (hidden_linkonce): New test.
3277 * configure: Regenerate.
3278 * config.in: Regenerate.
3280 2018-01-26 Julia Koval <julia.koval@intel.com>
3282 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3283 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3284 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3285 _mm_mask_bitshuffle_epi64_mask): Fix type.
3286 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3287 USI_FTYPE_V4DI_V4DI_USI): Remove.
3288 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3289 __builtin_ia32_vpshufbitqmb256_mask,
3290 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3291 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3292 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3294 2018-01-26 Alan Modra <amodra@gmail.com>
3296 PR target/84033
3297 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3298 UNSPEC_VBPERMQ. Sort other unspecs.
3300 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3302 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3304 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3306 PR middle-end/83055
3307 * predict.c (drop_profile): Do not push/pop cfun; update also
3308 node->count.
3309 (handle_missing_profiles): Fix logic looking for zero profiles.
3311 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3313 PR middle-end/83977
3314 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3315 on functions with #pragma omp declare simd or functions with simd
3316 attribute.
3317 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3318 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3319 Remove trailing \n from warning_at calls.
3321 2018-01-25 Tom de Vries <tom@codesourcery.com>
3323 PR target/84028
3324 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3325 for neutered workers.
3327 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3329 PR target/68467
3330 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3331 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3333 2018-01-24 Jeff Law <law@redhat.com>
3335 PR target/83994
3336 * i386.c (get_probe_interval): Move to earlier point.
3337 (ix86_compute_frame_layout): If -fstack-clash-protection and
3338 the frame is larger than the probe interval, then use pushes
3339 to save registers rather than reg->mem moves.
3340 (ix86_expand_prologue): Remove conditional for int_registers_saved
3341 assertion.
3343 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3345 PR target/84014
3346 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3347 min/max for never referenced object.
3349 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3351 PR middle-end/83977
3352 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3353 here.
3354 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3355 attributes from DECL_ATTRIBUTES (decl) without affecting
3356 DECL_ATTRIBUTES (current_function_decl).
3357 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3358 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3360 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3362 PR tree-optimization/83979
3363 * fold-const.c (fold_comparison): Use constant_boolean_node
3364 instead of boolean_{true,false}_node.
3366 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3368 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3369 with zero counts.
3371 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3373 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3374 Simplify the clause that sets the length attribute.
3375 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3376 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3377 clause that sets the length attribute.
3378 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3380 2018-01-24 Tom de Vries <tom@codesourcery.com>
3382 PR target/83589
3383 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3384 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3385 Add strict parameter.
3386 (prevent_branch_around_nothing): Insert dummy insn between branch to
3387 label and label with no ptx insn inbetween.
3388 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3390 2018-01-24 Tom de Vries <tom@codesourcery.com>
3392 PR target/81352
3393 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3394 for neutered threads in warp.
3395 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3397 2018-01-24 Richard Biener <rguenther@suse.de>
3399 PR tree-optimization/83176
3400 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3401 operands.
3403 2018-01-24 Richard Biener <rguenther@suse.de>
3405 PR tree-optimization/82819
3406 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3407 code generating pluses that are no-ops in the target precision.
3409 2018-01-24 Richard Biener <rguenther@suse.de>
3411 PR middle-end/84000
3412 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3414 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3416 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3417 to merge probabilities.
3418 * predict.c (probably_never_executed): Also mark as cold functions
3419 with global 0 profile and guessed local profile.
3420 * profile-count.c (profile_probability::combine_with_count): New
3421 member function.
3422 * profile-count.h (profile_probability::operator*,
3423 profile_probability::operator*=, profile_probability::operator/,
3424 profile_probability::operator/=): Reduce precision to adjusted
3425 and set value to guessed on contradictory divisions.
3426 (profile_probability::combine_with_freq): Remove.
3427 (profile_probability::combine_wiht_count): Declare.
3428 (profile_count::force_nonzero):: Set to adjusted.
3429 (profile_count::probability_in):: Set quality to adjusted.
3430 * tree-ssa-tail-merge.c (replace_block_by): Use
3431 combine_with_count.
3433 2018-01-23 Andrew Waterman <andrew@sifive.com>
3434 Jim Wilson <jimw@sifive.com>
3436 * config/riscv/riscv.c (riscv_stack_boundary): New.
3437 (riscv_option_override): Set riscv_stack_boundary. Handle
3438 riscv_preferred_stack_boundary_arg.
3439 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3440 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3441 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3442 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3443 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3444 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3446 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3448 PR target/83905
3449 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3450 of struct ix86_frame.
3451 (ix86_expand_epilogue): Likewise. Add a local variable for
3452 the reg_save_offset field in struct ix86_frame.
3454 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3456 PR tree-optimization/82604
3457 * tree-loop-distribution.c (enum partition_kind): New enum item
3458 PKIND_PARTIAL_MEMSET.
3459 (partition_builtin_p): Support above new enum item.
3460 (generate_code_for_partition): Ditto.
3461 (compute_access_range): Differentiate cases that equality can be
3462 proven at all loops, the innermost loops or no loops.
3463 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3464 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3465 (finalize_partitions, distribute_loop): Don't fuse partition of
3466 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3467 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3468 parloop is enabled.
3470 2018-01-23 Martin Liska <mliska@suse.cz>
3472 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3473 order to ignore the predictor.
3474 (PRED_POLYMORPHIC_CALL): Likewise.
3475 (PRED_RECURSIVE_CALL): Likewise.
3477 2018-01-23 Martin Liska <mliska@suse.cz>
3479 * tree-profile.c (tree_profiling): Print function header to
3480 aware reader which function we are working on.
3481 * value-prof.c (gimple_find_values_to_profile): Do not print
3482 not interesting value histograms.
3484 2018-01-23 Martin Liska <mliska@suse.cz>
3486 * profile-count.h (enum profile_quality): Add
3487 profile_uninitialized as the first value. Do not number values
3488 as they are zero based.
3489 (profile_count::verify): Update sanity check.
3490 (profile_probability::verify): Likewise.
3492 2018-01-23 Nathan Sidwell <nathan@acm.org>
3494 * doc/invoke.texi (ffor-scope): Deprecate.
3496 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3498 PR tree-optimization/83510
3499 * domwalk.c (set_all_edges_as_executable): New function.
3500 (dom_walker::dom_walker): Convert bool param
3501 "skip_unreachable_blocks" to enum reachability. Move setup of
3502 edge flags to set_all_edges_as_executable and only do it when
3503 reachability is REACHABLE_BLOCKS.
3504 * domwalk.h (enum dom_walker::reachability): New enum.
3505 (dom_walker::dom_walker): Convert bool param
3506 "skip_unreachable_blocks" to enum reachability.
3507 (set_all_edges_as_executable): New decl.
3508 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3509 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3510 "reachability".
3511 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3512 but converting true to REACHABLE_BLOCKS.
3513 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3514 * tree-vrp.c
3515 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3516 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3517 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3518 REACHABLE_BLOCKS.
3519 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3520 if check_all_array_refs will be called.
3522 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3524 * tree.c (selftest::test_location_wrappers): Add more test
3525 coverage.
3527 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3529 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3530 (selftest::test_bit_in_range): Likewise.
3532 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3534 PR testsuite/83888
3535 * doc/sourcebuild.texi (vect_float): Say that the selector
3536 only describes the situation when -funsafe-math-optimizations is on.
3537 (vect_float_strict): Document.
3539 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3541 PR tree-optimization/83965
3542 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3543 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3544 instead of checking only for a reduction.
3545 (vect_recog_widen_sum_pattern): Likewise.
3547 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3549 * predict.c (probably_never_executed): Only use precise profile info.
3550 (compute_function_frequency): Skip after inlining hack since we now
3551 have quality checking.
3553 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3555 * profile-count.h (profile_probability::very_unlikely,
3556 profile_probability::unlikely, profile_probability::even): Set
3557 precision to guessed.
3559 2018-01-23 Richard Biener <rguenther@suse.de>
3561 PR tree-optimization/83963
3562 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3563 Properly terminate dominator walk when crossing the exit edge not
3564 when visiting its source block.
3566 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3568 PR c++/83918
3569 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3570 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3572 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3574 PR tree-optimization/83957
3575 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3576 semicolon after for body surrounded by braces.
3578 PR tree-optimization/83081
3579 * profile-count.h (profile_probability::split): New method.
3580 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3581 Use profile_probability::split.
3582 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3583 when splitting a single conditional jump into 2.
3585 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3587 PR tree-optimization/69452
3588 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3589 decl.
3591 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3593 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3594 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3595 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3597 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3599 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3600 * config/rl78/rl78.md (movdi): New define_expand.
3601 * config/rl78/rl78.c (rl78_split_movdi): New function.
3603 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3605 PR target/83862
3606 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3607 no longer used.
3608 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3609 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3610 128-bit to produce an UNSPEC move to get the double word with the
3611 signbit and then a shift directly to do signbit.
3612 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3613 implementation with a new version that just does either a direct
3614 move or a regular move. Move memory interface to separate insns.
3615 Move insns so they are next to the expander.
3616 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3617 with signbit move. Split big and little endian case.
3618 (signbit<mode>2_dm_mem_le): Likewise.
3619 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3620 (signbit<mode>2_dm2): Likewise.
3622 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3624 * config/rl78/rl78.md (anddi3): New define_expand.
3626 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3628 * config/rl78/rl78.md (umindi3): New define_expand.
3630 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3632 * config/rl78/rl78.md (smindi3): New define_expand.
3634 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3636 * config/rl78/rl78.md (smaxdi3): New define_expand.
3638 2018-01-22 Carl Love <cel@us.ibm.com>
3640 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3641 LVX_V1TI): Add macro expansion.
3642 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3643 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3644 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3645 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3646 Change check to determine if the instruction is a byte reversing
3647 entry. Fix typo in comment.
3648 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3649 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3650 Add def_builtin calls for new builtins.
3651 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3652 Add define_insn expansion.
3654 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3656 * config/rl78/rl78.md (umaxdi3): New define_expand.
3658 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3660 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3661 for non-QImode registers.
3663 2018-01-22 Richard Biener <rguenther@suse.de>
3665 PR tree-optimization/83963
3666 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3667 including the loop exit block.
3668 (scop_detection::merge_sese): Likewise.
3669 (scop_detection::add_scop): Do it here instead.
3671 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3673 * doc/sourcebuild.texi (arm_softfloat): Document.
3675 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3677 PR gcc/77734
3678 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3679 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3680 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3682 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3683 David Edelsohn <dje.gcc@gmail.com>
3685 PR target/83946
3686 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3687 Change "crset eq" to "crset 2".
3688 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3689 (*call_indirect_aix<mode>_nospec): Likewise.
3690 (*call_value_indirect_aix<mode>_nospec): Likewise.
3691 (*call_indirect_elfv2<mode>_nospec): Likewise.
3692 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3693 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3694 change assembly output from . to $.
3695 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3696 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3697 (*tablejump<mode>_internal1_nospec): Likewise.
3699 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3701 PR target/80870
3702 * config/sh/sh_optimize_sett_clrt.cc:
3703 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3705 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3707 PR tree-optimization/83940
3708 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3709 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3710 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3711 use it to pass back the definition type.
3712 (vect_check_store_rhs): Likewise rhs_dt_out.
3713 (vect_build_gather_load_calls): Add a mask_dt argument and use
3714 it instead of a call to vect_is_simple_use.
3715 (vectorizable_store): Update calls to vect_check_load_store_mask
3716 and vect_check_store_rhs. Use the dt returned by the latter instead
3717 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3718 instead of calls to vect_is_simple_use. Pass the scalar rather
3719 than the vector operand to vect_is_simple_use when handling
3720 second and subsequent copies of an rhs value.
3721 (vectorizable_load): Update calls to vect_check_load_store_mask
3722 and vect_build_gather_load_calls. Use the cached mask_dt and
3723 gs_info.offset_dt instead of calls to vect_is_simple_use.
3725 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3727 PR middle-end/83945
3728 * tree-emutls.c: Include gimplify.h.
3729 (lower_emutls_2): New function.
3730 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3731 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3732 it before further processing.
3734 PR target/83930
3735 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3736 UINTVAL (trueop1) instead of INTVAL (op1).
3738 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3740 PR debug/81570
3741 PR debug/83728
3742 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3743 INCOMING_FRAME_SP_OFFSET if not defined.
3744 (scan_trace): Add ENTRY argument. If true and
3745 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3746 emit a note to adjust the CFA offset.
3747 (create_cfi_notes): Adjust scan_trace callers.
3748 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3749 INCOMING_FRAME_SP_OFFSET in the CIE.
3750 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3751 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3752 Likewise.
3753 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3754 * doc/tm.texi: Regenerated.
3756 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3758 PR rtl-optimization/83147
3759 * lra-constraints.c (remove_inheritance_pseudos): Use
3760 lra_substitute_pseudo_within_insn.
3762 2018-01-19 Tom de Vries <tom@codesourcery.com>
3763 Cesar Philippidis <cesar@codesourcery.com>
3765 PR target/83920
3766 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3768 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3770 PR target/83790
3771 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3772 spaces for function labels.
3774 2018-01-19 Martin Liska <mliska@suse.cz>
3776 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3777 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3778 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3779 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3780 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3781 (PRED_CONST_RETURN): Change from 69 to 65.
3782 (PRED_NULL_RETURN): Change from 91 to 71.
3783 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3784 (PRED_LOOP_GUARD): Change from 66 to 73.
3786 2018-01-19 Martin Liska <mliska@suse.cz>
3788 * predict.c (predict_insn_def): Add new assert.
3789 (struct branch_predictor): Change type to signed integer.
3790 (test_prediction_value_range): Amend test to cover
3791 PROB_UNINITIALIZED.
3792 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3793 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3794 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3795 (PRED_LOOP_IV_COMPARE): Likewise.
3796 * predict.h (PROB_UNINITIALIZED): Define new constant.
3798 2018-01-19 Martin Liska <mliska@suse.cz>
3800 * predict.c (dump_prediction): Add new format for
3801 analyze_brprob.py script which is enabled with -details
3802 suboption.
3803 * profile-count.h (precise_p): New function.
3805 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3807 PR tree-optimization/83922
3808 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3809 there are no statements that need masking.
3810 (vect_active_double_reduction_p): New function.
3811 (vect_analyze_loop_operations): Use it when handling phis that
3812 are not in the loop header.
3814 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3816 PR tree-optimization/83914
3817 * tree-vect-loop.c (vectorizable_induction): Don't convert
3818 init_expr or apply the peeling adjustment for inductions
3819 that are nested within the vectorized loop.
3821 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3823 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3824 instead of NEG.
3826 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3828 PR sanitizer/81715
3829 PR testsuite/83882
3830 * function.h (gimplify_parameters): Add gimple_seq * argument.
3831 * function.c: Include gimple.h and options.h.
3832 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3833 for the added local temporaries if needed.
3834 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3835 if there are any parameter cleanups, wrap whole body into a
3836 try/finally with the cleanups.
3838 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3840 PR target/82964
3841 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3842 Use GET_MODE_CLASS for scalar floating point.
3844 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3846 PR ipa/82256
3847 patch by PaX Team
3848 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3849 Fix call of call_cgraph_insertion_hooks.
3851 2018-01-18 Martin Sebor <msebor@redhat.com>
3853 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3855 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3857 PR ipa/83619
3858 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3859 frequencies.
3861 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3863 PR other/70268
3864 * common.opt: (-ffile-prefix-map): New option.
3865 * opts.c (common_handle_option): Defer it.
3866 * opts-global.c (handle_common_deferred_options): Handle it.
3867 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3868 * file-prefix-map.h: New file.
3869 (remap_debug_filename, add_debug_prefix_map): ...here.
3870 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3871 * final.c (debug_prefix_map, add_debug_prefix_map
3872 remap_debug_filename): Move to...
3873 * file-prefix-map.c: New file.
3874 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3875 generalize, get rid of alloca(), use strrchr() instead of strchr().
3876 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3877 Implement in terms of add_prefix_map().
3878 (remap_macro_filename, remap_debug_filename): Implement in term of
3879 remap_filename().
3880 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3881 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3882 * dbxout.c: Include file-prefix-map.h.
3883 * varasm.c: Likewise.
3884 * vmsdbgout.c: Likewise.
3885 * xcoffout.c: Likewise.
3886 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3887 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3888 * doc/invoke.texi (-ffile-prefix-map): Document.
3889 (-fdebug-prefix-map): Update description.
3891 2018-01-18 Martin Liska <mliska@suse.cz>
3893 * config/i386/i386.c (indirect_thunk_name): Document that also
3894 lfence is emitted.
3895 (output_indirect_thunk): Document why both instructions
3896 (pause and lfence) are generated.
3898 2018-01-18 Richard Biener <rguenther@suse.de>
3900 PR tree-optimization/83887
3901 * graphite-scop-detection.c
3902 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3903 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3904 (scop_detection::merge_sese): Re-implement with a flood-fill
3905 algorithm that properly finds a SESE region if it exists.
3907 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3909 PR c/61240
3910 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3911 pointer_diff optimizations use view_convert instead of convert.
3913 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3915 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3916 Generate different code for -mno-speculate-indirect-jumps.
3917 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3918 (*call_indirect_aix<mode>): Disable for
3919 -mno-speculate-indirect-jumps.
3920 (*call_indirect_aix<mode>_nospec): New define_insn.
3921 (*call_value_indirect_aix<mode>): Disable for
3922 -mno-speculate-indirect-jumps.
3923 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3924 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3925 -mno-speculate-indirect-jumps.
3926 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3928 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3930 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3931 long double type, set the flags for noting the default long double
3932 type, even if we don't pass or return a long double type.
3934 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3936 PR ipa/83051
3937 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3938 failure.
3940 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3942 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3943 support for merge[hl].
3944 (fold_mergehl_helper): New helper function.
3945 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3946 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3947 (altivec_vmrglw_direct): Add xxmrglw insn.
3949 2018-01-17 Andrew Waterman <andrew@sifive.com>
3951 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3952 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3954 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3956 PR lto/83121
3957 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3958 call the lto_location_cache before reading the
3959 DECL_SOURCE_LOCATION of the types.
3961 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3962 Richard Sandiford <richard.sandiford@linaro.org>
3964 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3965 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3966 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3967 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3968 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3969 Add declaration.
3970 * config/aarch64/constraints.md (aarch64_movti_operand):
3971 Limit immediates.
3972 * config/aarch64/predicates.md (Uti): Add new constraint.
3974 2018-01-17 Carl Love <cel@us.ibm.com>
3976 * config/rs6000/vsx.md (define_expand xl_len_r,
3977 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3978 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3979 lxvll.
3980 (define_expand, define_insn): Move the shift left from the
3981 define_insn to the define_expand for lxvl and stxvl instructions.
3982 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3983 and XL_LEN_R definitions to PURE.
3985 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3987 * config/i386/i386.c (indirect_thunk_name): Declare regno
3988 as unsigned int. Compare regno with INVALID_REGNUM.
3989 (output_indirect_thunk): Ditto.
3990 (output_indirect_thunk_function): Ditto.
3991 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3992 in the call to output_indirect_thunk_function.
3994 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3996 PR middle-end/83884
3997 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3998 rather than the size of inner_type to determine the stack slot size
3999 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4001 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4003 PR target/83546
4004 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4005 to PTA_SILVERMONT.
4007 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4009 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4010 endian Linux systems to optionally enable multilibs for selecting
4011 the long double type if the user configured an explicit type.
4012 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4013 have no long double multilibs if not defined.
4014 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4015 warn if the user used -mabi={ieee,ibm}longdouble and we built
4016 multilibs for long double.
4017 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4018 appropriate multilib option.
4019 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4020 multilib options.
4021 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4022 for building long double multilibs.
4023 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4025 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4027 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4028 copies.
4030 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4031 64 bits.
4032 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4033 128 bits.
4035 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4036 variables.
4038 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4039 return value.
4041 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4043 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4044 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4046 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4048 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4049 different rtl trees depending on TARGET_64BIT.
4050 (rs6000_gen_lvx): Likewise.
4052 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4054 * config/visium/visium.md (nop): Tweak comment.
4055 (hazard_nop): Likewise.
4057 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4059 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4060 -mspeculate-indirect-jumps.
4061 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4062 for -mno-speculate-indirect-jumps.
4063 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4064 (*call_value_indirect_elfv2<mode>): Disable for
4065 -mno-speculate-indirect-jumps.
4066 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4067 (indirect_jump): Emit different RTL for
4068 -mno-speculate-indirect-jumps.
4069 (*indirect_jump<mode>): Disable for
4070 -mno-speculate-indirect-jumps.
4071 (*indirect_jump<mode>_nospec): New define_insn.
4072 (tablejump): Emit different RTL for
4073 -mno-speculate-indirect-jumps.
