1 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3 PR rtl-optimization/84169
4 PR rtl-optimization/84780
5 * combine.c (can_combine_p): Check for a 2-insn combination whether
6 the destination register is used between the two insns, too.
8 2018-03-12 Richard Biener <rguenther@suse.de>
10 PR tree-optimization/84803
11 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
12 for refs DR analysis didn't process.
14 2018-03-12 Richard Biener <rguenther@suse.de>
16 PR tree-optimization/84777
17 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
18 force-vectorize loops ignore whether we are optimizing for size.
20 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
22 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
23 (TARGET_MD_ASM_ADJUST): Define.
25 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
26 Kito Cheng <kito.cheng@gmail.com>
27 Chung-Ju Wu <jasonwucj@gmail.com>
29 * config/nds32/nds32.c (nds32_compute_stack_frame,
30 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
31 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
32 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
33 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
34 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
35 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
36 * config/nds32/nds32.md (prologue, epilogue): Use macro
37 NDS32_V3PUSH_AVAILABLE_P to do checking.
39 2018-03-11 Jakub Jelinek <jakub@redhat.com>
42 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
43 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
44 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
45 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
46 addition of most attributes on !orig_type_die or the attribute not
47 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
49 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
50 Chung-Ju Wu <jasonwucj@gmail.com>
52 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
54 * config/nds32/nds32.opt (mvh): New option.
56 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
57 Chung-Ju Wu <jasonwucj@gmail.com>
59 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
61 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
62 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
65 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
66 Chung-Ju Wu <jasonwucj@gmail.com>
68 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
70 * config/nds32/nds32-multiple.md (strlensi): New pattern.
71 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
73 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
74 Kito Cheng <kito.cheng@gmail.com>
75 Chung-Ju Wu <jasonwucj@gmail.com>
77 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
78 UNSPEC_FFMISM and UNSPEC_FLMISM.
79 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
80 for ffb, ffmism and flmism.
81 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
82 (unspec_ffmism): Ditto.
83 (unspec_flmism): Ditto.
84 (nds32_expand_builtin_impl): Check if string extension is available.
85 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
86 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
88 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
91 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
94 * lra-assigns.c (assign_by_spills): Return a flag of reload
95 assignment failure. Do not process the reload assignment
96 failures. Do not spill other reload pseudos if they has the same
98 (lra_assign): Add a return arg. Set up from the result of
99 assign_by_spills call.
100 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
101 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
102 usage_insns if it is not NULL.
103 (spill_hard_reg_in_range): New function.
104 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
105 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
107 (lra_assign): Change prototype.
108 * lra.c (lra): Add code to deal with fails by splitting hard reg
111 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
114 * config/i386/i386.opt: Replace Enforcment with Enforcement.
116 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
119 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
120 (dw_val_node): Add val_symbolic_view.
121 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
122 (symview_upper_bound): New.
123 (new_line_info_table): Initialize symviews_since_reset.
124 (dwarf2out_source_line): Count symviews_since_reset and set
126 (dw_val_equal_p): Handle symview.
127 (add_AT_symview): New.
128 (print_dw_val): Handle symview.
129 (attr_checksum, attr_checksum_ordered): Likewise.
130 (same_dw_val_p, size_of_die): Likewise.
131 (value_format, output_die): Likewise.
132 (add_high_low_attributes): Use add_AT_symview for entry_view.
133 (dwarf2out_finish): Reset symview_upper_bound, clear
136 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
139 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
140 Add strict argument and use it.
141 (rs6000_split_multireg_move): Update for new strict argument.
142 (mem_operand_gpr): Disallow all non-offsettable addresses.
143 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
145 2018-03-09 Jakub Jelinek <jakub@redhat.com>
148 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
149 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
150 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
153 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
154 decl, use remap_type if we want to use the type.
156 2018-03-09 Martin Sebor <msebor@redhat.com>
158 PR tree-optimization/84526
159 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
161 (builtin_access::generic_overlap): Be prepared to handle non-array
164 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
166 PR rtl-optimization/84682
167 * lra-constraints.c (process_address_1): Check is_address flag
168 for address constraints.
169 (process_alt_operands): Likewise.
170 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
171 preprocess_constraints.
172 * recog.h (preprocess_constraints): Add oploc parameter.
174 * recog.c (preprocess_constraints): Test address_operand for
175 CT_ADDRESS constraints.
177 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
180 * lra-assigns.c (assign_by_spills): Return a flag of reload
181 assignment failure. Do not process the reload assignment
182 failures. Do not spill other reload pseudos if they has the same
184 (lra_assign): Add a return arg. Set up from the result of
185 assign_by_spills call.
186 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
187 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
188 usage_insns if it is not NULL.
189 (spill_hard_reg_in_range): New function.
190 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
191 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
193 (lra_assign): Change prototype.
194 * lra.c (lra): Add code to deal with fails by splitting hard reg
197 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
200 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
201 Accept complain bool parameter. Only emit errors if it is true.
202 (arm_parse_cpu_option_name): Likewise.
203 (arm_target_thumb_only): Adjust callers of the above.
204 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
205 prototype to take a default true bool parameter.
206 (arm_parse_arch_option_name): Likewise.
208 2018-03-09 David Malcolm <dmalcolm@redhat.com>
209 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
213 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
214 * configure: Regenerate.
215 * configure.ac ("linker --version-script option"): New.
216 ("linker soname option"): New.
218 2018-03-09 Richard Biener <rguenther@suse.de>
220 PR tree-optimization/84775
221 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
222 immediate uses of predicate stmts and mark them modified.
225 PR tree-optimization/84178
226 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
228 (version_loop_for_if_conversion): Delay update_ssa call.
229 (tree_if_conversion): Delay update_ssa until after predicate
232 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
235 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
236 when the function accesses prior frames.
238 2018-03-08 Jakub Jelinek <jakub@redhat.com>
241 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
242 gen_llsym, otherwise call maybe_gen_llsym.
245 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
246 has ',' character inside of it.
248 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
251 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
252 as clobbering CC_REGNUM.
254 2018-03-08 Richard Biener <rguenther@suse.de>
257 * tree-scalar-evolution.c: Include tree-into-ssa.h.
258 (follow_copies_to_constant): Do not follow SSA names registered
261 2018-03-08 Richard Biener <rguenther@suse.de>
263 PR tree-optimization/84178
264 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
266 (version_loop_for_if_conversion): Delay update_ssa call.