4074 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4075 (tablejumpsi_nospec): New define_expand.
4076 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4077 (tablejumpdi_nospec): New define_expand.
4078 (*tablejump<mode>_internal1): Disable for
4079 -mno-speculate-indirect-jumps.
4080 (*tablejump<mode>_internal1_nospec): New define_insn.
4081 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4082 option.
4084 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4086 * caller-save.c (insert_save): Drop unnecessary parameter. All
4087 callers updated.
4089 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4090 Richard Biener <rguenth@suse.de>
4092 PR libgomp/83590
4093 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4094 return early, inline manually is_gimple_sizepos. Make sure if we
4095 call gimplify_expr we don't end up with a gimple constant.
4096 * tree.c (variably_modified_type_p): Don't return true for
4097 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4098 * gimplify.h (is_gimple_sizepos): Remove.
4100 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4102 PR tree-optimization/83857
4103 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4104 vectorizable_live_operation for pure SLP statements.
4105 (vectorizable_live_operation): Handle PHIs.
4107 2018-01-16 Richard Biener <rguenther@suse.de>
4109 PR tree-optimization/83867
4110 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4111 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4113 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4115 PR c/83844
4116 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4117 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4118 If off is not INTEGER_CST, issue a may not be aligned warning
4119 rather than isn't aligned. Use isn%'t rather than isn't.
4120 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4121 into MULT_EXPR.
4122 <case MULT_EXPR>: Improve the case when bottom and one of the
4123 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4124 operand, in that case check if the other operand is multiple of
4125 bottom divided by the INTEGER_CST operand.
4127 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4129 PR target/83858
4130 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4131 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4132 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4133 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4134 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4135 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4136 (pa_function_arg_size): New function.
4138 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4140 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4141 in a separate statement.
4143 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4145 PR tree-optimization/83847
4146 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4147 group gathers and scatters.
4149 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4151 PR rtl-optimization/86620
4152 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4154 PR rtl-optimization/83213
4155 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4156 to last if both are JUMP_INSNs.
4158 PR tree-optimization/83843
4159 * gimple-ssa-store-merging.c
4160 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4161 store_immediate_info for bswap/nop orig_stores.
4163 2018-01-15 Andrew Waterman <andrew@sifive.com>
4165 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4166 !TARGET_MUL.
4167 <UDIV>: Increase cost if !TARGET_DIV.
4169 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4171 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4172 (define_attr "cr_logical_3op"): New.
4173 (cceq_ior_compare): Adjust.
4174 (cceq_ior_compare_complement): Adjust.
4175 (*cceq_rev_compare): Adjust.
4176 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4177 (is_cracked_insn): Adjust.
4178 (insn_must_be_first_in_group): Adjust.
4179 * config/rs6000/40x.md: Adjust.
4180 * config/rs6000/440.md: Adjust.
4181 * config/rs6000/476.md: Adjust.
4182 * config/rs6000/601.md: Adjust.
4183 * config/rs6000/603.md: Adjust.
4184 * config/rs6000/6xx.md: Adjust.
4185 * config/rs6000/7450.md: Adjust.
4186 * config/rs6000/7xx.md: Adjust.
4187 * config/rs6000/8540.md: Adjust.
4188 * config/rs6000/cell.md: Adjust.
4189 * config/rs6000/e300c2c3.md: Adjust.
4190 * config/rs6000/e500mc.md: Adjust.
4191 * config/rs6000/e500mc64.md: Adjust.
4192 * config/rs6000/e5500.md: Adjust.
4193 * config/rs6000/e6500.md: Adjust.
4194 * config/rs6000/mpc.md: Adjust.
4195 * config/rs6000/power4.md: Adjust.
4196 * config/rs6000/power5.md: Adjust.
4197 * config/rs6000/power6.md: Adjust.
4198 * config/rs6000/power7.md: Adjust.
4199 * config/rs6000/power8.md: Adjust.
4200 * config/rs6000/power9.md: Adjust.
4201 * config/rs6000/rs64.md: Adjust.
4202 * config/rs6000/titan.md: Adjust.
4204 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4206 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4207 ix86_indirect_branch_register logic.
4209 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4211 * config/i386/constraints.md (Bs): Update
4212 ix86_indirect_branch_register check. Don't check
4213 ix86_indirect_branch_register with GOT_memory_operand.
4214 (Bw): Likewise.
4215 * config/i386/predicates.md (GOT_memory_operand): Don't check
4216 ix86_indirect_branch_register here.
4217 (GOT32_symbol_operand): Likewise.
4219 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4221 * config/i386/predicates.md (constant_call_address_operand):
4222 Rewrite ix86_indirect_branch_register logic.
4223 (sibcall_insn_operand): Likewise.
4225 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4227 * config/i386/constraints.md (Bs): Replace
4228 ix86_indirect_branch_thunk_register with
4229 ix86_indirect_branch_register.
4230 (Bw): Likewise.
4231 * config/i386/i386.md (indirect_jump): Likewise.
4232 (tablejump): Likewise.
4233 (*sibcall_memory): Likewise.
4234 (*sibcall_value_memory): Likewise.
4235 Peepholes of indirect call and jump via memory: Likewise.
4236 * config/i386/i386.opt: Likewise.
4237 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4238 (GOT_memory_operand): Likewise.
4239 (call_insn_operand): Likewise.
4240 (sibcall_insn_operand): Likewise.
4241 (GOT32_symbol_operand): Likewise.
4243 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4245 PR middle-end/83837
4246 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4247 type rather than type addr's type points to.
4248 (expand_omp_atomic_mutex): Likewise.
4249 (expand_omp_atomic): Likewise.
4251 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4253 PR target/83839
4254 * config/i386/i386.c (output_indirect_thunk_function): Use
4255 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4256 for __x86_return_thunk.
4258 2018-01-15 Richard Biener <rguenther@suse.de>
4260 PR middle-end/83850
4261 * expmed.c (extract_bit_field_1): Fix typo.
4263 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4265 PR target/83687
4266 * config/arm/iterators.md (VF): New mode iterator.
4267 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4268 Remove integer-related logic from pattern.
4269 (neon_vabd<mode>_3): Likewise.
4271 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4273 PR middle-end/82694
4274 * common.opt (fstrict-overflow): No longer an alias.
4275 (fwrapv-pointer): New option.
4276 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4277 also for pointer types based on flag_wrapv_pointer.
4278 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4279 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4280 opts->x_flag_wrapv got set.
4281 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4282 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4283 POINTER_TYPE_OVERFLOW_UNDEFINED.
4284 * match.pd: Likewise in address comparison pattern.
4285 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4287 2018-01-15 Richard Biener <rguenther@suse.de>
4289 PR lto/83804
4290 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4291 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4292 Reset type names to their identifier if their TYPE_DECL doesn't
4293 have linkage (and thus is used for ODR and devirt).
4294 (save_debug_info_for_decl): Remove.
4295 (save_debug_info_for_type): Likewise.
4296 (add_tree_to_fld_list): Adjust.
4297 * tree-pretty-print.c (dump_generic_node): Make dumping of
4298 type names more robust.
4300 2018-01-15 Richard Biener <rguenther@suse.de>
4302 * BASE-VER: Bump to 8.0.1.
4304 2018-01-14 Martin Sebor <msebor@redhat.com>
4306 PR other/83508
4307 * builtins.c (check_access): Avoid warning when the no-warning bit
4308 is set.
4310 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4312 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4313 * ira-color (allocno_hard_regs_compare): Likewise.
4315 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4317 PR target/83013
4318 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4319 Use .pushsection/.popsection.
4321 2018-01-14 Martin Sebor <msebor@redhat.com>
4323 PR c++/81327
4324 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4326 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4328 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4329 entry from extra_headers.
4330 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4331 extra_headers, make the list bitwise identical to the i?86-*-* one.
4333 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4335 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4336 -mcmodel=large with -mindirect-branch=thunk,
4337 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4338 -mfunction-return=thunk-extern.
4339 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4340 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4341 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4343 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4345 * config/i386/i386.c (print_reg): Print the name of the full
4346 integer register without '%'.
4347 (ix86_print_operand): Handle 'V'.
4348 * doc/extend.texi: Document 'V' modifier.
4350 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4352 * config/i386/constraints.md (Bs): Disallow memory operand for
4353 -mindirect-branch-register.
4354 (Bw): Likewise.
4355 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4356 (GOT_memory_operand): Likewise.
4357 (call_insn_operand): Likewise.
4358 (sibcall_insn_operand): Likewise.
4359 (GOT32_symbol_operand): Likewise.
4360 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4361 for -mindirect-branch-register.
4362 (tablejump): Likewise.
4363 (*sibcall_memory): Likewise.
4364 (*sibcall_value_memory): Likewise.
4365 Disallow peepholes of indirect call and jump via memory for
4366 -mindirect-branch-register.
4367 (*call_pop): Replace m with Bw.
4368 (*call_value_pop): Likewise.
4369 (*sibcall_pop_memory): Replace m with Bs.
4370 * config/i386/i386.opt (mindirect-branch-register): New option.
4371 * doc/invoke.texi: Document -mindirect-branch-register option.
4373 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4375 * config/i386/i386-protos.h (ix86_output_function_return): New.
4376 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4377 set function_return_type.
4378 (indirect_thunk_name): Add ret_p to indicate thunk for function
4379 return.
4380 (output_indirect_thunk_function): Pass false to
4381 indirect_thunk_name.
4382 (ix86_output_indirect_branch_via_reg): Likewise.
4383 (ix86_output_indirect_branch_via_push): Likewise.
4384 (output_indirect_thunk_function): Create alias for function
4385 return thunk if regno < 0.
4386 (ix86_output_function_return): New function.
4387 (ix86_handle_fndecl_attribute): Handle function_return.
4388 (ix86_attribute_table): Add function_return.
4389 * config/i386/i386.h (machine_function): Add
4390 function_return_type.
4391 * config/i386/i386.md (simple_return_internal): Use
4392 ix86_output_function_return.
4393 (simple_return_internal_long): Likewise.
4394 * config/i386/i386.opt (mfunction-return=): New option.
4395 (indirect_branch): Mention -mfunction-return=.
4396 * doc/extend.texi: Document function_return function attribute.
4397 * doc/invoke.texi: Document -mfunction-return= option.
4399 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4401 * config/i386/i386-opts.h (indirect_branch): New.
4402 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4403 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4404 with local indirect jump when converting indirect call and jump.
4405 (ix86_set_indirect_branch_type): New.
4406 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4407 (indirectlabelno): New.
4408 (indirect_thunk_needed): Likewise.
4409 (indirect_thunk_bnd_needed): Likewise.
4410 (indirect_thunks_used): Likewise.
4411 (indirect_thunks_bnd_used): Likewise.
4412 (INDIRECT_LABEL): Likewise.
4413 (indirect_thunk_name): Likewise.
4414 (output_indirect_thunk): Likewise.
4415 (output_indirect_thunk_function): Likewise.
4416 (ix86_output_indirect_branch_via_reg): Likewise.
4417 (ix86_output_indirect_branch_via_push): Likewise.
4418 (ix86_output_indirect_branch): Likewise.
4419 (ix86_output_indirect_jmp): Likewise.
4420 (ix86_code_end): Call output_indirect_thunk_function if needed.
4421 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4422 needed.
4423 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4424 (ix86_attribute_table): Add indirect_branch.
4425 * config/i386/i386.h (machine_function): Add indirect_branch_type
4426 and has_local_indirect_jump.
4427 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4428 to true.
4429 (tablejump): Likewise.
4430 (*indirect_jump): Use ix86_output_indirect_jmp.
4431 (*tablejump_1): Likewise.
4432 (simple_return_indirect_internal): Likewise.
4433 * config/i386/i386.opt (mindirect-branch=): New option.
4434 (indirect_branch): New.
4435 (keep): Likewise.
4436 (thunk): Likewise.
4437 (thunk-inline): Likewise.
4438 (thunk-extern): Likewise.
4439 * doc/extend.texi: Document indirect_branch function attribute.
4440 * doc/invoke.texi: Document -mindirect-branch= option.
4442 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4444 PR ipa/83051
4445 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4447 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4449 * ipa-inline.c (want_inline_small_function_p): Return false if
4450 inlining has already failed with CIF_FINAL_ERROR.
4451 (update_caller_keys): Call want_inline_small_function_p before
4452 can_inline_edge_p.
4453 (update_callee_keys): Likewise.
4455 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4457 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4458 New function.
4459 (rs6000_quadword_masked_address_p): Likewise.
4460 (quad_aligned_load_p): Likewise.
4461 (quad_aligned_store_p): Likewise.
4462 (const_load_sequence_p): Add comment to describe the outer-most loop.
4463 (mimic_memory_attributes_and_flags): New function.
4464 (rs6000_gen_stvx): Likewise.
4465 (replace_swapped_aligned_store): Likewise.
4466 (rs6000_gen_lvx): Likewise.
4467 (replace_swapped_aligned_load): Likewise.
4468 (replace_swapped_load_constant): Capitalize argument name in
4469 comment describing this function.
4470 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4471 and stores that access quad-word aligned addresses and replace
4472 with stvx or lvx instructions when appropriate.
4473 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4474 New function prototype.
4475 (rs6000_quadword_masked_address_p): Likewise.
4476 (rs6000_gen_lvx): Likewise.
4477 (rs6000_gen_stvx): Likewise.
4478 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4479 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4480 when memory address is aligned.
4481 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4482 this split to select lvx instruction when memory address is aligned.
4483 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4484 instruction when memory address is aligned.
4485 (*vsx_le_perm_load_v16qi): Likewise.
4486 (four unnamed splitters): Modify to select the stvx instruction
4487 when memory is aligned.
4489 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4491 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4492 which appears in the queue multiple times.
4494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4495 Alan Hayward <alan.hayward@arm.com>
4496 David Sherwood <david.sherwood@arm.com>
4498 * tree-vectorizer.h (vec_lower_bound): New structure.
4499 (_loop_vec_info): Add check_nonzero and lower_bounds.
4500 (LOOP_VINFO_CHECK_NONZERO): New macro.
4501 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4502 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4503 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4504 fields. Make seg_len the distance travelled, not including the
4505 access size.
4506 (dr_direction_indicator): Declare.
4507 (dr_zero_step_indicator): Likewise.
4508 (dr_known_forward_stride_p): Likewise.
4509 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4510 tree-ssanames.h.
4511 (runtime_alias_check_p): Allow runtime alias checks with
4512 variable strides.
4513 (operator ==): Compare access_size and align.
4514 (prune_runtime_alias_test_list): Rework for new distinction between
4515 the access_size and seg_len.
4516 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4517 segment lengths.
4518 (get_segment_min_max): New function.
4519 (create_intersect_range_checks): Use it.
4520 (dr_step_indicator): New function.
4521 (dr_direction_indicator): Likewise.
4522 (dr_zero_step_indicator): Likewise.
4523 (dr_known_forward_stride_p): Likewise.
4524 * tree-loop-distribution.c (data_ref_segment_size): Return
4525 DR_STEP * (niters - 1).
4526 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4527 constructor.
4528 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4529 (vect_preserves_scalar_order_p): New function, split out from...
4530 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4531 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4532 (vect_vfa_access_size): New function.
4533 (vect_vfa_align): Likewise.
4534 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4535 (dump_lower_bound): New function.
4536 (vect_check_lower_bound): Likewise.
4537 (vect_small_gap_p): Likewise.
4538 (vectorizable_with_step_bound_p): Likewise.
4539 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4540 depencies if the vectorization factor is 1. Convert the checks
4541 for nonzero steps into checks on the bounds of DR_STEP. Try using
4542 a bunds check for variable steps if the minimum required step is
4543 relatively small. Update calls to the dr_with_seg_len
4544 constructor and to vect_compile_time_alias.
4545 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4546 function.
4547 (vect_loop_versioning): Call it.
4548 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4549 when retrying.
4550 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4553 Alan Hayward <alan.hayward@arm.com>
4554 David Sherwood <david.sherwood@arm.com>
4556 * doc/sourcebuild.texi (vect_scatter_store): Document.
4557 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4558 optabs.
4559 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4560 Document.
4561 * genopinit.c (main): Add supports_vec_scatter_store and
4562 supports_vec_scatter_store_cached to target_optabs.
4563 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4564 IFN_MASK_SCATTER_STORE.
4565 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4566 functions.
4567 * internal-fn.h (internal_store_fn_p): Declare.
4568 (internal_fn_stored_value_index): Likewise.
4569 * internal-fn.c (scatter_store_direct): New macro.
4570 (expand_scatter_store_optab_fn): New function.
4571 (direct_scatter_store_optab_supported_p): New macro.
4572 (internal_store_fn_p): New function.
4573 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4574 IFN_MASK_SCATTER_STORE.
4575 (internal_fn_mask_index): Likewise.
4576 (internal_fn_stored_value_index): New function.
4577 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4578 for scatter stores.
4579 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4580 * optabs-query.c (supports_vec_scatter_store_p): New function.
4581 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4582 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4583 true for scatter stores.
4584 (vect_gather_scatter_fn_p): Handle scatter stores too.
4585 (vect_check_gather_scatter): Consider using scatter stores if
4586 supports_vec_scatter_store_p.
4587 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4588 scatter stores too.
4589 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4590 internal_fn_stored_value_index.
4591 (check_load_store_masking): Handle scatter stores too.
4592 (vect_get_store_rhs): Make public.
4593 (vectorizable_call): Use internal_store_fn_p.
4594 (vectorizable_store): Handle scatter store internal functions.
4595 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4596 when deciding whether the end of the group has been reached.
4597 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4598 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4599 (mask_scatter_store<mode>): New insns.
4601 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4602 Alan Hayward <alan.hayward@arm.com>
4603 David Sherwood <david.sherwood@arm.com>
4605 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4606 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4607 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4608 function.
4609 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4610 Use vect_truncate_gather_scatter_offset if we can't treat the
4611 operation as a normal gather load or scatter store.
4612 (get_group_load_store_type): Take the gather_scatter_info
4613 as argument. Try using a gather load or scatter store for
4614 single-element groups.
4615 (get_load_store_type): Update calls to get_group_load_store_type
4616 and vect_use_strided_gather_scatters_p.
4618 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4619 Alan Hayward <alan.hayward@arm.com>
4620 David Sherwood <david.sherwood@arm.com>
4622 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4623 optional tree argument.
4624 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4625 null target hooks.
4626 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4627 but continue to use the current value as a fallback.
4628 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4629 to compare the updates.
4630 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4631 (get_load_store_type): Use it when handling a strided access.
4632 (vect_get_strided_load_store_ops): New function.
4633 (vect_get_data_ptr_increment): Likewise.
4634 (vectorizable_load): Handle strided gather loads. Always pass
4635 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4637 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4638 Alan Hayward <alan.hayward@arm.com>
4639 David Sherwood <david.sherwood@arm.com>
4641 * doc/md.texi (gather_load@var{m}): Document.
4642 (mask_gather_load@var{m}): Likewise.
4643 * genopinit.c (main): Add supports_vec_gather_load and
4644 supports_vec_gather_load_cached to target_optabs.
4645 * optabs-tree.c (init_tree_optimization_optabs): Use
4646 ggc_cleared_alloc to allocate target_optabs.
4647 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4648 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4649 functions.
4650 * internal-fn.h (internal_load_fn_p): Declare.
4651 (internal_gather_scatter_fn_p): Likewise.
4652 (internal_fn_mask_index): Likewise.
4653 (internal_gather_scatter_fn_supported_p): Likewise.
4654 * internal-fn.c (gather_load_direct): New macro.
4655 (expand_gather_load_optab_fn): New function.
4656 (direct_gather_load_optab_supported_p): New macro.
4657 (direct_internal_fn_optab): New function.
4658 (internal_load_fn_p): Likewise.
4659 (internal_gather_scatter_fn_p): Likewise.
4660 (internal_fn_mask_index): Likewise.
4661 (internal_gather_scatter_fn_supported_p): Likewise.
4662 * optabs-query.c (supports_at_least_one_mode_p): New function.
4663 (supports_vec_gather_load_p): Likewise.
4664 * optabs-query.h (supports_vec_gather_load_p): Declare.
4665 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4666 and memory_type field.
4667 (NUM_PATTERNS): Bump to 15.
4668 * tree-vect-data-refs.c: Include internal-fn.h.
4669 (vect_gather_scatter_fn_p): New function.
4670 (vect_describe_gather_scatter_call): Likewise.
4671 (vect_check_gather_scatter): Try using internal functions for
4672 gather loads. Recognize existing calls to a gather load function.
4673 (vect_analyze_data_refs): Consider using gather loads if
4674 supports_vec_gather_load_p.
4675 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4676 (vect_get_gather_scatter_offset_type): Likewise.
4677 (vect_convert_mask_for_vectype): Likewise.
4678 (vect_add_conversion_to_patterm): Likewise.