267 (tree_if_conversion): Delay update_ssa until after predicate
270 2018-03-08 David Malcolm <dmalcolm@redhat.com>
272 PR tree-optimization/84178
273 * tree-if-conv.c (release_bb_predicate): Remove the
274 the assertion that the stmts have NULL use_ops.
275 Discard the statements, asserting that they haven't
276 yet been added to a BB.
278 2018-03-08 Richard Biener <rguenther@suse.de>
280 PR tree-optimization/84746
281 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
282 (phi_translate): Pass in destination ANTIC_OUT set.
283 (phi_translate_1): Likewise. For a simplified result lookup
284 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
285 (phi_translate_set): Adjust.
286 (do_pre_regular_insertion): Likewise.
287 (do_pre_partial_partial_insertion): Likewise.
289 2018-03-08 Martin Liska <mliska@suse.cz>
291 PR gcov-profile/84735
292 * doc/gcov.texi: Document usage of profile files.
293 * gcov-io.h: Document changes in the format.
295 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
299 * dwarf2out.c (struct dw_line_info_table): Update comments for
301 (FORCE_RESET_NEXT_VIEW): New.
302 (FORCE_RESETTING_VIEW_P): New.
303 (RESETTING_VIEW_P): Check for -1 too.
304 (ZERO_VIEW_P): Likewise.
305 (new_line_info_table): Force-reset next view.
306 (dwarf2out_begin_function): Likewise.
307 (dwarf2out_source_line): Simplify zero_view_p initialization.
308 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
309 view directly. Omit view when omitting .loc at line 0.
311 2018-03-08 Jakub Jelinek <jakub@redhat.com>
313 PR tree-optimization/84740
314 * tree-switch-conversion.c (process_switch): Call build_constructors
315 only if info.phi_count is non-zero.
317 PR tree-optimization/84739
318 * tree-tailcall.c (find_tail_calls): Check call arguments against
319 DECL_ARGUMENTS (current_function_decl) rather than
320 DECL_ARGUMENTS (func) when checking for tail recursion.
322 2018-03-07 Jakub Jelinek <jakub@redhat.com>
324 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
325 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
326 Volker Reichelt's entry and add entries for people that perform
327 GCC fuzzy testing and report numerous bugs.
329 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
332 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
333 readonly data in sdata, if that is disabled.
334 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
335 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
336 -mreadonly-in-sdata option.
338 2018-03-07 Martin Sebor <msebor@redhat.com>
340 PR tree-optimization/84468
341 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
342 basic block when looking for nul assignment.
344 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
347 * except.h (output_function_exception_table): Adjust prototype.
348 * except.c (output_function_exception_table): Remove FNNAME parameter
349 and add SECTION parameter. Ouput one part of the table at a time.
350 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
351 the first part of the exception table and emit unwind directives.
352 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
353 (i386_pe_seh_cold_init): Likewise.
354 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
355 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
356 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
357 (ix86_output_call_insn): Emit a nop in one more case for SEH.
358 * config/i386/winnt.c: Include except.h.
359 (struct seh_frame_state): Add reg_offset, after_prologue and
360 in_cold_section fields.
361 (i386_pe_seh_end_prologue): Set seh->after_prologue.
362 (i386_pe_seh_cold_init): New function.
363 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
364 to seh->in_cold_section.
365 (seh_emit_push): Record the offset of the push.
366 (seh_emit_save): Record the offet of the save.
367 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
368 Test seh->after_prologue to disregard the epilogue.
369 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
370 (i386_pe_end_cold_function): New function.
372 2018-03-07 Jakub Jelinek <jakub@redhat.com>
375 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
376 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
379 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
381 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
382 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
385 * multiple_target.c: Include tree-inline.h and intl.h.
386 (expand_target_clones): Diagnose and fail if node->definition and
387 !tree_versionable_function_p (node->decl).
389 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
391 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
393 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
394 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
395 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
397 2018-03-06 Jakub Jelinek <jakub@redhat.com>
400 * combine.c (try_combine): Use reg_or_subregno instead of handling
401 just paradoxical SUBREGs and REGs.
403 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
405 * config/arc/arc.c (arc_finalize_pic): Remove function.
406 (arc_must_save_register): We use single base PIC register, remove
407 checks to save/restore the PIC register.
408 (arc_expand_prologue): Likewise.
409 * config/arc/arc-protos.h (arc_set_default_type_attributes):
411 (arc_verify_short): Likewise.
412 (arc_attr_type): Likewise.
413 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
414 (walk_stores): Likewise.
415 (arc_address_cost): Make it static.
416 (arc_verify_short): Likewise.
417 (branch_dest): Likewise.
418 (arc_attr_type): Likewise.
419 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
420 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
421 (arc_final_prescan_insn): Remove inserting the nops due to
422 hardware hazards. It is done in reorg step.
423 (insn_length_variant_t): Remove.
424 (insn_length_parameters_t): Likewise.
425 (arc_insn_length_parameters): Likewise.
426 (arc_get_insn_variants): Likewise.
427 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
429 2018-03-06 Jakub Jelinek <jakub@redhat.com>
432 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
435 PR tree-optimization/84687
436 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
438 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
440 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
442 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
443 Rename to ppc_speculation_barrier.
444 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
445 __builtin_ppc_speculation_barrier.
447 2018-03-05 Jakub Jelinek <jakub@redhat.com>
450 * combine.c (combine_simplify_rtx): Don't try to simplify if
451 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
454 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
456 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
457 to 32 bytes when compiling for POWER9.
459 2018-03-05 Jakub Jelinek <jakub@redhat.com>
462 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
463 regparm >= 3 with no arg reg available also for calls with
464 flag_force_indirect_call. Pass decl to ix86_function_regparm.
467 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
469 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
471 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
474 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
476 2018-03-05 Richard Biener <rguenther@suse.de>
478 PR tree-optimization/84486
479 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
480 When inserting a __builtin_assume_aligned call set the LHS
481 SSA name alignment info accordingly.
483 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
485 PR tree-optimization/84114
486 * config/aarch64/aarch64.c (aarch64_reassociation_width)
487 Avoid reassociation of FLOAT_MODE addition.
489 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
491 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
492 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
493 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
494 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
495 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
496 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
497 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
499 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
500 __builtin_ia32_wbinvd): New builtins.
501 (SPECIAL_ARGS2): New.
502 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
503 (SPECIAL_ARGS2): New.
504 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
505 (ix86_valid_target_attribute_inner_p): Ditto.
506 (ix86_init_mmx_sse_builtins): Add special_args2.
507 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
508 TARGET_WBNOINVD_P): New.
509 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
510 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
511 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
512 * config/i386/immintrin.h (_wbinvd): New intrinsic.
513 * config/i386/pconfigintrin.h: New file.
514 * config/i386/wbnoinvdintrin.h: Ditto.
515 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
516 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
518 2018-03-05 Richard Biener <rguenther@suse.de>
520 PR tree-optimization/84670
521 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
523 (BB_VISITED_WITH_VISITED_SUCCS): New define.
524 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
525 (compute_antic_aux): Only assert the number of values in ANTIC_IN
526 doesn't grow if all successors (recursively) were visited at least
529 2018-03-05 Richard Biener <rguenther@suse.de>
531 PR tree-optimization/84650
532 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
533 if executed in the loop pipeline.
535 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
537 * doc/configfiles.texi (Configuration Files): Move info about
538 conditionalizing $target-protos.h to...
539 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
540 differs from $target-protos.h.
542 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
543 Chung-Ju Wu <jasonwucj@gmail.com>
545 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
546 * config/nds32/nds32-multiple.md (setmemsi): Define.
547 * config/nds32/nds32-memory-manipulation.c
548 (nds32_gen_dup_4_byte_to_word_value): New.
549 (emit_setmem_word_loop): New.
550 (emit_setmem_byte_loop): New.
551 (nds32_expand_setmem_loop): New.
552 (nds32_expand_setmem_loop_v3m): New.
553 (nds32_expand_setmem_unroll): New.
554 (nds32_expand_setmem): New.
556 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
557 Chung-Ju Wu <jasonwucj@gmail.com>
559 * config/nds32/nds32-memory-manipulation.c
560 (nds32_emit_load_store): New.
561 (nds32_emit_post_inc_load_store): New.
562 (nds32_emit_mem_move): New.
563 (nds32_emit_mem_move_block): New.
564 (nds32_expand_movmemsi_loop_unknown_size): New.
565 (nds32_expand_movmemsi_loop_known_size): New.
566 (nds32_expand_movmemsi_loop): New.
567 (nds32_expand_movmemsi_unroll): New.
568 (nds32_expand_movmemqi): Rename ...
569 (nds32_expand_movmemsi): ... to this.
570 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
571 (movmemsi): ... to this.
572 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
573 (nds32_expand_movmemsi): ... to this.
575 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
576 Monk Chiang <sh.chiang04@gmail.com>
577 Chung-Ju Wu <jasonwucj@gmail.com>
579 * config/nds32/nds32-protos.h
580 (nds32_expand_load_multiple): New arguments.
581 (nds32_expand_store_multiple): Ditto.
582 (nds32_valid_multiple_load_store): Rename ...
583 (nds32_valid_multiple_load_store_p): ... to this.
584 * config/nds32/nds32-memory-manipulation.c
585 (nds32_expand_load_multiple): Refine implementation.
586 (nds32_expand_store_multiple): Ditto.
587 * config/nds32/nds32-multiple.md
588 (load_multiple): Update nds32_expand_load_multiple interface.
589 (store_multiple): Update nds32_expand_store_multiple interface.
590 * config/nds32/nds32-predicates.c
591 (nds32_valid_multiple_load_store): Rename ...
592 (nds32_valid_multiple_load_store_p): ... to this and refine
594 * config/nds32/predicates.md
595 (nds32_load_multiple_and_update_address_operation): New predicate.
596 (nds32_store_multiple_and_update_address_operation): New predicate.
598 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
599 Chung-Ju Wu <jasonwucj@gmail.com>
601 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
602 (combo): New attribute.
603 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
605 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
607 * config/nds32/nds32.opt: Change -mcmodel= default value.
609 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
610 Monk Chiang <sh.chiang04@gmail.com>
611 Chung-Ju Wu <jasonwucj@gmail.com>
613 * config/nds32/constants.md (unspec_element): New enum.
614 * config/nds32/constraints.md (Umw): New constraint.
615 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
616 * config/nds32/nds32-intrinsic.md: Likewise.
617 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
618 (nds32_valid_smw_lwm_base_p): New.
619 (nds32_output_smw_single_word): New.
620 (nds32_output_lmw_single_word): New.
621 (nds32_expand_unaligned_load): New.
622 (nds32_expand_unaligned_store): New.
623 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
624 (nds32_output_smw_single_word): Declare.
625 (nds32_output_lmw_single_word): Declare.
626 (nds32_expand_unaligned_load): Declare.
627 (nds32_expand_unaligned_store): Declare.
628 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
629 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
630 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
631 NDS32_BUILTIN_UASTORE_DW.
632 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
635 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
636 Kito Cheng <kito.cheng@gmail.com>
637 Chung-Ju Wu <jasonwucj@gmail.com>
639 * config/nds32/nds32-intrinsic.c
640 (nds32_expand_builtin_null_ftype_reg): Delete.
641 (nds32_expand_builtin_reg_ftype_imm): Ditto.
642 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
643 (nds32_read_argument): New.
644 (nds32_legitimize_target): Ditto.
645 (nds32_legitimize_argument): Ditto.
646 (nds32_check_constant_argument): Ditto.
647 (nds32_expand_unop_builtin): Ditto.
648 (nds32_expand_unopimm_builtin): Ditto.
649 (nds32_expand_binop_builtin): Ditto.
650 (nds32_builtin_decl_impl): Ditto.
651 (builtin_description): Ditto.
652 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
653 (nds32_init_builtins_impl): Ditto.
654 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
655 (nds32_builtin_decl): New.
656 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
657 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
659 2018-03-02 Jeff Law <law@redhat.com>
661 * reorg.c (stop_search_p): Handle DEBUG_INSN.
662 (redundant_insn, fill_simple_delay_slots): Likewise.
663 (fill_slots_from_thread): Likewise.
664 * resource.c (mark_referenced_resources): Likewise.
665 (mark_set_resources, find_dead_or_set_registers): Likewise.
667 2018-03-02 Jakub Jelinek <jakub@redhat.com>
669 * substring-locations.h (format_warning_va): Formatting fix for
671 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
673 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
674 * substring-locations.c: Include intl.h.