4679 (vect_try_gather_scatter_pattern): Likewise.
4680 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4681 (vect_vect_recog_func_ptrs): Add it.
4682 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4683 internal_fn_mask_index and internal_gather_scatter_fn_p.
4684 (check_load_store_masking): Take the gather_scatter_info as an
4685 argument and handle gather loads.
4686 (vect_get_gather_scatter_ops): New function.
4687 (vectorizable_call): Check internal_load_fn_p.
4688 (vectorizable_load): Likewise. Handle gather load internal
4689 functions.
4690 (vectorizable_store): Update call to check_load_store_masking.
4691 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4692 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4693 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4694 (aarch64_gather_scale_operand_d): New predicates.
4695 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4696 (mask_gather_load<mode>): New insns.
4698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4699 Alan Hayward <alan.hayward@arm.com>
4700 David Sherwood <david.sherwood@arm.com>
4702 * optabs.def (fold_left_plus_optab): New optab.
4703 * doc/md.texi (fold_left_plus_@var{m}): Document.
4704 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4705 * internal-fn.c (fold_left_direct): Define.
4706 (expand_fold_left_optab_fn): Likewise.
4707 (direct_fold_left_optab_supported_p): Likewise.
4708 * fold-const-call.c (fold_const_fold_left): New function.
4709 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4710 * tree-parloops.c (valid_reduction_p): New function.
4711 (gather_scalar_reductions): Use it.
4712 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4713 (vect_finish_replace_stmt): Declare.
4714 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4715 (needs_fold_left_reduction_p): New function, split out from...
4716 (vect_is_simple_reduction): ...here. Accept reductions that
4717 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4718 (vect_force_simple_reduction): Also store the reduction type in
4719 the assignment's STMT_VINFO_REDUC_TYPE.
4720 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4721 (merge_with_identity): New function.
4722 (vect_expand_fold_left): Likewise.
4723 (vectorize_fold_left_reduction): Likewise.
4724 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4725 scalar phi in place for it. Check for target support and reject
4726 cases that would reassociate the operation. Defer the transform
4727 phase to vectorize_fold_left_reduction.
4728 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4729 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4730 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4734 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4735 call to ifc_temp_var.
4737 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4738 Alan Hayward <alan.hayward@arm.com>
4739 David Sherwood <david.sherwood@arm.com>
4741 * target.def (legitimize_address_displacement): Take the original
4742 offset as a poly_int.
4743 * targhooks.h (default_legitimize_address_displacement): Update
4744 accordingly.
4745 * targhooks.c (default_legitimize_address_displacement): Likewise.
4746 * doc/tm.texi: Regenerate.
4747 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4748 as an argument, moving assert of ad->disp == ad->disp_term to...
4749 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4750 Try calling targetm.legitimize_address_displacement before expanding
4751 the address rather than afterwards, and adjust for the new interface.
4752 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4753 Match the new hook interface. Handle SVE addresses.
4754 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4755 new hook interface.
4757 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4759 * Makefile.in (OBJS): Add early-remat.o.
4760 * target.def (select_early_remat_modes): New hook.
4761 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4762 * doc/tm.texi: Regenerate.
4763 * targhooks.h (default_select_early_remat_modes): Declare.
4764 * targhooks.c (default_select_early_remat_modes): New function.
4765 * timevar.def (TV_EARLY_REMAT): New timevar.
4766 * passes.def (pass_early_remat): New pass.
4767 * tree-pass.h (make_pass_early_remat): Declare.
4768 * early-remat.c: New file.
4769 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4770 function.
4771 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4773 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4774 Alan Hayward <alan.hayward@arm.com>
4775 David Sherwood <david.sherwood@arm.com>
4777 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4778 vfm1 with a bound_epilog parameter.
4779 (vect_do_peeling): Update calls accordingly, and move the prologue
4780 call earlier in the function. Treat the base bound_epilog as 0 for
4781 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4782 this base when peeling for gaps.
4783 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4784 with fully-masked loops.
4785 (vect_estimate_min_profitable_iters): Handle the single peeled
4786 iteration in that case.
4788 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4789 Alan Hayward <alan.hayward@arm.com>
4790 David Sherwood <david.sherwood@arm.com>
4792 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4793 single-element interleaving even if the size is not a power of 2.
4794 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4795 accesses for single-element interleaving if the group size is
4796 not a power of 2.
4798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4799 Alan Hayward <alan.hayward@arm.com>
4800 David Sherwood <david.sherwood@arm.com>
4802 * doc/md.texi (fold_extract_last_@var{m}): Document.
4803 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4804 * optabs.def (fold_extract_last_optab): New optab.
4805 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4806 * internal-fn.c (fold_extract_direct): New macro.
4807 (expand_fold_extract_optab_fn): Likewise.
4808 (direct_fold_extract_optab_supported_p): Likewise.
4809 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4810 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4811 EXTRACT_LAST_REDUCTION.
4812 (get_initial_def_for_reduction): Do not create an initial vector
4813 for EXTRACT_LAST_REDUCTION reductions.
4814 (vectorizable_reduction): Leave the scalar phi in place for
4815 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4816 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4817 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4818 transform phase to vectorizable_condition.
4819 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4820 split out from...
4821 (vect_finish_stmt_generation): ...here.
4822 (vect_finish_replace_stmt): New function.
4823 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4824 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4825 pattern.
4826 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4829 Alan Hayward <alan.hayward@arm.com>
4830 David Sherwood <david.sherwood@arm.com>
4832 * doc/md.texi (extract_last_@var{m}): Document.
4833 * optabs.def (extract_last_optab): New optab.
4834 * internal-fn.def (EXTRACT_LAST): New internal function.
4835 * internal-fn.c (cond_unary_direct): New macro.
4836 (expand_cond_unary_optab_fn): Likewise.
4837 (direct_cond_unary_optab_supported_p): Likewise.
4838 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4839 loops using EXTRACT_LAST.
4840 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4841 (extract_last_<mode>): ...this optab.
4842 (vec_extract<mode><Vel>): Update accordingly.
4844 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4845 Alan Hayward <alan.hayward@arm.com>
4846 David Sherwood <david.sherwood@arm.com>
4848 * target.def (empty_mask_is_expensive): New hook.
4849 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4850 * doc/tm.texi: Regenerate.
4851 * targhooks.h (default_empty_mask_is_expensive): Declare.
4852 * targhooks.c (default_empty_mask_is_expensive): New function.
4853 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4854 if the target says that empty masks are expensive.
4855 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4856 New function.
4857 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4859 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4860 Alan Hayward <alan.hayward@arm.com>
4861 David Sherwood <david.sherwood@arm.com>
4863 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4864 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4865 (vect_use_loop_mask_for_alignment_p): New function.
4866 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4867 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4868 niters_skip argument. Make sure that the first niters_skip elements
4869 of the first iteration are inactive.
4870 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4871 Update call to vect_set_loop_masks_directly.
4872 (get_misalign_in_elems): New function, split out from...
4873 (vect_gen_prolog_loop_niters): ...here.
4874 (vect_update_init_of_dr): Take a code argument that specifies whether
4875 the adjustment should be added or subtracted.
4876 (vect_update_init_of_drs): Likewise.
4877 (vect_prepare_for_masked_peels): New function.
4878 (vect_do_peeling): Skip prologue peeling if we're using a mask
4879 instead. Update call to vect_update_inits_of_drs.
4880 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4881 mask_skip_niters.
4882 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4883 alignment. Do not include the number of peeled iterations in
4884 the minimum threshold in that case.
4885 (vectorizable_induction): Adjust the start value down by
4886 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4887 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4888 Take the number of skipped iterations into account when calculating
4889 the loop bounds.
4890 * tree-vect-stmts.c (vect_gen_while_not): New function.
4892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4893 Alan Hayward <alan.hayward@arm.com>
4894 David Sherwood <david.sherwood@arm.com>
4896 * doc/sourcebuild.texi (vect_fully_masked): Document.
4897 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4898 default value to 0.
4899 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4900 split out from...
4901 (vect_analyze_loop_2): ...here. Don't check the vectorization
4902 factor against the number of loop iterations if the loop is
4903 fully-masked.
4905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4906 Alan Hayward <alan.hayward@arm.com>
4907 David Sherwood <david.sherwood@arm.com>
4909 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4910 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4911 (dump_groups): Update accordingly.
4912 (iv_use::mem_type): New member variable.
4913 (address_p): New function.
4914 (record_use): Add a mem_type argument and initialize the new
4915 mem_type field.
4916 (record_group_use): Add a mem_type argument. Use address_p.
4917 Remove obsolete null checks of base_object. Update call to record_use.
4918 (find_interesting_uses_op): Update call to record_group_use.
4919 (find_interesting_uses_cond): Likewise.
4920 (find_interesting_uses_address): Likewise.
4921 (get_mem_type_for_internal_fn): New function.
4922 (find_address_like_use): Likewise.
4923 (find_interesting_uses_stmt): Try find_address_like_use before
4924 calling find_interesting_uses_op.
4925 (addr_offset_valid_p): Use the iv mem_type field as the type
4926 of the addressed memory.
4927 (add_autoinc_candidates): Likewise.
4928 (get_address_cost): Likewise.
4929 (split_small_address_groups_p): Use address_p.
4930 (split_address_groups): Likewise.
4931 (add_iv_candidate_for_use): Likewise.
4932 (autoinc_possible_for_pair): Likewise.
4933 (rewrite_groups): Likewise.
4934 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4935 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4936 (get_alias_ptr_type_for_ptr_address): New function.
4937 (rewrite_use_address): Rewrite address uses in calls that were
4938 identified by find_address_like_use.
4940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4941 Alan Hayward <alan.hayward@arm.com>
4942 David Sherwood <david.sherwood@arm.com>
4944 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4945 TARGET_MEM_REFs.
4946 * gimple-expr.h (is_gimple_addressable: Likewise.
4947 * gimple-expr.c (is_gimple_address): Likewise.
4948 * internal-fn.c (expand_call_mem_ref): New function.
4949 (expand_mask_load_optab_fn): Use it.
4950 (expand_mask_store_optab_fn): Likewise.
4952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4953 Alan Hayward <alan.hayward@arm.com>
4954 David Sherwood <david.sherwood@arm.com>
4956 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4957 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4958 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4959 (cond_umax@var{mode}): Document.
4960 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4961 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4962 (cond_umin_optab, cond_umax_optab): New optabs.
4963 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4964 (COND_IOR, COND_XOR): New internal functions.
4965 * internal-fn.h (get_conditional_internal_fn): Declare.
4966 * internal-fn.c (cond_binary_direct): New macro.
4967 (expand_cond_binary_optab_fn): Likewise.
4968 (direct_cond_binary_optab_supported_p): Likewise.
4969 (get_conditional_internal_fn): New function.
4970 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4971 Cope with reduction statements that are vectorized as calls rather
4972 than assignments.
4973 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4974 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4975 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4976 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4977 (UNSPEC_COND_EOR): New unspecs.
4978 (optab): Add mappings for them.
4979 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4980 (sve_int_op, sve_fp_op): New int attributes.
4982 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4983 Alan Hayward <alan.hayward@arm.com>
4984 David Sherwood <david.sherwood@arm.com>
4986 * optabs.def (while_ult_optab): New optab.
4987 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4988 * internal-fn.def (WHILE_ULT): New internal function.
4989 * internal-fn.h (direct_internal_fn_supported_p): New override
4990 that takes two types as argument.
4991 * internal-fn.c (while_direct): New macro.
4992 (expand_while_optab_fn): New function.
4993 (convert_optab_supported_p): Likewise.
4994 (direct_while_optab_supported_p): New macro.
4995 * wide-int.h (wi::udiv_ceil): New function.
4996 * tree-vectorizer.h (rgroup_masks): New structure.
4997 (vec_loop_masks): New typedef.
4998 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4999 and fully_masked_p.
5000 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5001 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5002 (vect_max_vf): New function.
5003 (slpeel_make_loop_iterate_ntimes): Delete.
5004 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5005 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5006 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5007 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5008 internal-fn.h, stor-layout.h and optabs-query.h.
5009 (vect_set_loop_mask): New function.
5010 (add_preheader_seq): Likewise.
5011 (add_header_seq): Likewise.
5012 (interleave_supported_p): Likewise.
5013 (vect_maybe_permute_loop_masks): Likewise.
5014 (vect_set_loop_masks_directly): Likewise.
5015 (vect_set_loop_condition_masked): Likewise.
5016 (vect_set_loop_condition_unmasked): New function, split out from
5017 slpeel_make_loop_iterate_ntimes.
5018 (slpeel_make_loop_iterate_ntimes): Rename to..
5019 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5020 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5021 (vect_do_peeling): Update call accordingly.
5022 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5023 loops.
5024 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5025 mask_compare_type, can_fully_mask_p and fully_masked_p.
5026 (release_vec_loop_masks): New function.
5027 (_loop_vec_info): Use it to free the loop masks.
5028 (can_produce_all_loop_masks_p): New function.
5029 (vect_get_max_nscalars_per_iter): Likewise.
5030 (vect_verify_full_masking): Likewise.
5031 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5032 retries, and free the mask rgroups before retrying. Check loop-wide
5033 reasons for disallowing fully-masked loops. Make the final decision
5034 about whether use a fully-masked loop or not.
5035 (vect_estimate_min_profitable_iters): Do not assume that peeling
5036 for the number of iterations will be needed for fully-masked loops.
5037 (vectorizable_reduction): Disable fully-masked loops.
5038 (vectorizable_live_operation): Likewise.
5039 (vect_halve_mask_nunits): New function.
5040 (vect_double_mask_nunits): Likewise.
5041 (vect_record_loop_mask): Likewise.
5042 (vect_get_loop_mask): Likewise.
5043 (vect_transform_loop): Handle the case in which the final loop
5044 iteration might handle a partial vector. Call vect_set_loop_condition
5045 instead of slpeel_make_loop_iterate_ntimes.
5046 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5047 (check_load_store_masking): New function.
5048 (prepare_load_store_mask): Likewise.
5049 (vectorizable_store): Handle fully-masked loops.
5050 (vectorizable_load): Likewise.
5051 (supportable_widening_operation): Use vect_halve_mask_nunits for
5052 booleans.
5053 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5054 (vect_gen_while): New function.
5055 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5056 (aarch64_uqdec<mode>): New insn.
5058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5059 Alan Hayward <alan.hayward@arm.com>
5060 David Sherwood <david.sherwood@arm.com>
5062 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5063 (reduc_xor_scal_optab): New optabs.
5064 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5065 (reduc_xor_scal_@var{m}): Document.
5066 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5067 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5068 internal functions.
5069 * fold-const-call.c (fold_const_call): Handle them.
5070 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5071 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5072 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5073 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5074 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5075 (UNSPEC_XORV): New unspecs.
5076 (optab): Add entries for them.
5077 (BITWISEV): New int iterator.
5078 (bit_reduc_op): New int attributes.
5080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5081 Alan Hayward <alan.hayward@arm.com>
5082 David Sherwood <david.sherwood@arm.com>
5084 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5085 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5086 * optabs.def (vec_shl_insert_optab): New optab.
5087 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5088 (duplicate_and_interleave): Likewise.
5089 * tree-vect-loop.c: Include internal-fn.h.
5090 (neutral_op_for_slp_reduction): New function, split out from
5091 get_initial_defs_for_reduction.
5092 (get_initial_def_for_reduction): Handle option 2 for variable-length
5093 vectors by loading the neutral value into a vector and then shifting
5094 the initial value into element 0.
5095 (get_initial_defs_for_reduction): Replace the code argument with
5096 the neutral value calculated by neutral_op_for_slp_reduction.
5097 Use gimple_build_vector for constant-length vectors.
5098 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5099 but the first group_size elements have a neutral value.
5100 Use duplicate_and_interleave otherwise.
5101 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5102 Update call to get_initial_defs_for_reduction. Handle SLP
5103 reductions for variable-length vectors by creating one vector
5104 result for each scalar result, with the elements associated
5105 with other scalar results stubbed out with the neutral value.
5106 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5107 Require IFN_VEC_SHL_INSERT for double reductions on
5108 variable-length vectors, or SLP reductions that have
5109 a neutral value. Require can_duplicate_and_interleave_p
5110 support for variable-length unchained SLP reductions if there
5111 is no neutral value, such as for MIN/MAX reductions. Also require
5112 the number of vector elements to be a multiple of the number of
5113 SLP statements when doing variable-length unchained SLP reductions.
5114 Update call to vect_create_epilog_for_reduction.
5115 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5116 and remove initial values.
5117 (duplicate_and_interleave): Make public.
5118 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5119 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5122 Alan Hayward <alan.hayward@arm.com>
5123 David Sherwood <david.sherwood@arm.com>
5125 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5126 (can_duplicate_and_interleave_p): New function.
5127 (vect_get_and_check_slp_defs): Take the vector of statements
5128 rather than just the current one. Remove excess parentheses.
5129 Restriction rejectinon of vect_constant_def and vect_external_def
5130 for variable-length vectors to boolean types, or types for which
5131 can_duplicate_and_interleave_p is false.
5132 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5133 (duplicate_and_interleave): New function.
5134 (vect_get_constant_vectors): Use gimple_build_vector for
5135 constant-length vectors and suitable variable-length constant
5136 vectors. Use duplicate_and_interleave for other variable-length
5137 vectors. Don't defer the update when inserting new statements.
5139 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5140 Alan Hayward <alan.hayward@arm.com>
5141 David Sherwood <david.sherwood@arm.com>
5143 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5144 min_profitable_iters doesn't go negative.
5146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5147 Alan Hayward <alan.hayward@arm.com>
5148 David Sherwood <david.sherwood@arm.com>
5150 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5151 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5152 * optabs.def (vec_mask_load_lanes_optab): New optab.
5153 (vec_mask_store_lanes_optab): Likewise.
5154 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5155 (MASK_STORE_LANES): Likewise.
5156 * internal-fn.c (mask_load_lanes_direct): New macro.
5157 (mask_store_lanes_direct): Likewise.
5158 (expand_mask_load_optab_fn): Handle masked operations.
5159 (expand_mask_load_lanes_optab_fn): New macro.
5160 (expand_mask_store_optab_fn): Handle masked operations.
5161 (expand_mask_store_lanes_optab_fn): New macro.
5162 (direct_mask_load_lanes_optab_supported_p): Likewise.
5163 (direct_mask_store_lanes_optab_supported_p): Likewise.
5164 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5165 parameter.
5166 (vect_load_lanes_supported): Likewise.
5167 * tree-vect-data-refs.c (strip_conversion): New function.
5168 (can_group_stmts_p): Likewise.
5169 (vect_analyze_data_ref_accesses): Use it instead of checking
5170 for a pair of assignments.
5171 (vect_store_lanes_supported): Take a masked_p parameter.
5172 (vect_load_lanes_supported): Likewise.
5173 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5174 vect_store_lanes_supported and vect_load_lanes_supported.
5175 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5176 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5177 parameter. Don't allow gaps for masked accesses.
5178 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5179 and vect_load_lanes_supported.
5180 (get_load_store_type): Take a masked_p parameter and update
5181 call to get_group_load_store_type.
5182 (vectorizable_store): Update call to get_load_store_type.
5183 Handle IFN_MASK_STORE_LANES.
5184 (vectorizable_load): Update call to get_load_store_type.
5185 Handle IFN_MASK_LOAD_LANES.
5187 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5188 Alan Hayward <alan.hayward@arm.com>
5189 David Sherwood <david.sherwood@arm.com>
5191 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5192 modes for SVE.
5193 * config/aarch64/aarch64-protos.h
5194 (aarch64_sve_struct_memory_operand_p): Declare.
5195 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5196 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5197 (VPRED, vpred): Handle SVE structure modes.
5198 * config/aarch64/constraints.md (Utx): New constraint.
5199 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5200 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5201 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5202 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5203 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5204 structure modes. Split into pieces after RA.
5205 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5206 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5207 New patterns.
5208 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5209 SVE structure modes.
5210 (aarch64_classify_address): Likewise.
5211 (sizetochar): Move earlier in file.
5212 (aarch64_print_operand): Handle SVE register lists.
5213 (aarch64_array_mode): New function.
5214 (aarch64_sve_struct_memory_operand_p): Likewise.
5215 (TARGET_ARRAY_MODE): Redefine.
5217 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5218 Alan Hayward <alan.hayward@arm.com>
5219 David Sherwood <david.sherwood@arm.com>
5221 * target.def (array_mode): New target hook.
5222 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5223 * doc/tm.texi: Regenerate.
5224 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5225 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5226 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5227 targetm.array_mode.
5228 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5229 type sizes.
5231 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5232 Alan Hayward <alan.hayward@arm.com>
5233 David Sherwood <david.sherwood@arm.com>
5235 * fold-const.c (fold_binary_loc): Check the argument types
5236 rather than the result type when testing for a vector operation.