675 (format_warning_va): Turned into small wrapper around
676 format_warning_n_va, renamed to ...
677 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
678 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
680 (format_warning_at_substring_n): New function.
681 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
682 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
683 format_warning_at_substring with just a shorter name instead of
684 const function pointer.
685 (fmtwarn_n): New function.
686 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
687 appropriate, get rid of all the fmtstr temporaries, move conditionals
688 with G_() wrapped string literals directly into fmtwarn arguments,
689 cast dir.len to (int), formatting fixes.
691 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
693 * doc/invoke.texi: Remove "Cilk Plus" references.
695 2018-03-02 Jakub Jelinek <jakub@redhat.com>
696 Richard Biener <rguenther@suse.de>
699 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
700 for error or warning attributes if CALL_FROM_THUNK_P is set.
703 2018-03-02 Jakub Jelinek <jakub@redhat.com>
706 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
707 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
710 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
711 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
713 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
714 instead of -1U in last predictors element's probability member.
716 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
719 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
720 arguments if they are comparable.
722 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
724 PR tree-optimization/84634
725 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
726 masks and masked_loop_p with a single loop_masks, making sure it's
727 null for bb vectorization.
729 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
731 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
732 (vect_analyze_data_ref_access): Use loop->safe_len rather than
733 loop->force_vectorize to check whether there is no alias.
735 2018-03-02 Jakub Jelinek <jakub@redhat.com>
738 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
740 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
742 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
743 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
744 instead of a loop around prev_real_insn.
745 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
749 * config/i386/i386.c (ix86_print_operand): Use conditional
750 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
753 2018-03-02 Richard Biener <rguenther@suse.de>
755 PR tree-optimization/84427
756 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
757 (bitmap_set_subtract_values): Rewrite to handle multiple
760 (prune_clobbered_mems): Likewise.
761 (phi_translate): Take edge instead of pred/phiblock.
762 (phi_translate_1): Likewise.
763 (phi_translate_set): Likewise. Insert all translated
764 exprs for a value into the set, keeping possibly multiple
765 expressions per value.
766 (compute_antic_aux): Adjust for phi_translate changes.
767 When intersecting union the expressions and prune those
768 not in the final value set, keeping possibly multiple
769 expressions per value. Do not use value-insertion
770 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
771 all expressions. Add verification that the value-sets
772 only shrink during iteration.
773 (compute_partial_antic_aux): Adjust for the phi_translate changes.
774 (do_pre_regular_insertion): Likewise.
775 (do_pre_partial_partial_insertion): Likewise.
777 2018-03-02 Richard Biener <rguenther@suse.de>
780 * config/darwin.c (saved_debug_info_level): New static global.
781 (darwin_asm_lto_start): Disable debug info generation for LTO out.
782 (darwin_asm_lto_end): Restore debug info generation settings.
784 2018-03-01 Martin Liska <mliska@suse.cz>
787 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
790 2018-03-01 Richard Biener <rguenther@suse.de>
793 * dwarf2out.c (gen_variable_die): Properly handle late VLA
794 type annotation with LTO when debug was disabled at compile-time.
796 2018-03-01 Matthew Fortune <mfortune@gmail.com>
798 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
800 (mips_final_postscan_insn): Likewise.
802 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
804 PR rtl-optimization/84528
805 * alias.c (init_alias_target): Add commentary.
806 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
807 a unique base value if the frame pointer is not eliminated
808 to the stack pointer.
810 2018-03-01 Tom de Vries <tom@codesourcery.com>
812 PR rtl-optimization/83327
813 * lra-int.h (hard_regs_spilled_into): Declare.
814 * lra.c (hard_regs_spilled_into): Define.
815 (init_reg_info): Init hard_regs_spilled_into.
816 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
817 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
818 (process_bb_lives): Handle hard_regs_spilled_into.
819 (lra_create_live_ranges_1): Before doing liveness propagation, clear
820 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
822 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
824 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
825 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
826 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
827 * config/rs6000/aix72.h: New file.
829 2018-02-28 Jakub Jelinek <jakub@redhat.com>
831 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
832 instead of warning_at with conditional singular and plural messages
836 * stor-layout.c (update_alignment_for_field): For
837 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
838 && !DECL_PACKED (field), do the alignment update, just use
839 only desired_align instead of MAX (type_align, desired_align)
841 (place_field): Don't do known_align < desired_align handling
842 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
843 is non-NULL, instead do it after rli->prev_field handling and
844 only if not within a bitfield word. For DECL_PACKED (field)
845 use type_align of BITS_PER_UNIT.
847 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
849 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
850 superfluous parentheses and trailing spaces.
852 2018-02-28 Richard Biener <rguenther@suse.de>
854 PR tree-optimization/84584
855 * graphite-scop-detection.c (scop_detection::add_scop): Discard
856 SCoPs with fake exit edge.
858 2018-02-28 Martin Liska <mliska@suse.cz>
861 * timevar.c (timer::print): Fix format to properly print 100%
864 2018-02-28 Richard Biener <rguenther@suse.de>
867 * genmatch.c (capture_info::walk_match): Do not mark
868 captured expressions without operands as expr_p given
869 they act more like predicates and should be subject to
870 "lost tail" side-effect preserving.
872 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
874 PR rtl-optimization/81611
875 * auto-inc-dec.c (attempt_change): Move dead note from
876 mem_insn if it's the next use of regno
877 (find_address): Take address use of reg holding
878 non-incremented value. Add parm to limit search to the named
880 (merge_in_block): Attempt to use a mem insn that is the next
881 use of the original regno.
883 2018-02-27 Martin Sebor <msebor@redhat.com>
886 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
887 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
889 2018-02-27 Martin Sebor <msebor@redhat.com>
892 * diagnostic-core.h (warning_n, error_n, inform_n): Change
893 n argument to unsigned HOST_WIDE_INT.
894 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
895 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
896 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
897 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
899 2018-02-27 Richard Biener <rguenther@suse.de>
901 PR tree-optimization/84512
902 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
903 Do not use the estimate returned from record_stmt_cost for
904 the scalar iteration cost but sum properly using add_stmt_cost.
906 2018-02-27 Richard Biener <rguenther@suse.de>
908 PR tree-optimization/84466
909 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
910 Adjust last change to less strictly validate use operands.
912 2018-02-27 Martin Liska <mliska@suse.cz>
914 PR gcov-profile/84548
915 * gcov.c (process_file): Allow partial overlap and consider it
916 also as group functions.
917 (output_lines): Properly calculate range of lines for a group.
919 2018-02-27 Martin Liska <mliska@suse.cz>
921 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
922 'ggc' suffixes. Change first column width.
923 (timer::print): Fix formatting of the column.
925 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
927 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
928 preserve inline entry blocks for the sake of debug inline
929 entry point markers alone.
930 (remove_unused_locals): Suggest in comments a better place to
931 force the preservation of inline entry blocks that are
932 otherwise unused, but do not preserve them.
934 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
936 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
938 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
941 * config/i386/constraints.md (Bs): Replace
942 ix86_indirect_branch_register with
943 TARGET_INDIRECT_BRANCH_REGISTER.
945 * config/i386/i386.md (indirect_jump): Likewise.
946 (tablejump): Likewise.
947 (*sibcall_memory): Likewise.
948 (*sibcall_value_memory): Likewise.
949 Peepholes of indirect call and jump via memory: Likewise.
950 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
951 (*sibcall_value_GOT_32): Likewise.
952 * config/i386/predicates.md (indirect_branch_operand): Likewise.
953 (GOT_memory_operand): Likewise.
954 (call_insn_operand): Likewise.
955 (sibcall_insn_operand): Likewise.
956 (GOT32_symbol_operand): Likewise.
957 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
959 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
961 PR rtl-optimization/83496
962 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
963 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
964 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
965 redundant insn, if any.
966 (relax_delay_slots): Likewise.
967 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
969 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
971 PR tree-optimization/83965
972 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
973 that grouped statements are part of a reduction chain. Return
974 true if the statement is not marked as a reduction itself but
976 (vect_recog_dot_prod_pattern): Don't check whether the statement
977 is part of a group here.
978 (vect_recog_sad_pattern): Likewise.
979 (vect_recog_widen_sum_pattern): Likewise.
981 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
984 * final.c (rest_of_clean_state): Also look for calls inside sequences.
986 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
989 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
991 (ix86_output_indirect_function_return): New prototype.
992 (ix86_split_simple_return_pop_internal): Likewise.
993 * config/i386/i386.c (indirect_return_via_cx): New.
994 (indirect_return_via_cx_bnd): Likewise.
995 (indirect_thunk_name): Handle return va CX_REG.
996 (output_indirect_thunk_function): Create alias for
997 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
998 (ix86_output_indirect_jmp): Remove the bool argument.
999 (ix86_output_indirect_function_return): New function.
1000 (ix86_split_simple_return_pop_internal): Likewise.
1001 * config/i386/i386.md (*indirect_jump): Don't pass false
1002 to ix86_output_indirect_jmp.
1003 (*tablejump_1): Likewise.
1004 (simple_return_pop_internal): Change it to define_insn_and_split.
1005 Call ix86_split_simple_return_pop_internal to split it for
1007 (simple_return_indirect_internal): Call
1008 ix86_output_indirect_function_return instead of
1009 ix86_output_indirect_jmp.
1011 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1014 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1015 memset and value initialization afterwards.
1017 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1019 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1021 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1024 * common/config/aarch64/aarch64-common.c
1025 (aarch_option_optimization_table[]): Switch
1026 off fomit-frame-pointer
1028 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1029 Chung-Ju Wu <jasonwucj@gmail.com>
1031 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1033 (store_multiple): Ditto.
1035 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1037 * config.gcc: Add --with-cpu support for nds32 target.
1038 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1039 * config/nds32/nds32.opt: Add -mcpu= option.
1041 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1043 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1044 isel=yes): Warn for these deprecated options.
1046 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1048 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1049 ISA_2_5_MASKS_EMBEDDED.
1051 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1053 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1054 p->max as pointers rather than using iterative_hash_expr.
1056 2018-02-23 Carl Love <cel@us.ibm.com>
1058 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1059 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1061 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1062 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1063 P8V_BUILTIN_VEC_VUNSIGNED2.
1065 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1068 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1069 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1071 * lra-constraints.c (curr_insn_transform): Set up
1072 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1074 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1075 * lra-eliminations.c (spill_pseudos): Ditto.
1076 (process_insn_for_elimination): Ditto.
1077 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1078 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1079 LRA_NON_CLOBBERED_ALT.
1081 2018-02-22 Martin Sebor <msebor@redhat.com>
1083 PR tree-optimization/84480
1084 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1085 to maybe_diag_stxncpy_trunc. Call it.
1086 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1087 from gimple_fold_builtin_strcpy. Print inlining stack.
1088 (handle_builtin_stxncpy): Print inlining stack.
1089 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1091 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1094 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1095 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1096 and -fcheck-pointer-bounds are used together.
1097 (indirect_thunk_prefix): New enum.
1098 (indirect_thunk_need_prefix): New function.
1099 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1100 "_nt" instead of "_bnd" for NOTRACK prefix.
1101 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1102 (output_indirect_thunk_function): Likewise.
1104 (ix86_code_end): Update output_indirect_thunk_function calls.
1105 (ix86_output_indirect_branch_via_reg): Replace
1106 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1107 (ix86_output_indirect_branch_via_push): Likewise.
1108 (ix86_output_function_return): Likewise.
1109 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1110 incompatible with -fcf-protection=branch and
1111 -fcheck-pointer-bounds.
1113 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1116 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1117 Change gcc_assert call to output_operand_lossage.
1119 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1121 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1123 2018-02-22 DJ Delorie <dj@redhat.com>
1124 Sebastian Perta <sebastian.perta@renesas.com>
1125 Oleg Endo <olegendo@gcc.gnu.org>
1127 * config/rx/rx.c (rx_rtx_costs): New function.
1128 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1130 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1132 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1134 2018-02-22 Martin Liska <mliska@suse.cz>
1137 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1138 Add "native" as a possible value.
1140 2018-02-22 Martin Liska <mliska@suse.cz>
1143 * config/i386/i386.c (ix86_option_override_internal):
1144 Add "native" as a possible value for -march and -mtune.
1146 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1149 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1150 to all type variants.
1152 PR tree-optimization/84503
1153 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1154 width as info->bitpos + info->bitsize - start.
1155 (merged_store_group::merge_overlapping): Simplify width computation.
1156 (check_no_overlap): New function.
1157 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1158 start + width and last_order of the group, fail if check_no_overlap
1160 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1161 to group if check_no_overlap fails.