5238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5240 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5241 * doc/tm.texi: Regenerate.
5243 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5244 Alan Hayward <alan.hayward@arm.com>
5245 David Sherwood <david.sherwood@arm.com>
5247 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5248 (sve): Document new AArch64 extension.
5249 * doc/md.texi (w): Extend the description of the AArch64
5250 constraint to include SVE vectors.
5251 (Upl, Upa): Document new AArch64 predicate constraints.
5252 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5253 enum.
5254 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5255 (msve-vector-bits=): New option.
5256 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5257 SVE when these are disabled.
5258 (sve): New extension.
5259 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5260 modes. Adjust their number of units based on aarch64_sve_vg.
5261 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5262 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5263 aarch64_addr_query_type.
5264 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5265 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5266 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5267 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5268 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5269 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5270 (aarch64_simd_imm_zero_p): Delete.
5271 (aarch64_check_zero_based_sve_index_immediate): Declare.
5272 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5273 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5274 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5275 (aarch64_sve_float_mul_immediate_p): Likewise.
5276 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5277 rather than an rtx.
5278 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5279 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5280 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5281 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5282 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5283 (aarch64_regmode_natural_size): Likewise.
5284 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5285 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5286 left one place.
5287 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5288 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5289 for VG and the SVE predicate registers.
5290 (V_ALIASES): Add a "z"-prefixed alias.
5291 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5292 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5293 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5294 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5295 (REG_CLASS_NAMES): Add entries for them.
5296 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5297 and the predicate registers.
5298 (aarch64_sve_vg): Declare.
5299 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5300 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5301 (REGMODE_NATURAL_SIZE): Define.
5302 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5303 SVE macros.
5304 * config/aarch64/aarch64.c: Include cfgrtl.h.
5305 (simd_immediate_info): Add a constructor for series vectors,
5306 and an associated step field.
5307 (aarch64_sve_vg): New variable.
5308 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5309 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5310 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5311 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5312 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5313 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5314 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5315 (aarch64_get_mask_mode): New functions.
5316 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5317 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5318 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5319 predicate modes and predicate registers. Explicitly restrict
5320 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5321 to store a vector mode if it is recognized by
5322 aarch64_classify_vector_mode.
5323 (aarch64_regmode_natural_size): New function.
5324 (aarch64_hard_regno_caller_save_mode): Return the original mode
5325 for predicates.
5326 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5327 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5328 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5329 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5330 functions.
5331 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5332 does not overlap dest if the function is frame-related. Handle
5333 SVE constants.
5334 (aarch64_split_add_offset): New function.
5335 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5336 them aarch64_add_offset.
5337 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5338 and update call to aarch64_sub_sp.
5339 (aarch64_add_cfa_expression): New function.
5340 (aarch64_expand_prologue): Pass extra temporary registers to the
5341 functions above. Handle the case in which we need to emit new
5342 DW_CFA_expressions for registers that were originally saved
5343 relative to the stack pointer, but now have to be expressed
5344 relative to the frame pointer.
5345 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5346 functions above.
5347 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5348 IP0 and IP1 values for SVE frames.
5349 (aarch64_expand_vec_series): New function.
5350 (aarch64_expand_sve_widened_duplicate): Likewise.
5351 (aarch64_expand_sve_const_vector): Likewise.
5352 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5353 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5354 into the register, rather than emitting a SET directly.
5355 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5356 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5357 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5358 (offset_9bit_signed_scaled_p): New functions.
5359 (aarch64_replicate_bitmask_imm): New function.
5360 (aarch64_bitmask_imm): Use it.
5361 (aarch64_cannot_force_const_mem): Reject expressions involving
5362 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5363 (aarch64_classify_index): Handle SVE indices, by requiring
5364 a plain register index with a scale that matches the element size.
5365 (aarch64_classify_address): Handle SVE addresses. Assert that
5366 the mode of the address is VOIDmode or an integer mode.
5367 Update call to aarch64_classify_symbol.
5368 (aarch64_classify_symbolic_expression): Update call to
5369 aarch64_classify_symbol.
5370 (aarch64_const_vec_all_in_range_p): New function.
5371 (aarch64_print_vector_float_operand): Likewise.
5372 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5373 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5374 and the FP immediates 1.0 and 0.5.
5375 (aarch64_print_address_internal): Handle SVE addresses.
5376 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5377 (aarch64_regno_regclass): Handle predicate registers.
5378 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5379 data modes.
5380 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5381 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5382 (aarch64_convert_sve_vector_bits): New function.
5383 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5384 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5385 rather than an rtx.
5386 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5387 Handle SVE vector and predicate modes. Accept VL-based constants
5388 that need only one temporary register, and VL offsets that require
5389 no temporary registers.
5390 (aarch64_conditional_register_usage): Mark the predicate registers
5391 as fixed if SVE isn't available.
5392 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5393 Return true for SVE vector and predicate modes.
5394 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5395 rather than an unsigned int. Handle SVE modes.
5396 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5397 SVE modes.
5398 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5399 if SVE is enabled.
5400 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5401 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5402 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5403 (aarch64_sve_float_mul_immediate_p): New functions.
5404 (aarch64_sve_valid_immediate): New function.
5405 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5406 Explicitly reject structure modes. Check for INDEX constants.
5407 Handle PTRUE and PFALSE constants.
5408 (aarch64_check_zero_based_sve_index_immediate): New function.
5409 (aarch64_simd_imm_zero_p): Delete.
5410 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5411 vector modes. Accept constants in the range of CNT[BHWD].
5412 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5413 ask for an Advanced SIMD mode.
5414 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5415 (aarch64_simd_vector_alignment): Handle SVE predicates.
5416 (aarch64_vectorize_preferred_vector_alignment): New function.
5417 (aarch64_simd_vector_alignment_reachable): Use it instead of
5418 the vector size.
5419 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5420 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5421 functions.
5422 (MAX_VECT_LEN): Delete.
5423 (expand_vec_perm_d): Add a vec_flags field.
5424 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5425 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5426 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5427 for SVE modes.
5428 (aarch64_evpc_rev): Rename to...
5429 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5430 (aarch64_evpc_rev_global): New function.
5431 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5432 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5433 MAX_VECT_LEN.
5434 (aarch64_evpc_sve_tbl): New function.
5435 (aarch64_expand_vec_perm_const_1): Update after rename of
5436 aarch64_evpc_rev. Handle SVE permutes too, trying
5437 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5438 than aarch64_evpc_tbl.
5439 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5440 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5441 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5442 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5443 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5444 (aarch64_expand_sve_vcond): New functions.
5445 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5446 of aarch64_vector_mode_p.
5447 (aarch64_dwarf_poly_indeterminate_value): New function.
5448 (aarch64_compute_pressure_classes): Likewise.
5449 (aarch64_can_change_mode_class): Likewise.
5450 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5451 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5452 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5453 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5454 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5455 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5456 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5457 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5458 constraints.
5459 (Dn, Dl, Dr): Accept const as well as const_vector.
5460 (Dz): Likewise. Compare against CONST0_RTX.
5461 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5462 of "vector" where appropriate.
5463 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5464 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5465 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5466 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5467 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5468 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5469 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5470 (v_int_equiv): Extend to SVE modes.
5471 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5472 mode attributes.
5473 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5474 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5475 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5476 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5477 (SVE_COND_FP_CMP): New int iterators.
5478 (perm_hilo): Handle the new unpack unspecs.
5479 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5480 attributes.
5481 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5482 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5483 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5484 (aarch64_equality_operator, aarch64_constant_vector_operand)
5485 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5486 (aarch64_sve_nonimmediate_operand): Likewise.
5487 (aarch64_sve_general_operand): Likewise.
5488 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5489 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5490 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5491 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5492 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5493 (aarch64_sve_float_arith_immediate): Likewise.
5494 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5495 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5496 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5497 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5498 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5499 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5500 (aarch64_sve_float_arith_operand): Likewise.
5501 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5502 (aarch64_sve_float_mul_operand): Likewise.
5503 (aarch64_sve_vec_perm_operand): Likewise.
5504 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5505 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5506 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5507 as well as const_vector.
5508 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5509 in file. Use CONST0_RTX and CONSTM1_RTX.
5510 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5511 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5512 Use aarch64_simd_imm_zero.
5513 * config/aarch64/aarch64-sve.md: New file.
5514 * config/aarch64/aarch64.md: Include it.
5515 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5516 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5517 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5518 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5519 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5520 (sve): New attribute.
5521 (enabled): Disable instructions with the sve attribute unless
5522 TARGET_SVE.
5523 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5524 aarch64_expand_mov_immediate.
5525 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5526 CNT[BHSD] immediates.
5527 (movti): Split CONST_POLY_INT moves into two halves.
5528 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5529 Split additions that need a temporary here if the destination
5530 is the stack pointer.
5531 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5532 (*add<mode>3_poly_1): New instruction.
5533 (set_clobber_cc): New expander.
5535 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5537 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5538 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5539 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5540 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5541 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5542 Change innermode from fixed_mode_size to machine_mode.
5543 (simplify_subreg): Update call accordingly. Handle a constant-sized
5544 subreg of a variable-length CONST_VECTOR.
5546 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5547 Alan Hayward <alan.hayward@arm.com>
5548 David Sherwood <david.sherwood@arm.com>
5550 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5551 (add_offset_to_base): New function, split out from...
5552 (create_mem_ref): ...here. When handling a scale other than 1,
5553 check first whether the address is valid without the offset.
5554 Add it into the base if so, leaving the index and scale as-is.
5556 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5558 PR c++/83778
5559 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5560 fold_for_warn before checking if arg2 is INTEGER_CST.
5562 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5564 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5565 (store_multiple_operation): Delete.
5566 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5567 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5568 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5569 guarded by TARGET_STRING.
5570 (rs6000_output_load_multiple): Delete.
5571 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5572 OPTION_MASK_STRING / TARGET_STRING handling.
5573 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5574 (const rs6000_opt_masks) <"string">: Change mask to 0.
5575 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5576 (MASK_STRING): Delete.
5577 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5578 parts. Simplify.
5579 (load_multiple): Delete.
5580 (*ldmsi8): Delete.
5581 (*ldmsi7): Delete.
5582 (*ldmsi6): Delete.
5583 (*ldmsi5): Delete.
5584 (*ldmsi4): Delete.
5585 (*ldmsi3): Delete.
5586 (store_multiple): Delete.
5587 (*stmsi8): Delete.
5588 (*stmsi7): Delete.
5589 (*stmsi6): Delete.
5590 (*stmsi5): Delete.
5591 (*stmsi4): Delete.
5592 (*stmsi3): Delete.
5593 (movmemsi_8reg): Delete.
5594 (corresponding unnamed define_insn): Delete.
5595 (movmemsi_6reg): Delete.
5596 (corresponding unnamed define_insn): Delete.
5597 (movmemsi_4reg): Delete.
5598 (corresponding unnamed define_insn): Delete.
5599 (movmemsi_2reg): Delete.
5600 (corresponding unnamed define_insn): Delete.
5601 (movmemsi_1reg): Delete.
5602 (corresponding unnamed define_insn): Delete.
5603 * config/rs6000/rs6000.opt (mno-string): New.
5604 (mstring): Replace by deprecation warning stub.
5605 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5607 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5609 * regrename.c (regrename_do_replace): If replacing the same
5610 reg multiple times, try to reuse last created gen_raw_REG.
5612 PR debug/81155
5613 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5614 main to workaround a bug in GDB.
5616 2018-01-12 Tom de Vries <tom@codesourcery.com>
5618 PR target/83737
5619 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5621 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5623 PR rtl-optimization/80481
5624 * ira-color.c (get_cap_member): New function.
5625 (allocnos_conflict_by_live_ranges_p): Use it.
5626 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5627 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5629 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5631 PR target/83628
5632 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5633 (*saddl_se_1): Ditto.
5634 (*ssubsi_1): Ditto.
5635 (*ssubl_se_1): Ditto.
5637 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5639 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5640 rather than wi::to_widest for DR_INITs.
5641 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5642 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5643 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5644 INTEGER_CSTs.
5645 (vect_analyze_group_access_1): Note that here.
5647 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5649 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5650 polynomial type sizes.
5652 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5654 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5655 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5656 (gimple_add_tmp_var): Likewise.
5658 2018-01-12 Martin Liska <mliska@suse.cz>
5660 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5661 (gimple_alloc_sizes): Likewise.
5662 (dump_gimple_statistics): Use PRIu64 in printf format.
5663 * gimple.h: Change uint64_t to int.
5665 2018-01-12 Martin Liska <mliska@suse.cz>
5667 * tree-core.h: Use uint64_t instead of int.
5668 * tree.c (tree_node_counts): Likewise.
5669 (tree_node_sizes): Likewise.
5670 (dump_tree_statistics): Use PRIu64 in printf format.
5672 2018-01-12 Martin Liska <mliska@suse.cz>
5674 * Makefile.in: As qsort_chk is implemented in vec.c, add
5675 vec.o to linkage of gencfn-macros.
5676 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5677 passing the info to record_node_allocation_statistics.
5678 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5679 and pass the info.
5680 * ggc-common.c (struct ggc_usage): Add operator== and use
5681 it in operator< and compare function.
5682 * mem-stats.h (struct mem_usage): Likewise.
5683 * vec.c (struct vec_usage): Remove operator< and compare
5684 function. Can be simply inherited.
5686 2018-01-12 Martin Jambor <mjambor@suse.cz>
5688 PR target/81616
5689 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5690 * tree-ssa-math-opts.c: Include domwalk.h.
5691 (convert_mult_to_fma_1): New function.
5692 (fma_transformation_info): New type.
5693 (fma_deferring_state): Likewise.
5694 (cancel_fma_deferring): New function.
5695 (result_of_phi): Likewise.
5696 (last_fma_candidate_feeds_initial_phi): Likewise.
5697 (convert_mult_to_fma): Added deferring logic, split actual
5698 transformation to convert_mult_to_fma_1.
5699 (math_opts_dom_walker): New type.
5700 (math_opts_dom_walker::after_dom_children): New method, body moved
5701 here from pass_optimize_widening_mul::execute, added deferring logic
5702 bits.
5703 (pass_optimize_widening_mul::execute): Moved most of code to
5704 math_opts_dom_walker::after_dom_children.
5705 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5706 * config/i386/i386.c (ix86_option_override_internal): Added
5707 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5709 2018-01-12 Richard Biener <rguenther@suse.de>
5711 PR debug/83157
5712 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5713 inline instance vars.
5715 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5717 PR target/81819
5718 * config/rx/rx.c (rx_is_restricted_memory_address):
5719 Handle SUBREG case.
5721 2018-01-12 Richard Biener <rguenther@suse.de>
5723 PR tree-optimization/80846
5724 * target.def (split_reduction): New target hook.
5725 * targhooks.c (default_split_reduction): New function.
5726 * targhooks.h (default_split_reduction): Declare.
5727 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5728 target requests first reduce vectors by combining low and high
5729 parts.
5730 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5731 (get_vectype_for_scalar_type_and_size): Export.
5732 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5733 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5734 * doc/tm.texi: Regenerate.
5735 * config/i386/i386.c (ix86_split_reduction): Implement
5736 TARGET_VECTORIZE_SPLIT_REDUCTION.
5738 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5740 PR target/83368
5741 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5742 in PIC mode except for TARGET_VXWORKS_RTP.
5743 * config/sparc/sparc.c: Include cfgrtl.h.
5744 (TARGET_INIT_PIC_REG): Define.
5745 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5746 (sparc_pic_register_p): New predicate.
5747 (sparc_legitimate_address_p): Use it.
5748 (sparc_legitimize_pic_address): Likewise.
5749 (sparc_delegitimize_address): Likewise.
5750 (sparc_mode_dependent_address_p): Likewise.
5751 (gen_load_pcrel_sym): Remove 4th parameter.
5752 (load_got_register): Adjust call to above. Remove obsolete stuff.
5753 (sparc_expand_prologue): Do not call load_got_register here.
5754 (sparc_flat_expand_prologue): Likewise.
5755 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5756 (sparc_use_pseudo_pic_reg): New function.
5757 (sparc_init_pic_reg): Likewise.
5758 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5759 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5761 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5763 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5764 Add item for branch_cost.
5766 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5768 PR rtl-optimization/83565
5769 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5770 not extend the result to a larger mode for rotate operations.
5771 (num_sign_bit_copies1): Likewise.
5773 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5775 PR target/40411
5776 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5777 -symbolic.
5778 Use values-Xc.o for -pedantic.
5779 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5781 2018-01-12 Martin Liska <mliska@suse.cz>
5783 PR ipa/83054
5784 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5785 New function.
5786 (possible_polymorphic_call_targets): Use it.
5787 (ipa_devirt): Likewise.
5789 2018-01-12 Martin Liska <mliska@suse.cz>
5791 * profile-count.h (enum profile_quality): Use 0 as invalid
5792 enum value of profile_quality.
5794 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5796 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5797 -mext-string options.
5799 2018-01-12 Richard Biener <rguenther@suse.de>
5801 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5802 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5803 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5804 Likewise.
5805 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5807 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5809 * configure.ac (--with-long-double-format): Add support for the
5810 configuration option to change the default long double format on
5811 PowerPC systems.
5812 * config.gcc (powerpc*-linux*-*): Likewise.
5813 * configure: Regenerate.
5814 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5815 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5816 used without modification.
5818 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5820 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5821 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5822 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5823 MISC_BUILTIN_SPEC_BARRIER.
5824 (rs6000_init_builtins): Likewise.
5825 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5826 enum value.
5827 (speculation_barrier): New define_insn.
5828 * doc/extend.texi: Document __builtin_speculation_barrier.
5830 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5832 PR target/83203
5833 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5834 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5835 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5836 iterators.
5837 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5838 integral modes instead of "ss" and "sd".
5839 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5840 vectors with 32-bit and 64-bit elements.
5841 (vecdupssescalarmodesuffix): New mode attribute.
5842 (vec_dup<mode>): Use it.
5844 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5846 PR target/83330
5847 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5848 frame if argument is passed on stack.
5850 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5852 PR target/82682
5853 * ree.c (combine_reaching_defs): Optimize also
5854 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5855 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5857 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5859 PR middle-end/83189
5860 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5862 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5864 PR middle-end/83718
5865 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5866 after they are computed.
5868 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5870 PR tree-optimization/83695
5871 * gimple-loop-linterchange.cc
5872 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5873 reset cached scev information after interchange.
5874 (pass_linterchange::execute): Remove call to scev_reset_htab.
5876 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5878 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5879 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5880 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5881 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5882 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5883 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5884 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5885 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5886 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5887 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5888 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5889 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5890 (V_lane_reg): Likewise.
5891 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5892 New define_expand.
5893 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5894 (vfmal_lane_low<mode>_intrinsic,
5895 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5896 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5897 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5898 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5899 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5900 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5902 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5904 * config/arm/arm-cpus.in (fp16fml): New feature.
5905 (ALL_SIMD): Add fp16fml.
5906 (armv8.2-a): Add fp16fml as an option.
5907 (armv8.3-a): Likewise.
5908 (armv8.4-a): Add fp16fml as part of fp16.
5909 * config/arm/arm.h (TARGET_FP16FML): Define.
5910 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5911 when appropriate.
5912 * config/arm/arm-modes.def (V2HF): Define.
5913 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5914 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5915 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5916 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5917 vfmsl_low, vfmsl_high): New set of builtins.
5918 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5919 (vfml_op): New code attribute.
5920 (VFMLHALVES): New int iterator.
5921 (VFML, VFMLSEL): New mode attributes.
5922 (V_reg): Define mapping for V2HF.
5923 (V_hi, V_lo): New mode attributes.
5924 (VF_constraint): Likewise.
5925 (vfml_half, vfml_half_selector): New int attributes.
5926 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5927 define_expand.
5928 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5929 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5930 New define_insn.
5931 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5932 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5933 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5934 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5935 documentation.
5936 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5937 Document new effective target and option set.
5939 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5941 * config/arm/arm-cpus.in (armv8_4): New feature.
5942 (ARMv8_4a): New fgroup.
5943 (armv8.4-a): New arch.
5944 * config/arm/arm-tables.opt: Regenerate.
5945 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5946 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5947 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5948 Add matching rules for -march=armv8.4-a and extensions.
5949 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5951 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5953 PR target/81821
5954 * config/rx/rx.md (BW): New mode attribute.
5955 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5957 2018-01-11 Richard Biener <rguenther@suse.de>
5959 PR tree-optimization/83435
5960 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5961 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5962 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5964 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5965 Alan Hayward <alan.hayward@arm.com>
5966 David Sherwood <david.sherwood@arm.com>
5968 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5969 field.
5970 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5971 (aarch64_print_address_internal): Use it to check for a zero offset.