1163 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1165 * config/rs6000/altivec.md: Delete contraint arguments to
1166 define_expand, define_split, and define_peephole2, and in
1167 define_insn_and_split if always unused.
1168 * config/rs6000/darwin.md: Ditto.
1169 * config/rs6000/dfp.md: Ditto.
1170 * config/rs6000/rs6000.md: Ditto.
1171 * config/rs6000/sync.md: Ditto.
1172 * config/rs6000/vector.md: Ditto.
1173 * config/rs6000/vsx.md: Ditto.
1175 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1177 * config/rs6000/altivec.md: Write output control strings as braced
1178 blocks instead of double-quoted strings.
1179 * config/rs6000/darwin.md: Ditto.
1180 * config/rs6000/rs6000.md: Ditto.
1181 * config/rs6000/vector.md: Ditto.
1182 * config/rs6000/vsx.md: Ditto.
1184 2018-02-21 Jason Merrill <jason@redhat.com>
1186 PR c++/84314 - ICE with templates and fastcall attribute.
1187 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1189 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1191 * ipa-cp.c (determine_versionability): Fix comment typos.
1193 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1196 * ipa-cp.c (determine_versionability): Do not version functions caling
1199 2018-02-21 Martin Liska <mliska@suse.cz>
1202 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1203 Add "native" as a possible value.
1204 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1205 the macro when native cpu detection is available.
1207 2018-02-21 Martin Liska <mliska@suse.cz>
1210 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1211 Add "native" as a possible value.
1212 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1213 when native cpu detection is available.
1215 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1216 Martin Sebor <msebor@redhat.com>
1218 PR tree-optimization/84478
1219 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1221 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1222 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1223 support which is conservatively correct, for 2 only stay conservative
1224 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1225 argument to the 2 argument get_range_strlen, adjust 6 arg
1226 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1228 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1229 (gimple_fold_builtin_strlen): Pass true as last argument to
1232 2018-02-20 Martin Sebor <msebor@redhat.com>
1235 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1236 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1237 (builtin_memref::builtin_memref): Factor out parts into
1238 set_base_and_offset and call it.
1240 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1243 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1244 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1245 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1246 search at the associated MODE_INT.
1248 2018-02-20 Jeff Law <law@redhat.com>
1251 PR tree-optimization/81592
1253 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1254 for range data rather than using global data.
1255 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1256 range data rather than using global data.
1257 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1258 pass it to children as needed.
1259 (struct directive::fmtresult): Similarly.
1260 (struct directive::set_width): Similarly.
1261 (struct directive::set_precision): Similarly.
1262 (format_integer, format_directive, parse_directive): Similarly.
1263 (format_none): Accept unnamed vr_values parameter.
1264 (format_percent, format_floating, format_character): Similarly.
1265 (format_string, format_plain): Similarly.
1266 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1267 the EVRP range analyzer for range data rather than using global data.
1268 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1269 gimple-ssa-evrp-analyze.h
1270 (class sprintf_dom_walker): Add after_dom_children member function.
1271 Add evrp_range_analyzer member.
1272 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1273 range analyzer as needed.
1274 (sprintf_dom_walker::after_dom_children): New member function.
1275 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1277 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1278 (evrp_range_analyzer::pop_to_marker): Likewise.
1280 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1282 PR tree-optimization/84419
1283 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1284 with the required type if its current type is compatible but
1287 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1290 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1291 after vectorization.
1293 2018-02-20 Martin Liska <mliska@suse.cz>
1296 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1297 possible values if we don't have a hint.
1299 2018-02-20 Martin Liska <mliska@suse.cz>
1303 * final.c (shorten_branches): Build align_tab array with one
1305 * opts.c (finish_options): Add alignment option limit check.
1306 (MAX_CODE_ALIGN): Likewise.
1307 (MAX_CODE_ALIGN_VALUE): Likewise.
1308 * doc/invoke.texi: Document maximum allowed option value for
1309 all -falign-* options.
1311 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1314 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1315 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1316 * var-tracking.c (emit_note_insn_var_location): Remove all references
1317 to NOTE_INSN_CALL_ARG_LOCATION.
1318 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1319 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1320 Use copy_rtx_if_shared.
1321 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1322 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1323 (dwarf2out_var_location): Remove handling of
1324 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1326 * final.c (final_scan_insn): Remove all references to
1327 NOTE_INSN_CALL_ARG_LOCATION.
1328 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1329 before dumping final insns.
1330 * except.c (emit_note_eh_region_end): Remove all references to
1331 NOTE_INSN_CALL_ARG_LOCATION.
1332 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1333 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1334 * config/arc/arc.c (hwloop_optimize): Likewise.
1335 * config/arm/arm.c (create_fix_barrier): Likewise.
1336 * config/s390/s390.c (s390_chunkify_start): Likewise.
1337 * config/sh/sh.c (find_barrier): Likewise.
1338 * config/i386/i386.c (rest_of_insert_endbranch,
1339 ix86_seh_fixup_eh_fallthru): Likewise.
1340 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1341 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1342 * config/frv/frv.c (frv_function_prologue): Likewise.
1343 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1345 (note_outside_basic_block_p): Remove all references to
1346 NOTE_INSN_CALL_ARG_LOCATION.
1347 * gengtype.c (adjust_field_rtx_def): Likewise.
1348 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1350 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1351 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1354 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1357 PR tree-optimization/84452
1358 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1359 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1362 2018-02-19 Martin Liska <mliska@suse.cz>
1365 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1367 2018-02-19 Martin Liska <mliska@suse.cz>
1368 Richard Sandiford <richard.sandiford@linaro.org>
1370 PR tree-optimization/82491
1371 * gimple-fold.c (get_base_constructor): Make earlier bail out
1374 2018-02-19 Carl Love <cel@us.ibm.com>
1376 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1377 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1379 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1380 P8V_BUILTIN_VEC_NEG.
1382 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1384 * config/rl78/rl78.md (movdf): New define expand.
1386 2018-02-19 Martin Liska <mliska@suse.cz>
1389 * doc/invoke.texi: Fix typo.
1390 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1392 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1394 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1395 handle rs6000_single_float and rs6000_double_float specially for
1398 2018-02-16 Jeff Law <law@redhat.com>
1400 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1403 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1406 * ipa-inline.c (inline_small_functions): Fix a typo.
1408 2018-02-16 Nathan Sidwell <nathan@acm.org>
1410 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1412 2018-02-16 Carl Love <cel@us.ibm.com>
1414 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1415 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1416 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1417 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1418 expansion to P8V_BUILTIN_VEC_FLOAT2.
1420 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1422 PR rtl-optimization/70023
1423 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1424 src_regno into account.
1426 2018-02-16 Carl Love <cel@us.ibm.com>
1428 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1429 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1430 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1431 * config/rs6000/rs6000.c: Remove case statements for
1432 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1433 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1434 and P9V_BUILTIN_VEC_VINSERT4B.
1435 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1436 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1437 * config/rs6000/vsx.md:
1438 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1441 2018-02-16 Carl Love <cel@us.ibm.com>
1443 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1445 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1447 * config/rs6000/rs6000-c.c: Add the definitions for
1448 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1449 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1450 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1451 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1452 definition for insert4b and define insn *insert3b_internal.
1453 * doc/extend.texi: Add documentation for vec_extract4b.
1455 2018-02-16 Nathan Sidwell <nathan@acm.org>
1457 * doc/extend.texi (Backwards Compatibility): Mention friend
1458 injection. Note for-scope is deprecated.
1459 * doc/invoke.texi (-ffriend-injection): Deprecate.
1461 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1463 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1464 that moved to I2, also allow destinations that are a paradoxical
1465 subreg (instead of a normal reg).
1467 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1470 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1473 2018-02-16 Richard Biener <rguenther@suse.de>
1475 PR tree-optimization/84037
1476 PR tree-optimization/84016
1478 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1479 Adjust vec_construct for the fact we need additional higher latency
1480 128bit inserts for AVX256 and AVX512 vector builds.
1481 (ix86_add_stmt_cost): Scale vector construction cost for
1484 2018-02-16 Richard Biener <rguenther@suse.de>
1486 PR tree-optimization/84417
1487 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1488 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1489 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1491 2018-02-16 Martin Liska <mliska@suse.cz>
1494 * internal-fn.def (ASAN_CHECK): Set proper flags.
1495 (ASAN_MARK): Likewise.
1497 2018-02-16 Julia Koval <julia.koval@intel.com>
1499 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1500 from PTA_CANNONLAKE.
1502 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1505 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1506 Use ++iter rather than iter++ for std::list iterators.
1507 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1508 defer deleting them until all nodes in the forest are processed. Do
1509 free even leaf nodes. Change to_process into auto_vec.
1512 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1513 * vec.h (vec_default_construct): Use memset instead of placement new
1514 if BROKEN_VALUE_INITIALIZATION is defined.
1515 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1516 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1519 PR rtl-optimization/83723
1520 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1521 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1522 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1524 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1526 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1528 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1530 PR rtl-optimization/81443
1531 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1532 from inner REGs to paradoxical SUBREGs.
1534 2018-02-16 Richard Biener <rguenther@suse.de>
1536 PR tree-optimization/84399
1537 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1538 For operands we can analyze at their definition make sure we can
1539 analyze them at each use as well.
1541 2018-02-16 Richard Biener <rguenther@suse.de>
1543 PR tree-optimization/84190
1544 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1545 volatile accesses if the decl isn't volatile.
1547 2018-02-15 Jason Merrill <jason@redhat.com>
1549 PR c++/84314 - ICE with templates and fastcall attribute.
1550 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1551 TYPE_CANONICAL on an existing type.
1553 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1555 PR tree-optimization/84383
1556 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1557 dstoff nor call operand_equal_p if dstbase is NULL.
1559 PR tree-optimization/84334
1560 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1561 also a CONSTANT_CLASS_P, punt.
1563 2018-02-14 Jim Wilson <jimw@sifive.com>
1565 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1566 first SMALL_OPERAND check. New local min_second_step. Move assert
1567 to where locals are set. Add TARGET_RVC support.
1568 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1570 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1572 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1574 2018-02-14 Martin Sebor <msebor@redhat.com>
1576 PR tree-optimization/83698
1577 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1578 arrays constrain the offset range to their bounds.
1579 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1580 (builtin_access::overlap): Avoid setting the size of overlap if it's
1582 (maybe_diag_overlap): Also consider arrays when deciding what values
1583 of offsets to include in diagnostics.
1585 2018-02-14 Martin Sebor <msebor@redhat.com>
1588 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1589 that correspond to the kind of a declaration.
1591 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1594 * config/pa/pa.md: Load address of PIC label using the linkage table
1595 if the label is nonlocal.
1597 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1599 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1600 warning message if user requests -maltivec=be.
1601 * doc/invoke.texi: Document deprecation of -maltivec=be.
1603 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1606 * config/rs6000/rs6000-c.c: Update definitions for
1607 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1608 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1610 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1613 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1614 add _get_ssp intrinsics. Remove argument from
1615 __builtin_ia32_rdssp[d|q].
1616 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1617 * config/i386/i386-builtin.def: Remove argument from
1618 __builtin_ia32_rdssp[d|q].
1619 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1620 ix86_expand_special_args_builtin for _rdssp[d|q].
1621 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1622 Clear register before usage.
1623 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1624 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1626 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1628 PR tree-optimization/84357
1629 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1630 operand 1 of an ARRAY_REF too.
1632 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1635 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1636 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1638 (set_of_reg): New struct.
1639 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1640 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1641 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1643 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1644 Split into bitclr, bitset, bitinvert patterns if appropriate.
1645 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1646 use rx_fuse_in_memory_bitop.
1647 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1648 to named insn, correct maximum insn length.
1650 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1653 * machmode.def: Define a complex mode for PARTIAL_INT.
1654 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1656 * doc/rtl.texi: Document CSPImode.
1657 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1659 (msp430_hard_regno_nregs_with_padding): Likewise.
1661 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1664 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1666 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1668 PR rtl-optimization/84169
1669 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1670 we generated a parallel as new i3 and we split that to new i2 and i3
1671 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1672 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1673 those to i2, not i1. Partially rewrite this scan code.
1675 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1678 * stor-layout.c (place_field): For variable length fields, adjust
1679 offset_align afterwards not just based on the field's alignment,
1680 but also on the size.
1683 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1684 of exps and logs in the use_exp2 case.
1686 2018-02-13 Jeff Law <law@redhat.com>
1688 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1691 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1694 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1698 * common.opt (gas-loc-support, gas-locview-support): New.
1699 (ginline-points, ginternal-reset-location-views): New.