5973 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5974 Alan Hayward <alan.hayward@arm.com>
5975 David Sherwood <david.sherwood@arm.com>
5977 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5978 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5979 Return a poly_int64 rather than a HOST_WIDE_INT.
5980 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5981 rather than a HOST_WIDE_INT.
5982 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5983 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5984 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5985 final_offset from HOST_WIDE_INT to poly_int64.
5986 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5987 to_constant when getting the number of units in an Advanced SIMD
5988 mode.
5989 (aarch64_builtin_vectorized_function): Check for a constant number
5990 of units.
5991 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5992 GET_MODE_SIZE.
5993 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5994 attribute instead of GET_MODE_NUNITS.
5995 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5996 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5997 GET_MODE_SIZE for fixed-size registers.
5998 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5999 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6000 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6001 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6002 (aarch64_print_operand, aarch64_print_address_internal)
6003 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6004 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6005 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6006 Handle polynomial GET_MODE_SIZE.
6007 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6008 wider than SImode without modification.
6009 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6010 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6011 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6012 passing and returning SVE modes.
6013 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6014 rather than GEN_INT.
6015 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6016 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6017 (aarch64_allocate_and_probe_stack_space): Likewise.
6018 (aarch64_layout_frame): Cope with polynomial offsets.
6019 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6020 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6021 polynomial offsets.
6022 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6023 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6024 poly_int64 rather than a HOST_WIDE_INT.
6025 (aarch64_get_separate_components, aarch64_process_components)
6026 (aarch64_expand_prologue, aarch64_expand_epilogue)
6027 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6028 (aarch64_anchor_offset): New function, split out from...
6029 (aarch64_legitimize_address): ...here.
6030 (aarch64_builtin_vectorization_cost): Handle polynomial
6031 TYPE_VECTOR_SUBPARTS.
6032 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6033 GET_MODE_NUNITS.
6034 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6035 number of elements from the PARALLEL rather than the mode.
6036 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6037 rather than GET_MODE_BITSIZE.
6038 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6039 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6040 (aarch64_expand_vec_perm_const_1): Handle polynomial
6041 d->perm.length () and d->perm elements.
6042 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6043 Apply to_constant to d->perm elements.
6044 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6045 polynomial CONST_VECTOR_NUNITS.
6046 (aarch64_move_pointer): Take amount as a poly_int64 rather
6047 than an int.
6048 (aarch64_progress_pointer): Avoid temporary variable.
6049 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6050 the mode attribute instead of GET_MODE.
6052 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6053 Alan Hayward <alan.hayward@arm.com>
6054 David Sherwood <david.sherwood@arm.com>
6056 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6057 x exists before using it.
6058 (aarch64_add_constant_internal): Rename to...
6059 (aarch64_add_offset_1): ...this. Replace regnum with separate
6060 src and dest rtxes. Handle the case in which they're different,
6061 including when the offset is zero. Replace scratchreg with an rtx.
6062 Use 2 additions if there is no spare register into which we can
6063 move a 16-bit constant.
6064 (aarch64_add_constant): Delete.
6065 (aarch64_add_offset): Replace reg with separate src and dest
6066 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6067 Use aarch64_add_offset_1.
6068 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6069 an rtx rather than an int. Take the delta as a poly_int64
6070 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6071 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6072 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6073 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6074 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6075 and aarch64_add_sp.
6076 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6077 aarch64_add_constant.
6079 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6081 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6082 Use scalar_float_mode.
6084 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6086 * config/aarch64/aarch64-simd.md
6087 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6088 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6089 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6090 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6091 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6092 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6093 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6094 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6095 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6096 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6098 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6100 PR target/83514
6101 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6102 targ_options->x_arm_arch_string is non NULL.
6104 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6106 * config/aarch64/aarch64.h
6107 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6109 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6111 PR target/82096
6112 * expmed.c (emit_store_flag_force): Swap if const op0
6113 and change VOIDmode to mode of op0.
6115 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6117 PR rtl-optimization/83761
6118 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6119 than bytes to mode_for_size.
6121 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6123 PR middle-end/83189
6124 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6125 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6126 profile.
6128 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6130 PR middle-end/83575
6131 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6132 when in layout mode.
6133 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6134 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6135 partition fixup.
6137 2018-01-10 Michael Collison <michael.collison@arm.com>
6139 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6140 * config/aarch64/aarch64-option-extension.def: Add
6141 AARCH64_OPT_EXTENSION of 'fp16fml'.
6142 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6143 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6144 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6145 * config/aarch64/constraints.md (Ui7): New constraint.
6146 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6147 (VFMLA_SEL_W): Ditto.
6148 (f16quad): Ditto.
6149 (f16mac1): Ditto.
6150 (VFMLA16_LOW): New int iterator.
6151 (VFMLA16_HIGH): Ditto.
6152 (UNSPEC_FMLAL): New unspec.
6153 (UNSPEC_FMLSL): Ditto.
6154 (UNSPEC_FMLAL2): Ditto.
6155 (UNSPEC_FMLSL2): Ditto.
6156 (f16mac): New code attribute.
6157 * config/aarch64/aarch64-simd-builtins.def
6158 (aarch64_fmlal_lowv2sf): Ditto.
6159 (aarch64_fmlsl_lowv2sf): Ditto.
6160 (aarch64_fmlalq_lowv4sf): Ditto.
6161 (aarch64_fmlslq_lowv4sf): Ditto.
6162 (aarch64_fmlal_highv2sf): Ditto.
6163 (aarch64_fmlsl_highv2sf): Ditto.
6164 (aarch64_fmlalq_highv4sf): Ditto.
6165 (aarch64_fmlslq_highv4sf): Ditto.
6166 (aarch64_fmlal_lane_lowv2sf): Ditto.
6167 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6168 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6169 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6170 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6171 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6172 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6173 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6174 (aarch64_fmlal_lane_highv2sf): Ditto.
6175 (aarch64_fmlsl_lane_highv2sf): Ditto.
6176 (aarch64_fmlal_laneq_highv2sf): Ditto.
6177 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6178 (aarch64_fmlalq_lane_highv4sf): Ditto.
6179 (aarch64_fmlsl_lane_highv4sf): Ditto.
6180 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6181 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6182 * config/aarch64/aarch64-simd.md:
6183 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6184 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6185 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6186 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6187 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6188 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6189 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6190 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6191 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6192 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6193 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6194 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6195 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6196 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6197 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6198 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6199 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6200 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6201 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6202 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6203 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6204 (vfmlsl_low_u32): Ditto.
6205 (vfmlalq_low_u32): Ditto.
6206 (vfmlslq_low_u32): Ditto.
6207 (vfmlal_high_u32): Ditto.
6208 (vfmlsl_high_u32): Ditto.
6209 (vfmlalq_high_u32): Ditto.
6210 (vfmlslq_high_u32): Ditto.
6211 (vfmlal_lane_low_u32): Ditto.
6212 (vfmlsl_lane_low_u32): Ditto.
6213 (vfmlal_laneq_low_u32): Ditto.
6214 (vfmlsl_laneq_low_u32): Ditto.
6215 (vfmlalq_lane_low_u32): Ditto.
6216 (vfmlslq_lane_low_u32): Ditto.
6217 (vfmlalq_laneq_low_u32): Ditto.
6218 (vfmlslq_laneq_low_u32): Ditto.
6219 (vfmlal_lane_high_u32): Ditto.
6220 (vfmlsl_lane_high_u32): Ditto.
6221 (vfmlal_laneq_high_u32): Ditto.
6222 (vfmlsl_laneq_high_u32): Ditto.
6223 (vfmlalq_lane_high_u32): Ditto.
6224 (vfmlslq_lane_high_u32): Ditto.
6225 (vfmlalq_laneq_high_u32): Ditto.
6226 (vfmlslq_laneq_high_u32): Ditto.
6227 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6228 (AARCH64_FL_FOR_ARCH8_4): New.
6229 (AARCH64_ISA_F16FML): New ISA flag.
6230 (TARGET_F16FML): New feature flag for fp16fml.
6231 (doc/invoke.texi): Document new fp16fml option.
6233 2018-01-10 Michael Collison <michael.collison@arm.com>
6235 * config/aarch64/aarch64-builtins.c:
6236 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6237 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6238 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6239 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6240 (AARCH64_ISA_SHA3): New ISA flag.
6241 (TARGET_SHA3): New feature flag for sha3.
6242 * config/aarch64/iterators.md (sha512_op): New int attribute.
6243 (CRYPTO_SHA512): New int iterator.
6244 (UNSPEC_SHA512H): New unspec.
6245 (UNSPEC_SHA512H2): Ditto.
6246 (UNSPEC_SHA512SU0): Ditto.
6247 (UNSPEC_SHA512SU1): Ditto.
6248 * config/aarch64/aarch64-simd-builtins.def
6249 (aarch64_crypto_sha512hqv2di): New builtin.
6250 (aarch64_crypto_sha512h2qv2di): Ditto.
6251 (aarch64_crypto_sha512su0qv2di): Ditto.
6252 (aarch64_crypto_sha512su1qv2di): Ditto.
6253 (aarch64_eor3qv8hi): Ditto.
6254 (aarch64_rax1qv2di): Ditto.
6255 (aarch64_xarqv2di): Ditto.
6256 (aarch64_bcaxqv8hi): Ditto.
6257 * config/aarch64/aarch64-simd.md:
6258 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6259 (aarch64_crypto_sha512su0qv2di): Ditto.
6260 (aarch64_crypto_sha512su1qv2di): Ditto.
6261 (aarch64_eor3qv8hi): Ditto.
6262 (aarch64_rax1qv2di): Ditto.
6263 (aarch64_xarqv2di): Ditto.
6264 (aarch64_bcaxqv8hi): Ditto.
6265 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6266 (vsha512h2q_u64): Ditto.
6267 (vsha512su0q_u64): Ditto.
6268 (vsha512su1q_u64): Ditto.
6269 (veor3q_u16): Ditto.
6270 (vrax1q_u64): Ditto.
6271 (vxarq_u64): Ditto.
6272 (vbcaxq_u16): Ditto.
6273 * config/arm/types.md (crypto_sha512): New type attribute.
6274 (crypto_sha3): Ditto.
6275 (doc/invoke.texi): Document new sha3 option.
6277 2018-01-10 Michael Collison <michael.collison@arm.com>
6279 * config/aarch64/aarch64-builtins.c:
6280 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6281 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6282 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6283 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6284 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6285 (AARCH64_ISA_SM4): New ISA flag.
6286 (TARGET_SM4): New feature flag for sm4.
6287 * config/aarch64/aarch64-simd-builtins.def
6288 (aarch64_sm3ss1qv4si): Ditto.
6289 (aarch64_sm3tt1aq4si): Ditto.
6290 (aarch64_sm3tt1bq4si): Ditto.
6291 (aarch64_sm3tt2aq4si): Ditto.
6292 (aarch64_sm3tt2bq4si): Ditto.
6293 (aarch64_sm3partw1qv4si): Ditto.
6294 (aarch64_sm3partw2qv4si): Ditto.
6295 (aarch64_sm4eqv4si): Ditto.
6296 (aarch64_sm4ekeyqv4si): Ditto.
6297 * config/aarch64/aarch64-simd.md:
6298 (aarch64_sm3ss1qv4si): Ditto.
6299 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6300 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6301 (aarch64_sm4eqv4si): Ditto.
6302 (aarch64_sm4ekeyqv4si): Ditto.
6303 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6304 (sm3part_op): Ditto.
6305 (CRYPTO_SM3TT): Ditto.
6306 (CRYPTO_SM3PART): Ditto.
6307 (UNSPEC_SM3SS1): New unspec.
6308 (UNSPEC_SM3TT1A): Ditto.
6309 (UNSPEC_SM3TT1B): Ditto.
6310 (UNSPEC_SM3TT2A): Ditto.
6311 (UNSPEC_SM3TT2B): Ditto.
6312 (UNSPEC_SM3PARTW1): Ditto.
6313 (UNSPEC_SM3PARTW2): Ditto.
6314 (UNSPEC_SM4E): Ditto.
6315 (UNSPEC_SM4EKEY): Ditto.
6316 * config/aarch64/constraints.md (Ui2): New constraint.
6317 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6318 * config/arm/types.md (crypto_sm3): New type attribute.
6319 (crypto_sm4): Ditto.
6320 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6321 (vsm3tt1aq_u32): Ditto.
6322 (vsm3tt1bq_u32): Ditto.
6323 (vsm3tt2aq_u32): Ditto.
6324 (vsm3tt2bq_u32): Ditto.
6325 (vsm3partw1q_u32): Ditto.
6326 (vsm3partw2q_u32): Ditto.
6327 (vsm4eq_u32): Ditto.
6328 (vsm4ekeyq_u32): Ditto.
6329 (doc/invoke.texi): Document new sm4 option.
6331 2018-01-10 Michael Collison <michael.collison@arm.com>
6333 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6334 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6335 (AARCH64_FL_FOR_ARCH8_4): New.
6336 (AARCH64_FL_V8_4): New flag.
6337 (doc/invoke.texi): Document new armv8.4-a option.
6339 2018-01-10 Michael Collison <michael.collison@arm.com>
6341 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6342 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6343 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6344 * config/aarch64/aarch64-option-extension.def: Add
6345 AARCH64_OPT_EXTENSION of 'sha2'.
6346 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6347 (crypto): Disable sha2 and aes if crypto disabled.
6348 (crypto): Enable aes and sha2 if enabled.
6349 (simd): Disable sha2 and aes if simd disabled.
6350 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6351 New flags.
6352 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6353 (TARGET_SHA2): New feature flag for sha2.
6354 (TARGET_AES): New feature flag for aes.
6355 * config/aarch64/aarch64-simd.md:
6356 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6357 conditional on TARGET_AES.
6358 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6359 (aarch64_crypto_sha1hsi): Make pattern conditional
6360 on TARGET_SHA2.
6361 (aarch64_crypto_sha1hv4si): Ditto.
6362 (aarch64_be_crypto_sha1hv4si): Ditto.
6363 (aarch64_crypto_sha1su1v4si): Ditto.
6364 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6365 (aarch64_crypto_sha1su0v4si): Ditto.
6366 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6367 (aarch64_crypto_sha256su0v4si): Ditto.
6368 (aarch64_crypto_sha256su1v4si): Ditto.
6369 (doc/invoke.texi): Document new aes and sha2 options.
6371 2018-01-10 Martin Sebor <msebor@redhat.com>
6373 PR tree-optimization/83781
6374 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6375 as string arrays.
6377 2018-01-11 Martin Sebor <msebor@gmail.com>
6378 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6380 PR tree-optimization/83501
6381 PR tree-optimization/81703
6383 * tree-ssa-strlen.c (get_string_cst): Rename...
6384 (get_string_len): ...to this. Handle global constants.
6385 (handle_char_store): Adjust.
6387 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6388 Jim Wilson <jimw@sifive.com>
6390 * config/riscv/riscv-protos.h (riscv_output_return): New.
6391 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6392 (riscv_attribute_table, riscv_output_return),
6393 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6394 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6395 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6396 function.
6397 (riscv_expand_prologue): Add early return for naked function.
6398 (riscv_expand_epilogue): Likewise.
6399 (riscv_function_ok_for_sibcall): Return false for naked function.
6400 (riscv_set_current_function): New.
6401 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6402 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6403 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6404 * doc/extend.texi (RISC-V Function Attributes): New.
6406 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6408 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6409 check for 128-bit long double before checking TCmode.
6410 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6411 128-bit long doubles before checking TFmode or TCmode.
6412 (FLOAT128_IBM_P): Likewise.
6414 2018-01-10 Martin Sebor <msebor@redhat.com>
6416 PR tree-optimization/83671
6417 * builtins.c (c_strlen): Unconditionally return zero for the empty
6418 string.
6419 Use -Warray-bounds for warnings.
6420 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6421 for non-constant array indices with COMPONENT_REF, arrays of
6422 arrays, and pointers to arrays.
6423 (gimple_fold_builtin_strlen): Determine and set length range for
6424 non-constant character arrays.
6426 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6428 PR middle-end/81897
6429 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6430 empty blocks.
6432 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6434 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6436 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6438 PR target/83399
6439 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6440 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6441 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6442 indexed_or_indirect_operand predicate.
6443 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6444 (*vsx_le_perm_load_v8hi): Likewise.
6445 (*vsx_le_perm_load_v16qi): Likewise.
6446 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6447 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6448 (*vsx_le_perm_store_v8hi): Likewise.
6449 (*vsx_le_perm_store_v16qi): Likewise.
6450 (eight unnamed splitters): Likewise.
6452 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6454 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6455 * config/rs6000/emmintrin.h: Likewise.
6456 * config/rs6000/mmintrin.h: Likewise.
6457 * config/rs6000/xmmintrin.h: Likewise.
6459 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6461 PR c++/43486
6462 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6463 "public_flag".
6464 * tree.c (tree_nop_conversion): Return true for location wrapper
6465 nodes.
6466 (maybe_wrap_with_location): New function.
6467 (selftest::check_strip_nops): New function.
6468 (selftest::test_location_wrappers): New function.
6469 (selftest::tree_c_tests): Call it.
6470 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6471 (maybe_wrap_with_location): New decl.
6472 (EXPR_LOCATION_WRAPPER_P): New macro.
6473 (location_wrapper_p): New inline function.
6474 (tree_strip_any_location_wrapper): New inline function.
6476 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6478 PR target/83735
6479 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6480 stack_realign_offset for the largest alignment of stack slot
6481 actually used.
6482 (ix86_find_max_used_stack_alignment): New function.
6483 (ix86_finalize_stack_frame_flags): Use it. Set
6484 max_used_stack_alignment if we don't realign stack.
6485 * config/i386/i386.h (machine_function): Add
6486 max_used_stack_alignment.
6488 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6490 * config/arm/arm.opt (-mbranch-cost): New option.
6491 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6492 account.
6494 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6496 PR target/83629
6497 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6498 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6500 2018-01-10 Richard Biener <rguenther@suse.de>
6502 PR debug/83765
6503 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6504 early out so it also covers the case where we have a non-NULL
6505 origin.
6507 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6509 PR tree-optimization/83753
6510 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6511 for non-strided grouped accesses if the number of elements is 1.
6513 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6515 PR target/81616
6516 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6517 * i386.h (TARGET_USE_GATHER): Define.
6518 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6520 2018-01-10 Martin Liska <mliska@suse.cz>
6522 PR bootstrap/82831
6523 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6524 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6525 partitioning.
6526 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6527 CLEANUP_NO_PARTITIONING is not set.
6529 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6531 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6532 for vectors, as a partial revert of r254296.
6533 * rtl.h (const_vec_p): Delete.
6534 (const_vec_duplicate_p): Don't test for vector CONSTs.
6535 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6536 * expmed.c (make_tree): Likewise.
6538 Revert:
6539 * common.md (E, F): Use CONSTANT_P instead of checking for
6540 CONST_VECTOR.
6541 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6542 checking for CONST_VECTOR.
6544 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6546 PR middle-end/83575
6547 * predict.c (force_edge_cold): Handle in more sane way edges
6548 with no prediction.
6550 2018-01-09 Carl Love <cel@us.ibm.com>
6552 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6553 V4SI, V4SF types.
6554 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6555 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6556 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6557 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6558 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6559 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6560 * config/rs6000/rs6000-protos.h: Add extern defition for
6561 rs6000_generate_float2_double_code.
6562 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6563 function.
6564 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6565 (float2_v2df): Add define_expand.
6567 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6569 PR target/83628
6570 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6571 op_mode in the force_to_mode call.
6573 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6575 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6576 instead of checking each element individually.
6577 (aarch64_evpc_uzp): Likewise.
6578 (aarch64_evpc_zip): Likewise.
6579 (aarch64_evpc_ext): Likewise.
6580 (aarch64_evpc_rev): Likewise.
6581 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6582 instead of checking each element individually. Return true without
6583 generating rtl if
6584 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6585 whether all selected elements come from the same input, instead of
6586 checking each element individually. Remove calls to gen_rtx_REG,
6587 start_sequence and end_sequence and instead assert that no rtl is
6588 generated.
6590 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6592 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6593 order of HIGH and CONST checks.
6595 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6597 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6598 if the destination isn't an SSA_NAME.
6600 2018-01-09 Richard Biener <rguenther@suse.de>
6602 PR tree-optimization/83668
6603 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6604 move prologue...
6605 (canonicalize_loop_form): ... here, renamed from ...
6606 (canonicalize_loop_closed_ssa_form): ... this and amended to
6607 swap successor edges for loop exit blocks to make us use
6608 the RPO order we need for initial schedule generation.
6610 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6612 PR tree-optimization/64811
6613 * match.pd: When optimizing comparisons with Inf, avoid
6614 introducing or losing exceptions from comparisons with NaN.
6616 2018-01-09 Martin Liska <mliska@suse.cz>
6618 PR sanitizer/82517
6619 * asan.c (shadow_mem_size): Add gcc_assert.
6621 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6623 Don't save registers in main().
6625 PR target/83738
6626 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6627 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6628 * config/avr/avr.c (avr_set_current_function): Don't error if
6629 naked, OS_task or OS_main are specified at the same time.
6630 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6631 OS_main.
6632 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6633 attribute.
6634 * common/config/avr/avr-common.c (avr_option_optimization_table):
6635 Switch on -mmain-is-OS_task for optimizing compilations.
6637 2018-01-09 Richard Biener <rguenther@suse.de>
6639 PR tree-optimization/83572
6640 * graphite.c: Include cfganal.h.
6641 (graphite_transform_loops): Connect infinite loops to exit
6642 and remove fake edges at the end.
6644 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6646 * ipa-inline.c (edge_badness): Revert accidental checkin.
6648 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6650 PR ipa/80763
6651 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6652 symbols; not inline clones.
6654 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6656 PR target/83507
6657 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6658 hard registers. Formatting fixes.
6660 PR preprocessor/83722
6661 * gcc.c (try_generate_repro): Pass
6662 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6663 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6664 do_report_bug.
6666 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6667 Kito Cheng <kito.cheng@gmail.com>
6669 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6670 (riscv_leaf_function_p): Delete.
6671 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6673 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6675 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6676 function.
6677 (do_ifelse): New function.
6678 (do_isel): New function.
6679 (do_sub3): New function.
6680 (do_add3): New function.
6681 (do_load_mask_compare): New function.
6682 (do_overlap_load_compare): New function.
6683 (expand_compare_loop): New function.
6684 (expand_block_compare): Call expand_compare_loop() when appropriate.
6685 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6686 option description.
6687 (-mblock-compare-inline-loop-limit): New option.
6689 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6691 PR target/83677
6692 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6693 Reverse order of second and third operands in first alternative.
6694 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6695 of first and second elements in UNSPEC_VPERMR vector.
6696 (altivec_expand_vec_perm_le): Likewise.
6698 2018-01-08 Jeff Law <law@redhat.com>
6700 PR rtl-optimizatin/81308
6701 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6702 (process_switch): If group_case_labels makes a change, then set
6703 cfg_altered.
6704 (pass_convert_switch::execute): If a switch is converted, then
6705 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6707 PR rtl-optimization/81308
6708 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6709 splitting insns.
6711 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6713 PR target/83663 - Revert r255946
6714 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6715 generation for cases where splatting a value is not useful.
6716 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6717 across a vec_duplicate and a paradoxical subreg forming a vector
6718 mode to a vec_concat.
6720 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6722 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6723 -march=armv8.3-a variants.
6724 * config/arm/t-multilib: Likewise.
6725 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6727 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6729 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6730 to generate rtl.
6731 (cceq_ior_compare_complement): Give it a name so I can use it, and
6732 change boolean_or_operator predicate to boolean_operator so it can
6733 be used to generate a crand.
6734 (eqne): New code iterator.
6735 (bd/bd_neg): New code_attrs.
6736 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6737 a single define_insn.
6738 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6739 decrement (bdnzt/bdnzf/bdzt/bdzf).
6740 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6741 with the new names of the branch decrement patterns, and added the
6742 names of the branch decrement conditional patterns.
6744 2018-01-08 Richard Biener <rguenther@suse.de>
6746 PR tree-optimization/83563
6747 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6748 cache.
6750 2018-01-08 Richard Biener <rguenther@suse.de>
6752 PR middle-end/83713
6753 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6755 2018-01-08 Richard Biener <rguenther@suse.de>
6757 PR tree-optimization/83685
6758 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6759 references to abnormals.
6761 2018-01-08 Richard Biener <rguenther@suse.de>
6763 PR lto/83719
6764 * dwarf2out.c (output_indirect_strings): Handle empty
6765 skeleton_debug_str_hash.
6766 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6768 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6770 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6771 (emit_store_direct): Likewise.
6772 (arc_trampoline_adjust_address): Likewise.
6773 (arc_asm_trampoline_template): New function.
6774 (arc_initialize_trampoline): Use asm_trampoline_template.
6775 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6776 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6777 * config/arc/arc.md (flush_icache): Delete pattern.
6779 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6781 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6782 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6783 munaligned-access.
6785 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6787 PR target/83681
6788 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6789 by not USED_FOR_TARGET.
6790 (make_pass_resolve_sw_modes): Likewise.
6792 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6794 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6795 USED_FOR_TARGET.
6797 2018-01-08 Richard Biener <rguenther@suse.de>
6799 PR middle-end/83580
6800 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6802 2018-01-08 Richard Biener <rguenther@suse.de>
6804 PR middle-end/83517
6805 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6807 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6809 PR middle-end/81897
6810 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6811 basic blocks with a small number of successors.
6812 (convert_control_dep_chain_into_preds): Improve handling of
6813 forwarder blocks.
6814 (dump_predicates): Split apart into...
6815 (dump_pred_chain): ...here...
6816 (dump_pred_info): ...and here.
6817 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6818 (can_chain_union_be_invalidated_p): Improve check for invalidation
6819 of paths.
6820 (uninit_uses_cannot_happen): Avoid unnecessary if
6821 convert_control_dep_chain_into_preds yielded nothing.
6823 2018-01-06 Martin Sebor <msebor@redhat.com>
6825 PR tree-optimization/83640
6826 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6827 subtracting negative offset from size.
6828 (builtin_access::overlap): Adjust offset bounds of the access to fall
6829 within the size of the object if possible.
6831 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6833 PR rtl-optimization/83699
6834 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6835 extract_bit_field_as_subreg to cases in which the extracted
6836 value is also a vector.
6838 * lra-constraints.c (process_alt_operands): Test for the equivalence
6839 substitutions when detecting a possible reload cycle.
6841 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6843 PR debug/83480
6844 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6845 by default if flag_selective_schedling{,2}. Formatting fixes.
6847 PR rtl-optimization/83682
6848 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6849 if it has non-VECTOR_MODE element mode.
6850 (vec_duplicate_p): Likewise.
6852 PR middle-end/83694
6853 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6854 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6856 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6858 PR target/83604
6859 * config/i386/i386-builtin.def
6860 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6861 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6862 Require also OPTION_MASK_ISA_AVX512F in addition to
6863 OPTION_MASK_ISA_GFNI.
6864 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6865 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6866 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6867 to OPTION_MASK_ISA_GFNI.
6868 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6869 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6870 OPTION_MASK_ISA_AVX512BW.
6871 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6872 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6873 addition to OPTION_MASK_ISA_GFNI.
6874 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6875 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6876 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6877 to OPTION_MASK_ISA_GFNI.
6878 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6879 a requirement for all ISAs rather than any of them with a few
6880 exceptions.
6881 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6882 processing.
6883 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6884 bitmasks to be enabled with 3 exceptions, instead of requiring any
6885 enabled ISA with lots of exceptions.
6886 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6887 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6888 Change avx512bw in isa attribute to avx512f.
6889 * config/i386/sgxintrin.h: Add license boilerplate.
6890 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6891 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6892 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6893 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6894 defined.
6895 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6896 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6897 temporarily sse2 rather than sse if not enabled already.
6899 PR target/83604
6900 * config/i386/sse.md (VI248_VLBW): Rename to ...
6901 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6902 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6903 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6904 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6905 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6906 mode iterator instead of VI248_VLBW.
6908 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6910 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6911 (record_modified): Skip clobbers; add debug output.
6912 (param_change_prob): Use sreal frequencies.
6914 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6916 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6917 punt for user-aligned variables.
6919 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6921 * tree-chrec.c (chrec_contains_symbols): Return true for
6922 POLY_INT_CST.
6924 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6926 PR target/82439
6927 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6928 of (x|y) == x for BICS pattern.
6930 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6932 PR tree-optimization/83605
6933 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6934 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6935 can throw.
6937 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6939 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6940 * config/epiphany/rtems.h: New file.
6942 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6943 Uros Bizjak <ubizjak@gmail.com>
6945 PR target/83554
6946 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6947 QIreg_operand instead of register_operand predicate.
6948 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6949 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6950 comments instead of -fmitigate[-_]rop.
6952 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6954 PR bootstrap/81926
6955 * cgraphunit.c (symbol_table::compile): Switch to text_section
6956 before calling assembly_start debug hook.
6957 * run-rtl-passes.c (run_rtl_passes): Likewise.
6958 Include output.h.
6960 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6962 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6963 range_int_cst_p rather than !symbolic_range_p before calling
6964 extract_range_from_multiplicative_op_1.
6966 2018-01-04 Jeff Law <law@redhat.com>
6968 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6969 redundant test in assertion.
6971 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6973 * doc/rtl.texi: Document machine_mode wrapper classes.
6975 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6977 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6978 using tree_to_uhwi.
6980 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6982 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6983 the VEC_PERM_EXPR fold to fail.
6985 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6987 PR debug/83585
6988 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6989 to switched_sections.
6991 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6993 PR target/83680
6994 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6995 test for d.testing.
6997 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6999 PR target/83387
7000 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7001 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7003 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7005 PR debug/83666
7006 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7007 is BLKmode and bitpos not zero or mode change is needed.
7009 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7011 PR target/83675
7012 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7013 TARGET_VIS2.
7015 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7017 PR target/83628
7018 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7019 instead of MULT rtx. Update all corresponding splitters.
7020 (*saddl_se): Ditto.
7021 (*ssub<modesuffix>): Ditto.
7022 (*ssubl_se): Ditto.
7023 (*cmp_sadd_di): Update split patterns.
7024 (*cmp_sadd_si): Ditto.
7025 (*cmp_sadd_sidi): Ditto.
7026 (*cmp_ssub_di): Ditto.
7027 (*cmp_ssub_si): Ditto.
7028 (*cmp_ssub_sidi): Ditto.
7029 * config/alpha/predicates.md (const23_operand): New predicate.
7030 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7031 Look for ASHIFT, not MULT inner operand.
7032 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7034 2018-01-04 Martin Liska <mliska@suse.cz>
7036 PR gcov-profile/83669
7037 * gcov.c (output_intermediate_file): Add version to intermediate
7038 gcov file.
7039 * doc/gcov.texi: Document new field 'version' in intermediate
7040 file format. Fix location of '-k' option of gcov command.
7042 2018-01-04 Martin Liska <mliska@suse.cz>
7044 PR ipa/82352
7045 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7047 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7049 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7051 2018-01-03 Martin Sebor <msebor@redhat.com>
7053 PR tree-optimization/83655
7054 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7055 checking calls with invalid arguments.
7057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7059 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7060 (vectorizable_mask_load_store): Delete.
7061 (vectorizable_call): Return false for masked loads and stores.
7062 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7063 instead of gimple_assign_rhs1.
7064 (vectorizable_load): Handle IFN_MASK_LOAD.
7065 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7069 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7070 split out from..,
7071 (vectorizable_mask_load_store): ...here.
7072 (vectorizable_load): ...and here.
7074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7076 * tree-vect-stmts.c (vect_build_all_ones_mask)
7077 (vect_build_zero_merge_argument): New functions, split out from...
7078 (vectorizable_load): ...here.
7080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7082 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7083 split out from...
7084 (vectorizable_mask_load_store): ...here.
7085 (vectorizable_store): ...and here.
7087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7089 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7090 split out from...
7091 (vectorizable_mask_load_store): ...here.
7093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7095 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7096 (vect_model_store_cost): Take a vec_load_store_type instead of a
7097 vect_def_type.
7098 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7099 (vect_model_store_cost): Take a vec_load_store_type instead of a
7100 vect_def_type.
7101 (vectorizable_mask_load_store): Update accordingly.
7102 (vectorizable_store): Likewise.
7103 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7107 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7108 IFN_MASK_LOAD calls here rather than...
7109 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7112 Alan Hayward <alan.hayward@arm.com>
7113 David Sherwood <david.sherwood@arm.com>
7115 * expmed.c (extract_bit_field_1): For vector extracts,
7116 fall back to extract_bit_field_as_subreg if vec_extract
7117 isn't available.
7119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7120 Alan Hayward <alan.hayward@arm.com>
7121 David Sherwood <david.sherwood@arm.com>
7123 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7124 they are variable or constant sized.
7125 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7126 slots for constant-sized data.
7128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7129 Alan Hayward <alan.hayward@arm.com>
7130 David Sherwood <david.sherwood@arm.com>
7132 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7133 handling COND_EXPRs with boolean comparisons, try to find a better
7134 basis for the mask type than the boolean itself.
7136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7138 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7139 is calculated and how it can be overridden.
7140 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7141 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7142 if defined.
7143 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7144 if nonzero.
7146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7147 Alan Hayward <alan.hayward@arm.com>
7148 David Sherwood <david.sherwood@arm.com>
7150 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7151 Remove the mode argument.
7152 (aarch64_simd_valid_immediate): Remove the mode and inverse
7153 arguments.
7154 * config/aarch64/iterators.md (bitsize): New iterator.
7155 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7156 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7157 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7158 aarch64_simd_valid_immediate.
7159 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7160 (aarch64_reg_or_bic_imm): Likewise.
7161 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7162 with an insn_type enum and msl with a modifier_type enum.
7163 Replace element_width with a scalar_mode. Change the shift
7164 to unsigned int. Add constructors for scalar_float_mode and
7165 scalar_int_mode elements.
7166 (aarch64_vect_float_const_representable_p): Delete.
7167 (aarch64_can_const_movi_rtx_p)
7168 (aarch64_simd_scalar_immediate_valid_for_move)
7169 (aarch64_simd_make_constant): Update call to
7170 aarch64_simd_valid_immediate.
7171 (aarch64_advsimd_valid_immediate_hs): New function.
7172 (aarch64_advsimd_valid_immediate): Likewise.
7173 (aarch64_simd_valid_immediate): Remove mode and inverse
7174 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7175 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7176 and aarch64_float_const_representable_p on the result.
7177 (aarch64_output_simd_mov_immediate): Remove mode argument.
7178 Update call to aarch64_simd_valid_immediate and use of
7179 simd_immediate_info.
7180 (aarch64_output_scalar_simd_mov_immediate): Update call
7181 accordingly.
7183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7184 Alan Hayward <alan.hayward@arm.com>
7185 David Sherwood <david.sherwood@arm.com>
7187 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7188 (mode_nunits): Likewise CONST_MODE_NUNITS.
7189 * machmode.def (ADJUST_NUNITS): Document.
7190 * genmodes.c (mode_data::need_nunits_adj): New field.
7191 (blank_mode): Update accordingly.
7192 (adj_nunits): New variable.
7193 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7194 parameter.
7195 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7196 listed in adj_nunits.
7197 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7198 listed in adj_nunits. Don't emit case statements for such modes.
7199 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7200 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7201 nothing if adj_nunits is nonnull.
7202 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7203 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7204 (emit_mode_fbit): Update use of print_maybe_const_decl.
7205 (emit_move_size): Likewise. Treat the array as non-const
7206 if adj_nunits.
7207 (emit_mode_adjustments): Handle adj_nunits.
7209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7211 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7212 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7213 (VECTOR_MODES): Use it.
7214 (make_vector_modes): Take the prefix as an argument.
7216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7217 Alan Hayward <alan.hayward@arm.com>
7218 David Sherwood <david.sherwood@arm.com>
7220 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7221 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7222 for MODE_VECTOR_BOOL.
7223 * machmode.def (VECTOR_BOOL_MODE): Document.
7224 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7225 (make_vector_bool_mode): New function.
7226 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7227 MODE_VECTOR_BOOL.
7228 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7229 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7230 Likewise.
7231 * stor-layout.c (int_mode_for_mode): Likewise.
7232 * tree.c (build_vector_type_for_mode): Likewise.
7233 * varasm.c (output_constant_pool_2): Likewise.
7234 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7235 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7236 for MODE_VECTOR_BOOL.
7237 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7238 of mode class checks.
7239 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7240 instead of a list of mode class checks.
7241 (expand_vector_scalar_condition): Likewise.
7242 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7245 Alan Hayward <alan.hayward@arm.com>
7246 David Sherwood <david.sherwood@arm.com>
7248 * machmode.h (mode_size): Change from unsigned short to
7249 poly_uint16_pod.
7250 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7251 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7252 or if measurement_type is not polynomial.
7253 (fixed_size_mode::includes_p): Check for constant-sized modes.
7254 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7255 return a poly_uint16 rather than an unsigned short.
7256 (emit_mode_size): Change the type of mode_size from unsigned short
7257 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7258 (emit_mode_adjustments): Cope with polynomial vector sizes.
7259 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7260 for GET_MODE_SIZE.
7261 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7262 for GET_MODE_SIZE.
7263 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7264 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7265 * caller-save.c (setup_save_areas): Likewise.
7266 (replace_reg_with_saved_mem): Likewise.
7267 * calls.c (emit_library_call_value_1): Likewise.
7268 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7269 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7270 (gen_lowpart_for_combine): Likewise.
7271 * convert.c (convert_to_integer_1): Likewise.
7272 * cse.c (equiv_constant, cse_insn): Likewise.
7273 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7274 (cselib_subst_to_values): Likewise.
7275 * dce.c (word_dce_process_block): Likewise.
7276 * df-problems.c (df_word_lr_mark_ref): Likewise.
7277 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7278 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7279 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7280 (rtl_for_decl_location): Likewise.
7281 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7282 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7283 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7284 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7285 (expand_expr_real_1): Likewise.
7286 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7287 (pad_below): Likewise.
7288 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7289 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7290 * ira.c (get_subreg_tracking_sizes): Likewise.
7291 * ira-build.c (ira_create_allocno_objects): Likewise.
7292 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7293 (ira_sort_regnos_for_alter_reg): Likewise.
7294 * ira-costs.c (record_operand_costs): Likewise.
7295 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7296 (resolve_simple_move): Likewise.
7297 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7298 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7299 (lra_constraints): Likewise.
7300 (CONST_POOL_OK_P): Reject variable-sized modes.
7301 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7302 (add_pseudo_to_slot, lra_spill): Likewise.
7303 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7304 * optabs-query.c (get_best_extraction_insn): Likewise.
7305 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7306 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7307 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7308 * recog.c (offsettable_address_addr_space_p): Likewise.
7309 * regcprop.c (maybe_mode_change): Likewise.
7310 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7311 * regrename.c (build_def_use): Likewise.
7312 * regstat.c (dump_reg_info): Likewise.
7313 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7314 (find_reloads, find_reloads_subreg_address): Likewise.
7315 * reload1.c (eliminate_regs_1): Likewise.
7316 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7317 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7318 (simplify_binary_operation_1, simplify_subreg): Likewise.
7319 * targhooks.c (default_function_arg_padding): Likewise.
7320 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7321 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7322 (verify_gimple_assign_ternary): Likewise.
7323 * tree-inline.c (estimate_move_cost): Likewise.
7324 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7325 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7326 (get_address_cost_ainc): Likewise.
7327 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7328 (vect_supportable_dr_alignment): Likewise.
7329 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7330 (vectorizable_reduction): Likewise.
7331 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7332 (vectorizable_operation, vectorizable_load): Likewise.
7333 * tree.c (build_same_sized_truth_vector_type): Likewise.
7334 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7335 * var-tracking.c (emit_note_insn_var_location): Likewise.
7336 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7337 (ADDR_VEC_ALIGN): Likewise.
7339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7340 Alan Hayward <alan.hayward@arm.com>
7341 David Sherwood <david.sherwood@arm.com>
7343 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7344 unsigned short.
7345 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7346 or if measurement_type is polynomial.
7347 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7348 * combine.c (make_extraction): Likewise.
7349 * dse.c (find_shift_sequence): Likewise.
7350 * dwarf2out.c (mem_loc_descriptor): Likewise.
7351 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7352 (extract_bit_field, extract_low_bits): Likewise.
7353 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7354 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7355 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7356 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7357 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7358 * reload.c (find_reloads): Likewise.
7359 * reload1.c (alter_reg): Likewise.
7360 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7361 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7362 * tree-if-conv.c (predicate_mem_writes): Likewise.
7363 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7364 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7365 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7366 * valtrack.c (dead_debug_insert_temp): Likewise.
7367 * varasm.c (mergeable_constant_section): Likewise.
7368 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7371 Alan Hayward <alan.hayward@arm.com>
7372 David Sherwood <david.sherwood@arm.com>
7374 * expr.c (expand_assignment): Cope with polynomial mode sizes
7375 when assigning to a CONCAT.
7377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7378 Alan Hayward <alan.hayward@arm.com>
7379 David Sherwood <david.sherwood@arm.com>
7381 * machmode.h (mode_precision): Change from unsigned short to
7382 poly_uint16_pod.
7383 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7384 short.
7385 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7386 or if measurement_type is not polynomial.
7387 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7388 in which the mode is already known to be a scalar_int_mode.
7389 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7390 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7391 initializer.
7392 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7393 for GET_MODE_PRECISION.
7394 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7395 for GET_MODE_PRECISION.
7396 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7397 as polynomial.
7398 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7399 (expand_field_assignment, make_extraction): Likewise.
7400 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7401 (get_last_value): Likewise.
7402 * convert.c (convert_to_integer_1): Likewise.
7403 * cse.c (cse_insn): Likewise.
7404 * expr.c (expand_expr_real_1): Likewise.
7405 * lra-constraints.c (simplify_operand_subreg): Likewise.
7406 * optabs-query.c (can_atomic_load_p): Likewise.
7407 * optabs.c (expand_atomic_load): Likewise.
7408 (expand_atomic_store): Likewise.
7409 * ree.c (combine_reaching_defs): Likewise.
7410 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7411 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7412 * tree.h (type_has_mode_precision_p): Likewise.
7413 * ubsan.c (instrument_si_overflow): Likewise.
7415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7416 Alan Hayward <alan.hayward@arm.com>
7417 David Sherwood <david.sherwood@arm.com>
7419 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7420 polynomial numbers of units.
7421 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7422 (valid_vector_subparts_p): New function.
7423 (build_vector_type): Remove temporary shim and take the number
7424 of units as a poly_uint64 rather than an int.
7425 (build_opaque_vector_type): Take the number of units as a
7426 poly_uint64 rather than an int.
7427 * tree.c (build_vector_from_ctor): Handle polynomial
7428 TYPE_VECTOR_SUBPARTS.
7429 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7430 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7431 (build_vector_from_val): If the number of units is variable,
7432 use build_vec_duplicate_cst for constant operands and
7433 VEC_DUPLICATE_EXPR otherwise.
7434 (make_vector_type): Remove temporary is_constant ().
7435 (build_vector_type, build_opaque_vector_type): Take the number of
7436 units as a poly_uint64 rather than an int.
7437 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7438 VECTOR_CST_NELTS.
7439 * cfgexpand.c (expand_debug_expr): Likewise.
7440 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7441 (store_constructor, expand_expr_real_1): Likewise.
7442 (const_scalar_mask_from_tree): Likewise.
7443 * fold-const-call.c (fold_const_reduction): Likewise.
7444 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7445 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7446 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7447 (fold_relational_const): Likewise.
7448 (native_interpret_vector): Likewise. Change the size from an
7449 int to an unsigned int.
7450 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7451 TYPE_VECTOR_SUBPARTS.
7452 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7453 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7454 duplicating a non-constant operand into a variable-length vector.
7455 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7456 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7457 * ipa-icf.c (sem_variable::equals): Likewise.
7458 * match.pd: Likewise.
7459 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7460 * print-tree.c (print_node): Likewise.
7461 * stor-layout.c (layout_type): Likewise.
7462 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7463 * tree-cfg.c (verify_gimple_comparison): Likewise.
7464 (verify_gimple_assign_binary): Likewise.
7465 (verify_gimple_assign_ternary): Likewise.
7466 (verify_gimple_assign_single): Likewise.
7467 * tree-pretty-print.c (dump_generic_node): Likewise.
7468 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7469 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7470 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7471 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7472 (vect_shift_permute_load_chain): Likewise.
7473 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7474 (expand_vector_condition, optimize_vector_constructor): Likewise.
7475 (lower_vec_perm, get_compute_type): Likewise.
7476 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7477 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7478 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7479 (vect_recog_mask_conversion_pattern): Likewise.
7480 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7481 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7482 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7483 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7484 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7485 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7486 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7487 (supportable_widening_operation): Likewise.
7488 (supportable_narrowing_operation): Likewise.
7489 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7490 Likewise.
7491 * varasm.c (output_constant): Likewise.
7493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7494 Alan Hayward <alan.hayward@arm.com>
7495 David Sherwood <david.sherwood@arm.com>
7497 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7498 so that both the length == 3 and length != 3 cases set up their
7499 own permute vectors. Add comments explaining why we know the
7500 number of elements is constant.
7501 (vect_permute_load_chain): Likewise.
7503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7504 Alan Hayward <alan.hayward@arm.com>
7505 David Sherwood <david.sherwood@arm.com>
7507 * machmode.h (mode_nunits): Change from unsigned char to
7508 poly_uint16_pod.
7509 (ONLY_FIXED_SIZE_MODES): New macro.
7510 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7511 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7512 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7513 New typedefs.
7514 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7515 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7516 or if measurement_type is not polynomial.
7517 * genmodes.c (ZERO_COEFFS): New macro.
7518 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7519 poly_uint16.
7520 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7521 Use ZERO_COEFFS when emitting initializers.
7522 * data-streamer.h (bp_pack_poly_value): New function.
7523 (bp_unpack_poly_value): Likewise.
7524 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7525 for GET_MODE_NUNITS.
7526 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7527 for GET_MODE_NUNITS.
7528 * tree.c (make_vector_type): Remove temporary shim and make
7529 the real function take the number of units as a poly_uint64
7530 rather than an int.
7531 (build_vector_type_for_mode): Handle polynomial nunits.
7532 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7533 * emit-rtl.c (const_vec_series_p_1): Likewise.
7534 (gen_rtx_CONST_VECTOR): Likewise.
7535 * fold-const.c (test_vec_duplicate_folding): Likewise.
7536 * genrecog.c (validate_pattern): Likewise.
7537 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7538 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7539 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7540 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7541 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7542 * rtlanal.c (subreg_get_info): Likewise.
7543 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7544 (vect_grouped_load_supported): Likewise.
7545 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7546 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7547 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7548 (simplify_const_unary_operation, simplify_binary_operation_1)
7549 (simplify_const_binary_operation, simplify_ternary_operation)
7550 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7551 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7552 instead of CONST_VECTOR_NUNITS.
7553 * varasm.c (output_constant_pool_2): Likewise.
7554 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7555 explicit-encoded elements in the XVEC for variable-length vectors.
7557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7559 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7562 Alan Hayward <alan.hayward@arm.com>
7563 David Sherwood <david.sherwood@arm.com>
7565 * coretypes.h (fixed_size_mode): Declare.
7566 (fixed_size_mode_pod): New typedef.
7567 * builtins.h (target_builtins::x_apply_args_mode)
7568 (target_builtins::x_apply_result_mode): Change type to
7569 fixed_size_mode_pod.
7570 * builtins.c (apply_args_size, apply_result_size, result_vector)
7571 (expand_builtin_apply_args_1, expand_builtin_apply)
7572 (expand_builtin_return): Update accordingly.
7574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7576 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7577 * cselib.c (cselib_hash_rtx): Likewise.
7578 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7579 CONST_VECTOR encoding.
7581 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7582 Jeff Law <law@redhat.com>
7584 PR target/83641
7585 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7586 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7587 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7588 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7590 PR target/83641
7591 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7592 explicitly probe *sp in a noreturn function if there were any callee
7593 register saves or frame pointer is needed.
7595 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7597 PR debug/83621
7598 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7599 BLKmode for ternary, binary or unary expressions.
7601 PR debug/83645
7602 * var-tracking.c (delete_vta_debug_insn): New inline function.
7603 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7604 insns from get_insns () to NULL instead of each bb separately.
7605 Use delete_vta_debug_insn. No longer static.
7606 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7607 delete_vta_debug_insns callers.
7608 * rtl.h (delete_vta_debug_insns): Declare.
7609 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7610 instead of variable_tracking_main.
7612 2018-01-03 Martin Sebor <msebor@redhat.com>
7614 PR tree-optimization/83603
7615 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7616 arguments past the endof the argument list in functions declared
7617 without a prototype.
7618 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7619 Avoid checking when arguments are null.
7621 2018-01-03 Martin Sebor <msebor@redhat.com>
7623 PR c/83559
7624 * doc/extend.texi (attribute const): Fix a typo.
7625 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7626 issuing -Wsuggest-attribute for void functions.
7628 2018-01-03 Martin Sebor <msebor@redhat.com>
7630 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7631 offset_int::from instead of wide_int::to_shwi.
7632 (maybe_diag_overlap): Remove assertion.
7633 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7634 * gimple-ssa-sprintf.c (format_directive): Same.
7635 (parse_directive): Same.
7636 (sprintf_dom_walker::compute_format_length): Same.
7637 (try_substitute_return_value): Same.
7639 2018-01-03 Jeff Law <law@redhat.com>
7641 PR middle-end/83654
7642 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7643 non-constant residual for zero at runtime and avoid probing in
7644 that case. Reorganize code for trailing problem to mirror handling
7645 of the residual.
7647 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7649 PR tree-optimization/83501
7650 * tree-ssa-strlen.c (get_string_cst): New.
7651 (handle_char_store): Call get_string_cst.
7653 2018-01-03 Martin Liska <mliska@suse.cz>
7655 PR tree-optimization/83593
7656 * tree-ssa-strlen.c: Include tree-cfg.h.
7657 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7658 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7659 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7660 to false.
7661 (strlen_dom_walker::before_dom_children): Call
7662 gimple_purge_dead_eh_edges. Dump tranformation with details
7663 dump flags.
7664 (strlen_dom_walker::before_dom_children): Update call by adding
7665 new argument cleanup_eh.
7666 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7668 2018-01-03 Martin Liska <mliska@suse.cz>
7670 PR ipa/83549
7671 * cif-code.def (VARIADIC_THUNK): New enum value.
7672 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7673 thunks.
7675 2018-01-03 Jan Beulich <jbeulich@suse.com>
7677 * sse.md (mov<mode>_internal): Tighten condition for when to use
7678 vmovdqu<ssescalarsize> for TI and OI modes.
7680 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7682 Update copyright years.
7684 2018-01-03 Martin Liska <mliska@suse.cz>
7686 PR ipa/83594
7687 * ipa-visibility.c (function_and_variable_visibility): Skip
7688 functions with noipa attribure.
7690 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7692 * gcc.c (process_command): Update copyright notice dates.
7693 * gcov-dump.c (print_version): Ditto.
7694 * gcov.c (print_version): Ditto.
7695 * gcov-tool.c (print_version): Ditto.
7696 * gengtype.c (create_file): Ditto.
7697 * doc/cpp.texi: Bump @copying's copyright year.
7698 * doc/cppinternals.texi: Ditto.
7699 * doc/gcc.texi: Ditto.
7700 * doc/gccint.texi: Ditto.
7701 * doc/gcov.texi: Ditto.
7702 * doc/install.texi: Ditto.
7703 * doc/invoke.texi: Ditto.
7705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7707 * vector-builder.h (vector_builder::m_full_nelts): Change from
7708 unsigned int to poly_uint64.
7709 (vector_builder::full_nelts): Update prototype accordingly.
7710 (vector_builder::new_vector): Likewise.
7711 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7712 (vector_builder::operator ==): Likewise.
7713 (vector_builder::finalize): Likewise.
7714 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7715 Take the number of elements as a poly_uint64 rather than an
7716 unsigned int.
7717 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7718 from unsigned int to poly_uint64.
7719 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7720 (vec_perm_indices::new_vector): Likewise.
7721 (vec_perm_indices::length): Likewise.
7722 (vec_perm_indices::nelts_per_input): Likewise.
7723 (vec_perm_indices::input_nelts): Likewise.
7724 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7725 number of elements per input as a poly_uint64 rather than an
7726 unsigned int. Use the original encoding for variable-length
7727 vectors, rather than clamping each individual element.
7728 For the second and subsequent elements in each pattern,
7729 clamp the step and base before clamping their sum.
7730 (vec_perm_indices::series_p): Handle polynomial element counts.
7731 (vec_perm_indices::all_in_range_p): Likewise.
7732 (vec_perm_indices_to_tree): Likewise.
7733 (vec_perm_indices_to_rtx): Likewise.
7734 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7735 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7736 (tree_vector_builder::new_binary_operation): Handle polynomial
7737 element counts. Return false if we need to know the number
7738 of elements at compile time.
7739 * fold-const.c (fold_vec_perm): Punt if the number of elements
7740 isn't known at compile time.
7742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7744 * vec-perm-indices.h (vec_perm_builder): Change element type
7745 from HOST_WIDE_INT to poly_int64.
7746 (vec_perm_indices::element_type): Update accordingly.
7747 (vec_perm_indices::clamp): Handle polynomial element_types.
7748 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7749 (vec_perm_indices::all_in_range_p): Likewise.
7750 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7751 than shwi trees.
7752 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7753 polynomial vec_perm_indices element types.
7754 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7755 * fold-const.c (fold_vec_perm): Likewise.
7756 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7757 * tree-vect-generic.c (lower_vec_perm): Likewise.
7758 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7759 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7760 element type to HOST_WIDE_INT.
7762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7763 Alan Hayward <alan.hayward@arm.com>
7764 David Sherwood <david.sherwood@arm.com>
7766 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7767 rather than an int. Use plus_constant.
7768 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7769 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7772 Alan Hayward <alan.hayward@arm.com>
7773 David Sherwood <david.sherwood@arm.com>
7775 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7776 a HOST_WIDE_INT to a poly_int64.
7778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7779 Alan Hayward <alan.hayward@arm.com>
7780 David Sherwood <david.sherwood@arm.com>
7782 * calls.c (load_register_parameters): Cope with polynomial
7783 mode sizes. Require a constant size for BLKmode parameters
7784 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7785 forces a parameter to be padded at the lsb end in order to
7786 fill a complete number of words, require the parameter size
7787 to be ordered wrt UNITS_PER_WORD.
7789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7790 Alan Hayward <alan.hayward@arm.com>
7791 David Sherwood <david.sherwood@arm.com>
7793 * reload1.c (spill_stack_slot_width): Change element type
7794 from unsigned int to poly_uint64_pod.
7795 (alter_reg): Treat mode sizes as polynomial.
7797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7798 Alan Hayward <alan.hayward@arm.com>
7799 David Sherwood <david.sherwood@arm.com>
7801 * reload.c (complex_word_subreg_p): New function.
7802 (reload_inner_reg_of_subreg, push_reload): Use it.
7804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7805 Alan Hayward <alan.hayward@arm.com>
7806 David Sherwood <david.sherwood@arm.com>
7808 * lra-constraints.c (process_alt_operands): Reject matched
7809 operands whose sizes aren't ordered.
7810 (match_reload): Refer to this check here.
7812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7813 Alan Hayward <alan.hayward@arm.com>
7814 David Sherwood <david.sherwood@arm.com>
7816 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7817 that the mode size is in the set {1, 2, 4, 8, 16}.
7819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7820 Alan Hayward <alan.hayward@arm.com>
7821 David Sherwood <david.sherwood@arm.com>
7823 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7824 Use plus_constant instead of gen_rtx_PLUS.
7826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7827 Alan Hayward <alan.hayward@arm.com>
7828 David Sherwood <david.sherwood@arm.com>
7830 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7831 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7832 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7833 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7834 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7835 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7836 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7837 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7838 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7839 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7840 a poly_int64.
7841 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7842 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7843 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7844 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7845 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7846 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7847 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7848 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7849 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7850 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7851 function.
7852 * expr.c (emit_move_resolve_push): Treat the input and result
7853 of PUSH_ROUNDING as a poly_int64.
7854 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7855 (emit_push_insn): Likewise.
7856 * lra-eliminations.c (mark_not_eliminable): Likewise.
7857 * recog.c (push_operand): Likewise.
7858 * reload1.c (elimination_effects): Likewise.
7859 * rtlanal.c (nonzero_bits1): Likewise.
7860 * calls.c (store_one_arg): Likewise. Require the padding to be
7861 known at compile time.
7863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7864 Alan Hayward <alan.hayward@arm.com>
7865 David Sherwood <david.sherwood@arm.com>
7867 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7868 Use plus_constant instead of gen_rtx_PLUS.
7870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7871 Alan Hayward <alan.hayward@arm.com>
7872 David Sherwood <david.sherwood@arm.com>
7874 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7875 rather than an int.
7877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7878 Alan Hayward <alan.hayward@arm.com>
7879 David Sherwood <david.sherwood@arm.com>
7881 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7882 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7883 via stack temporaries. Treat the mode size as polynomial too.
7885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7886 Alan Hayward <alan.hayward@arm.com>
7887 David Sherwood <david.sherwood@arm.com>
7889 * expr.c (expand_expr_real_2): When handling conversions involving
7890 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7891 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7892 as a poly_uint64 too.
7894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7895 Alan Hayward <alan.hayward@arm.com>
7896 David Sherwood <david.sherwood@arm.com>
7898 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7901 Alan Hayward <alan.hayward@arm.com>
7902 David Sherwood <david.sherwood@arm.com>
7904 * combine.c (can_change_dest_mode): Handle polynomial
7905 REGMODE_NATURAL_SIZE.
7906 * expmed.c (store_bit_field_1): Likewise.
7907 * expr.c (store_constructor): Likewise.
7908 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7909 and polynomial REGMODE_NATURAL_SIZE.
7910 (gen_lowpart_common): Likewise.
7911 * reginfo.c (record_subregs_of_mode): Likewise.
7912 * rtlanal.c (read_modify_subreg_p): Likewise.
7914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7915 Alan Hayward <alan.hayward@arm.com>
7916 David Sherwood <david.sherwood@arm.com>
7918 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7919 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
7927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7928 Alan Hayward <alan.hayward@arm.com>
7929 David Sherwood <david.sherwood@arm.com>
7931 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7932 in a POINTER_PLUS_EXPR.
7934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7935 Alan Hayward <alan.hayward@arm.com>
7936 David Sherwood <david.sherwood@arm.com>
7938 * omp-simd-clone.c (simd_clone_subparts): New function.
7939 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7940 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7947 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7948 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7949 (expand_vector_condition, vector_element): Likewise.
7950 (subparts_gt): New function.
7951 (get_compute_type): Use subparts_gt.
7952 (count_type_subparts): Delete.
7953 (expand_vector_operations_1): Use subparts_gt instead of
7954 count_type_subparts.
7956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7957 Alan Hayward <alan.hayward@arm.com>
7958 David Sherwood <david.sherwood@arm.com>
7960 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7961 (vect_compile_time_alias): ...this new function. Do the calculation
7962 on poly_ints rather than trees.
7963 (vect_prune_runtime_alias_test_list): Update call accordingly.
7965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7966 Alan Hayward <alan.hayward@arm.com>
7967 David Sherwood <david.sherwood@arm.com>
7969 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7970 numbers of units.
7971 (vect_schedule_slp_instance): Likewise.
7973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7974 Alan Hayward <alan.hayward@arm.com>
7975 David Sherwood <david.sherwood@arm.com>
7977 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7978 constant and extern definitions for variable-length vectors.
7979 (vect_get_constant_vectors): Note that the number of units
7980 is known to be constant.
7982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7983 Alan Hayward <alan.hayward@arm.com>
7984 David Sherwood <david.sherwood@arm.com>
7986 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7987 of units as polynomial. Choose between WIDE and NARROW based
7988 on multiple_p.
7990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7991 Alan Hayward <alan.hayward@arm.com>
7992 David Sherwood <david.sherwood@arm.com>
7994 * tree-vect-stmts.c (simd_clone_subparts): New function.
7995 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7998 Alan Hayward <alan.hayward@arm.com>
7999 David Sherwood <david.sherwood@arm.com>
8001 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8002 vectors as polynomial. Use build_index_vector for
8003 IFN_GOMP_SIMD_LANE.
8005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8006 Alan Hayward <alan.hayward@arm.com>
8007 David Sherwood <david.sherwood@arm.com>
8009 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8010 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8011 for variable-length vectors.
8012 (vectorizable_mask_load_store): Treat the number of units as
8013 polynomial, asserting that it is constant if the condition has
8014 already been enforced.
8015 (vectorizable_store, vectorizable_load): Likewise.
8017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8018 Alan Hayward <alan.hayward@arm.com>
8019 David Sherwood <david.sherwood@arm.com>
8021 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8022 of units as polynomial. Punt if we can't tell at compile time
8023 which vector contains the final result.
8025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8026 Alan Hayward <alan.hayward@arm.com>
8027 David Sherwood <david.sherwood@arm.com>
8029 * tree-vect-loop.c (vectorizable_induction): Treat the number
8030 of units as polynomial. Punt on SLP inductions. Use an integer
8031 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8032 cast of such a series for variable-length floating-point
8033 reductions.
8035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8036 Alan Hayward <alan.hayward@arm.com>
8037 David Sherwood <david.sherwood@arm.com>
8039 * tree.h (build_index_vector): Declare.
8040 * tree.c (build_index_vector): New function.
8041 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8042 of units as polynomial, forcibly converting it to a constant if
8043 vectorizable_reduction has already enforced the condition.
8044 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8045 to create a {1,2,3,...} vector.
8046 (vectorizable_reduction): Treat the number of units as polynomial.
8047 Choose vectype_in based on the largest scalar element size rather
8048 than the smallest number of units. Enforce the restrictions
8049 relied on above.
8051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8052 Alan Hayward <alan.hayward@arm.com>
8053 David Sherwood <david.sherwood@arm.com>
8055 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8056 number of units as polynomial.
8058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8059 Alan Hayward <alan.hayward@arm.com>
8060 David Sherwood <david.sherwood@arm.com>
8062 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8063 * target.def (autovectorize_vector_sizes): Return the vector sizes
8064 by pointer, using vector_sizes rather than a bitmask.
8065 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8066 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8067 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8068 Likewise.
8069 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8070 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8071 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8072 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8073 * omp-general.c (omp_max_vf): Likewise.
8074 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8075 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8076 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8077 * tree-vect-slp.c (vect_slp_bb): Likewise.
8078 * doc/tm.texi: Regenerate.
8079 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8080 to a poly_uint64.
8081 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8082 the vector size as a poly_uint64 rather than an unsigned int.
8083 (current_vector_size): Change from an unsigned int to a poly_uint64.
8084 (get_vectype_for_scalar_type): Update accordingly.
8085 * tree.h (build_truth_vector_type): Take the size and number of
8086 units as a poly_uint64 rather than an unsigned int.
8087 (build_vector_type): Add a temporary overload that takes
8088 the number of units as a poly_uint64 rather than an unsigned int.
8089 * tree.c (make_vector_type): Likewise.
8090 (build_truth_vector_type): Take the number of units as a poly_uint64
8091 rather than an unsigned int.
8093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8094 Alan Hayward <alan.hayward@arm.com>
8095 David Sherwood <david.sherwood@arm.com>
8097 * target.def (get_mask_mode): Take the number of units and length
8098 as poly_uint64s rather than unsigned ints.
8099 * targhooks.h (default_get_mask_mode): Update accordingly.
8100 * targhooks.c (default_get_mask_mode): Likewise.
8101 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8102 * doc/tm.texi: Regenerate.
8104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8105 Alan Hayward <alan.hayward@arm.com>
8106 David Sherwood <david.sherwood@arm.com>
8108 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8109 * omp-general.c (omp_max_vf): Likewise.
8110 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8111 (expand_omp_simd): Handle polynomial safelen.
8112 * omp-low.c (omplow_simd_context): Add a default constructor.
8113 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8114 (lower_rec_simd_input_clauses): Update accordingly.
8115 (lower_rec_input_clauses): Likewise.
8117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8118 Alan Hayward <alan.hayward@arm.com>
8119 David Sherwood <david.sherwood@arm.com>
8121 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8122 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8123 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8124 (vect_analyze_slp_cost): Likewise.
8125 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8126 (vect_model_load_cost): Likewise.
8128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8129 Alan Hayward <alan.hayward@arm.com>
8130 David Sherwood <david.sherwood@arm.com>
8132 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8133 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8134 from an unsigned int * to a poly_uint64_pod *.
8135 (calculate_unrolling_factor): New function.
8136 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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 (_slp_instance::unrolling_factor): Change
8143 from an unsigned int to a poly_uint64.
8144 (_loop_vec_info::slp_unrolling_factor): Likewise.
8145 (_loop_vec_info::vectorization_factor): Change from an int
8146 to a poly_uint64.
8147 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8148 (vect_get_num_vectors): New function.
8149 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8150 (vect_get_num_copies): Use vect_get_num_vectors.
8151 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8152 to an unsigned int *.
8153 (vect_analyze_data_refs): Change min_vf from an int * to a
8154 poly_uint64 *.
8155 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8156 than an unsigned HOST_WIDE_INT.
8157 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8158 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8159 to an unsigned int *.
8160 (vect_analyze_data_ref_dependences): Likewise.
8161 (vect_compute_data_ref_alignment): Handle polynomial vf.
8162 (vect_enhance_data_refs_alignment): Likewise.
8163 (vect_prune_runtime_alias_test_list): Likewise.
8164 (vect_shift_permute_load_chain): Likewise.
8165 (vect_supportable_dr_alignment): Likewise.
8166 (dependence_distance_ge_vf): Take the vectorization factor as a
8167 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8168 (vect_analyze_data_refs): Change min_vf from an int * to a
8169 poly_uint64 *.
8170 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8171 vfm1 as a poly_uint64 rather than an int. Make the same change
8172 for the returned bound_scalar.
8173 (vect_gen_vector_loop_niters): Handle polynomial vf.
8174 (vect_do_peeling): Likewise. Update call to
8175 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8176 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8177 be constant.
8178 * tree-vect-loop.c (vect_determine_vectorization_factor)
8179 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8180 (vect_get_known_peeling_cost): Likewise.
8181 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8182 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8183 (vect_transform_loop): Likewise. Use the lowest possible VF when
8184 updating the upper bounds of the loop.
8185 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8186 rather than an int.
8187 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8188 polynomial unroll factors.
8189 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8190 (vect_make_slp_decision): Likewise.
8191 (vect_supported_load_permutation_p): Likewise, and polynomial
8192 vf too.
8193 (vect_analyze_slp_cost): Handle polynomial vf.
8194 (vect_slp_analyze_node_operations): Likewise.
8195 (vect_slp_analyze_bb_1): Likewise.
8196 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8197 than an unsigned HOST_WIDE_INT.
8198 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8199 (vectorizable_load): Handle polynomial vf.
8200 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8201 a poly_uint64.
8202 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8205 Alan Hayward <alan.hayward@arm.com>
8206 David Sherwood <david.sherwood@arm.com>
8208 * match.pd: Handle bit operations involving three constants
8209 and try to fold one pair.
8211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8213 * tree-vect-loop-manip.c: Include gimple-fold.h.
8214 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8215 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8216 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8217 Add a path that uses a step of VF instead of 1, but disable it
8218 for now.
8219 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8220 and niters_no_overflow parameters. Update calls to
8221 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8222 Create a new SSA name if the latter choses to use a ste other
8223 than zero, and return it via niters_vector_mult_vf_var.
8224 * tree-vect-loop.c (vect_transform_loop): Update calls to
8225 vect_do_peeling, vect_gen_vector_loop_niters and
8226 slpeel_make_loop_iterate_ntimes.
8227 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8228 (vect_gen_vector_loop_niters): Update declarations after above changes.
8230 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8232 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8233 128-bit round to integer instructions.
8234 (ceil<mode>2): Likewise.
8235 (btrunc<mode>2): Likewise.
8236 (round<mode>2): Likewise.
8238 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8240 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8241 unaligned VSX load/store on P8/P9.
8242 (expand_block_clear): Allow the use of unaligned VSX
8243 load/store on P8/P9.
8245 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8247 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8248 New function.
8249 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8250 swap associated with both a load and a store.
8252 2018-01-02 Andrew Waterman <andrew@sifive.com>
8254 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8255 * config/riscv/riscv.md (clear_cache): Use it.
8257 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8259 * web.c: Remove out-of-date comment.
8261 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8263 * expr.c (fixup_args_size_notes): Check that any existing
8264 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8265 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8266 (emit_single_push_insn): ...here.
8268 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8270 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8271 (const_vector_encoded_nelts): New function.
8272 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8273 (const_vector_int_elt, const_vector_elt): Declare.
8274 * emit-rtl.c (const_vector_int_elt_1): New function.
8275 (const_vector_elt): Likewise.
8276 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8277 of CONST_VECTOR_ELT.
8279 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8281 * expr.c: Include rtx-vector-builder.h.
8282 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8283 directly on the tree encoding.
8284 (const_vector_from_tree): Likewise.
8285 * optabs.c: Include rtx-vector-builder.h.
8286 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8287 sequence of "u" values.
8288 * vec-perm-indices.c: Include rtx-vector-builder.h.
8289 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8290 directly on the vec_perm_indices encoding.
8292 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8294 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8295 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8296 * rtx-vector-builder.h: New file.
8297 * rtx-vector-builder.c: Likewise.
8298 * rtl.h (rtx_def::u2): Add a const_vector field.
8299 (CONST_VECTOR_NPATTERNS): New macro.
8300 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8301 (CONST_VECTOR_DUPLICATE_P): Likewise.
8302 (CONST_VECTOR_STEPPED_P): Likewise.
8303 (CONST_VECTOR_ENCODED_ELT): Likewise.
8304 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8305 (unwrap_const_vec_duplicate): Likewise.
8306 (const_vec_series_p): Check for a non-duplicated vector encoding.
8307 Say that the function only returns true for integer vectors.
8308 * emit-rtl.c: Include rtx-vector-builder.h.
8309 (gen_const_vec_duplicate_1): Delete.
8310 (gen_const_vector): Call gen_const_vec_duplicate instead of
8311 gen_const_vec_duplicate_1.
8312 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8313 (gen_const_vec_duplicate): Use rtx_vector_builder.
8314 (gen_const_vec_series): Likewise.
8315 (gen_rtx_CONST_VECTOR): Likewise.
8316 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8317 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8318 Build a new vector rather than modifying a CONST_VECTOR in-place.
8319 (handle_special_swappables): Update call accordingly.
8320 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8321 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8322 Build a new vector rather than modifying a CONST_VECTOR in-place.
8323 (handle_special_swappables): Update call accordingly.
8325 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8327 * simplify-rtx.c (simplify_const_binary_operation): Use
8328 CONST_VECTOR_ELT instead of XVECEXP.
8330 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8332 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8333 the selector elements to be different from the data elements
8334 if the selector is a VECTOR_CST.
8335 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8336 ssizetype for the selector.
8338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8340 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8341 before testing each element individually.
8342 * tree-vect-generic.c (lower_vec_perm): Likewise.
8344 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8346 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8347 * selftest-run-tests.c (selftest::run_tests): Call it.
8348 * vector-builder.h (vector_builder::operator ==): New function.
8349 (vector_builder::operator !=): Likewise.
8350 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8351 (vec_perm_indices::all_from_input_p): New function.
8352 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8353 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8354 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8355 instead of reading the VECTOR_CST directly. Detect whether both
8356 vector inputs are the same before constructing the vec_perm_indices,
8357 and update the number of inputs argument accordingly. Use the
8358 utility functions added above. Only construct sel2 if we need to.
8360 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8362 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8363 the broadcast of the low byte.
8364 (expand_mult_highpart): Use an explicit encoding for the permutes.
8365 * optabs-query.c (can_mult_highpart_p): Likewise.
8366 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8367 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8368 (vectorizable_bswap): Likewise.
8369 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8370 explicit encoding for the power-of-2 permutes.
8371 (vect_permute_store_chain): Likewise.
8372 (vect_grouped_load_supported): Likewise.
8373 (vect_permute_load_chain): Likewise.
8375 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8377 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8378 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8379 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8380 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8381 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8382 (vect_gen_perm_mask_any): Likewise.
8384 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8386 * int-vector-builder.h: New file.
8387 * vec-perm-indices.h: Include int-vector-builder.h.
8388 (vec_perm_indices): Redefine as an int_vector_builder.
8389 (auto_vec_perm_indices): Delete.
8390 (vec_perm_builder): Redefine as a stand-alone class.
8391 (vec_perm_indices::vec_perm_indices): New function.
8392 (vec_perm_indices::clamp): Likewise.
8393 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8394 (vec_perm_indices::new_vector): New function.
8395 (vec_perm_indices::new_expanded_vector): Update for new
8396 vec_perm_indices class.
8397 (vec_perm_indices::rotate_inputs): New function.
8398 (vec_perm_indices::all_in_range_p): Operate directly on the
8399 encoded form, without computing elided elements.
8400 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8401 encoding. Update for new vec_perm_indices class.
8402 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8403 the given vec_perm_builder.
8404 (expand_vec_perm_var): Update vec_perm_builder constructor.
8405 (expand_mult_highpart): Use vec_perm_builder instead of
8406 auto_vec_perm_indices.
8407 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8408 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8409 or double series encoding as appropriate.
8410 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8411 vec_perm_indices instead of auto_vec_perm_indices.
8412 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8413 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8414 (vect_permute_store_chain): Likewise.
8415 (vect_grouped_load_supported): Likewise.
8416 (vect_permute_load_chain): Likewise.
8417 (vect_shift_permute_load_chain): Likewise.
8418 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8419 (vect_transform_slp_perm_load): Likewise.
8420 (vect_schedule_slp_instance): Likewise.
8421 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8422 (vectorizable_mask_load_store): Likewise.
8423 (vectorizable_bswap): Likewise.
8424 (vectorizable_store): Likewise.
8425 (vectorizable_load): Likewise.
8426 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8427 vec_perm_indices instead of auto_vec_perm_indices. Use
8428 tree_to_vec_perm_builder to read the vector from a tree.
8429 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8430 vec_perm_builder instead of a vec_perm_indices.
8431 (have_whole_vector_shift): Use vec_perm_builder and
8432 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8433 truncation to calc_vec_perm_mask_for_shift.
8434 (vect_create_epilog_for_reduction): Likewise.
8435 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8436 from auto_vec_perm_indices to vec_perm_indices.
8437 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8438 instead of changing individual elements.
8439 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8440 the vector in d.perm.
8441 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8442 from auto_vec_perm_indices to vec_perm_indices.
8443 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8444 instead of changing individual elements.
8445 (arm_vectorize_vec_perm_const): Use new_vector to install
8446 the vector in d.perm.
8447 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8448 Update vec_perm_builder constructor.
8449 (rs6000_expand_interleave): Likewise.
8450 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8451 (rs6000_expand_interleave): Likewise.
8453 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8455 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8456 to qimode could truncate the indices.
8457 * optabs.c (expand_vec_perm_var): Likewise.
8459 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8461 * Makefile.in (OBJS): Add vec-perm-indices.o.
8462 * vec-perm-indices.h: New file.
8463 * vec-perm-indices.c: Likewise.
8464 * target.h (vec_perm_indices): Replace with a forward class
8465 declaration.
8466 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8467 * optabs.h: Include vec-perm-indices.h.
8468 (expand_vec_perm): Delete.
8469 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8470 (expand_vec_perm_const): Declare.
8471 * target.def (vec_perm_const_ok): Replace with...
8472 (vec_perm_const): ...this new hook.
8473 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8474 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8475 * doc/tm.texi: Regenerate.
8476 * optabs.def (vec_perm_const): Delete.
8477 * doc/md.texi (vec_perm_const): Likewise.
8478 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8479 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8480 expand_vec_perm for constant permutation vectors. Assert that
8481 the mode of variable permutation vectors is the integer equivalent
8482 of the mode that is being permuted.
8483 * optabs-query.h (selector_fits_mode_p): Declare.
8484 * optabs-query.c: Include vec-perm-indices.h.
8485 (selector_fits_mode_p): New function.
8486 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8487 is defined, instead of checking whether the vec_perm_const_optab
8488 exists. Use targetm.vectorize.vec_perm_const instead of
8489 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8490 fit in the vector mode before using a variable permute.
8491 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8492 vec_perm_indices instead of an rtx.
8493 (expand_vec_perm): Replace with...
8494 (expand_vec_perm_const): ...this new function. Take the selector
8495 as a vec_perm_indices rather than an rtx. Also take the mode of
8496 the selector. Update call to shift_amt_for_vec_perm_mask.
8497 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8498 Use vec_perm_indices::new_expanded_vector to expand the original
8499 selector into bytes. Check whether the indices fit in the vector
8500 mode before using a variable permute.
8501 (expand_vec_perm_var): Make global.
8502 (expand_mult_highpart): Use expand_vec_perm_const.
8503 * fold-const.c: Includes vec-perm-indices.h.
8504 * tree-ssa-forwprop.c: Likewise.
8505 * tree-vect-data-refs.c: Likewise.
8506 * tree-vect-generic.c: Likewise.
8507 * tree-vect-loop.c: Likewise.
8508 * tree-vect-slp.c: Likewise.
8509 * tree-vect-stmts.c: Likewise.
8510 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8511 Delete.
8512 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8513 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8514 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8515 (aarch64_vectorize_vec_perm_const): ...this new function.
8516 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8517 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8518 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8519 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8520 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8521 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8522 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8523 into...
8524 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8525 check for NEON modes.
8526 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8527 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8528 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8529 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8530 into...
8531 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8532 the old VEC_PERM_CONST conditions.
8533 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8534 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8535 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8536 (ia64_vectorize_vec_perm_const_ok): Merge into...
8537 (ia64_vectorize_vec_perm_const): ...this new function.
8538 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8539 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8540 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8541 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8542 * config/mips/mips.c (mips_expand_vec_perm_const)
8543 (mips_vectorize_vec_perm_const_ok): Merge into...
8544 (mips_vectorize_vec_perm_const): ...this new function.
8545 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8546 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8547 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8548 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8549 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8550 (rs6000_expand_vec_perm_const): Delete.
8551 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8552 Delete.
8553 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8554 (altivec_expand_vec_perm_const_le): Take each operand individually.
8555 Operate on constant selectors rather than rtxes.
8556 (altivec_expand_vec_perm_const): Likewise. Update call to
8557 altivec_expand_vec_perm_const_le.
8558 (rs6000_expand_vec_perm_const): Delete.
8559 (rs6000_vectorize_vec_perm_const_ok): Delete.
8560 (rs6000_vectorize_vec_perm_const): New function.
8561 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8562 an element count and rtx array.
8563 (rs6000_expand_extract_even): Update call accordingly.
8564 (rs6000_expand_interleave): Likewise.
8565 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8566 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8567 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8568 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8569 (rs6000_expand_vec_perm_const): Delete.
8570 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8571 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8572 (altivec_expand_vec_perm_const_le): Take each operand individually.
8573 Operate on constant selectors rather than rtxes.
8574 (altivec_expand_vec_perm_const): Likewise. Update call to
8575 altivec_expand_vec_perm_const_le.
8576 (rs6000_expand_vec_perm_const): Delete.
8577 (rs6000_vectorize_vec_perm_const_ok): Delete.
8578 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8579 reference to the SPE evmerge intructions.
8580 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8581 an element count and rtx array.
8582 (rs6000_expand_extract_even): Update call accordingly.
8583 (rs6000_expand_interleave): Likewise.
8584 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8585 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8586 new function.
8587 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8589 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8591 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8592 vector mode and that that mode matches the mode of the data
8593 being permuted.
8594 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8595 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8596 directly using expand_vec_perm_1 when forcing selectors into
8597 registers.
8598 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8600 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8602 * optabs-query.h (can_vec_perm_p): Delete.
8603 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8604 * optabs-query.c (can_vec_perm_p): Split into...
8605 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8606 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8607 particular selector is valid.
8608 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8609 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8610 (vect_grouped_load_supported): Likewise.
8611 (vect_shift_permute_load_chain): Likewise.
8612 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8613 (vect_transform_slp_perm_load): Likewise.
8614 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8615 (vectorizable_bswap): Likewise.
8616 (vect_gen_perm_mask_checked): Likewise.
8617 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8618 implementations of variable permutation vectors into account
8619 when deciding which selector to use.
8620 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8621 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8622 with a false third argument.
8623 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8624 to test whether the constant selector is valid and can_vec_perm_var_p
8625 to test whether a variable selector is valid.
8627 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8629 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8630 * optabs-query.c (can_vec_perm_p): Likewise.
8631 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8632 instead of vec_perm_indices.
8633 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8634 (vect_gen_perm_mask_checked): Likewise,
8635 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8636 (vect_gen_perm_mask_checked): Likewise,
8638 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8640 * optabs-query.h (qimode_for_vec_perm): Declare.
8641 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8642 (qimode_for_vec_perm): ...this new function.
8643 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8645 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8647 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8648 does not have a conditional at the top.
8650 2018-01-02 Richard Biener <rguenther@suse.de>
8652 * ipa-inline.c (big_speedup_p): Fix expression.
8654 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8656 PR target/81616
8657 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8658 for generic 4->6.
8660 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8662 PR target/81616
8663 Generic tuning.
8664 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8665 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8666 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8667 cond_taken_branch_cost 3->4.
8669 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8671 PR tree-optimization/83581
8672 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8673 TODO_cleanup_cfg if any changes have been made.
8675 PR middle-end/83608
8676 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8677 convert_modes if target mode has the right side, but different mode
8678 class.
8680 PR middle-end/83609
8681 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8682 last argument when extracting from CONCAT. If either from_real or
8683 from_imag is NULL, use expansion through memory. If result is not
8684 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8685 the parts directly to inner mode, if even that fails, use expansion
8686 through memory.
8688 PR middle-end/83623
8689 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8690 check for bswap in mode rather than HImode and use that in expand_unop
8691 too.
8692 \f
8693 Copyright (C) 2018 Free Software Foundation, Inc.
8695 Copying and distribution of this file, with or without modification,
8696 are permitted in any medium without royalty provided the copyright
8697 notice and this notice are preserved.