1700 * doc/invoke.texi: Document them. Use @itemx where intended.
1701 (gvariable-location-views): Adjust.
1702 * target.def (reset_location_view): New.
1703 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1704 (TARGET_RESET_LOCATION_VIEW): New.
1705 * doc/tm.texi: Rebuilt.
1706 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1707 (dwarf2out_default_as_locview_support): New.
1708 (output_asm_line_debug_info): Use option variables.
1709 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1710 (output_loc_list): Likewise.
1711 (add_high_low_attributes): Check option variables.
1712 Don't output entry view attribute in strict mode.
1713 (gen_inlined_subroutine_die): Check option variables.
1714 (dwarf2out_inline_entry): Likewise.
1715 (init_sections_and_labels): Likewise.
1716 (dwarf2out_early_finish): Likewise.
1717 (maybe_reset_location_view): New, from...
1718 (dwarf2out_var_location): ... here. Call it.
1719 * debug.h (dwarf2out_default_as_loc_support): Declare.
1720 (dwarf2out_default_as_locview_support): Declare.
1721 * hooks.c (hook_int_rtx_insn_0): New.
1722 * hooks.h (hook_int_rtx_insn_0): Declare.
1723 * toplev.c (process_options): Take -gas-loc-support and
1724 -gas-locview-support from dwarf2out. Enable
1725 -gvariable-location-views by default only with locview
1726 assembler support. Enable -ginternal-reset-location-views by
1727 default only if the target defines the corresponding hook.
1728 Enable -ginline-points by default if location views are
1729 enabled; force it disabled if statement frontiers are
1731 * tree-inline.c (expand_call_inline): Check option variables.
1732 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1734 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1736 PR tree-optimization/84321
1737 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1738 handling. Also check whether the anti-range contains any values
1739 that satisfy the mask; switch to a VR_RANGE if not.
1741 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1744 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1746 2018-02-13 Martin Jambor <mjambor@suse.cz>
1749 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1750 of call statements, also set location of a load to a temporary.
1752 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1754 * config/rl78/rl78.c (add_vector_labels): New function.
1755 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1756 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1757 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1758 which checks that no arguments are passed.
1759 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1760 * doc/extend.texi: Documentation for the new attribute.
1762 2018-02-13 Andreas Schwab <schwab@suse.de>
1764 * config/riscv/linux.h (CPP_SPEC): Define.
1766 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1769 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1770 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1771 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1772 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1773 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1774 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1775 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1776 temporarily for AES and PCLMUL builtins.
1778 PR tree-optimization/84339
1779 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1780 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1784 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1785 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1786 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1788 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1790 * omp-simd-clone.h: New file.
1791 * omp-simd-clone.c: Include omp-simd-clone.h.
1792 (expand_simd_clones): No longer static.
1793 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1794 cgraph.h and omp-simd-clone.h.
1795 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1796 (vect_recog_widen_shift_pattern): Formatting fix.
1797 (vect_pattern_recog_1): Don't check optab for calls.
1800 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1801 operands[2] into a REG before using gen_lowpart on it.
1803 2018-02-12 Jeff Law <law@redhat.com>
1806 * config/sh/sh.c (find_barrier): Consider a sibling call
1809 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1810 successfully back substituting a reg.
1812 2018-02-12 Richard Biener <rguenther@suse.de>
1814 PR tree-optimization/84037
1815 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1816 parameter, move visited init to caller.
1817 (vect_slp_analyze_operations): Separate cost from validity
1818 check, initialize visited once for all instances.
1819 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1821 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1823 (vect_model_promotion_demotion_cost): Likewise.
1824 (vectorizable_bswap): Guard cost modeling with !slp_node
1825 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1827 (vectorizable_call): Likewise.
1828 (vectorizable_conversion): Likewise.
1829 (vectorizable_assignment): Likewise.
1830 (vectorizable_shift): Likewise.
1831 (vectorizable_operation): Likewise.
1832 (vectorizable_store): Likewise.
1833 (vectorizable_load): Likewise.
1834 (vectorizable_condition): Likewise.
1835 (vectorizable_comparison): Likewise.
1837 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1840 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1841 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1842 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1844 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1847 * params.def (inline-min-speedup): Increase from 8 to 15.
1848 (max-inline-insns-auto): Decrease from 40 to 30.
1849 * ipa-split.c (consider_split): Add some buffer for function to
1850 be considered inlining candidate.
1851 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1854 2018-02-12 Richard Biener <rguenther@suse.de>
1856 PR tree-optimization/84037
1857 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1858 matched stmts if we cannot swap the non-matched ones.
1860 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1862 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1863 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1864 _mm_maskz_scalef_round_ss): New intrinsics.
1865 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1866 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1867 __builtin_ia32_scalefss_round): Remove.
1868 (__builtin_ia32_scalefsd_mask_round,
1869 __builtin_ia32_scalefss_mask_round): New intrinsics.
1870 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1871 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1872 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1873 "<round_constraint>")): Changed to ...
1874 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1875 "<round_scalar_constraint>")): ... this.
1876 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1877 %0, %1, %2<round_op3>}"): Changed to ...
1878 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1879 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1880 %2<round_scalar_mask_op3>}"): ... this.
1881 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1883 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1885 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1886 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1887 (_mm_maskz_sqrt_round_ss): New intrinsics.
1888 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1889 (__builtin_ia32_sqrtsd_mask_round)
1890 (__builtin_ia32_sqrtss_mask_round): New builtins.
1891 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1892 (__builtin_ia32_sqrtss_round): Remove.
1893 (__builtin_ia32_sqrtsd_mask_round)
1894 (__builtin_ia32_sqrtss_mask_round): New builtins.
1895 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1896 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1897 ((match_operand:VF_128 1 "vector_operand"
1898 "xBm,<round_constraint>")): Changed to ...
1899 ((match_operand:VF_128 1 "vector_operand"
1900 "xBm,<round_scalar_constraint>")): ... this.
1901 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1902 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1903 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1904 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1905 %<iptr>1<round_scalar_mask_op3>}): ... this.
1906 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1907 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1909 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1912 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1913 Cast vec_cmpeq result to correct type.
1914 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1915 Cast vec_cmpgt result to correct type.
1917 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1919 * final.c (final_scan_insn_1): Renamed from...
1920 (final_scan_insn): ... this. New wrapper, to recover
1921 seen from the outermost call in recursive ones.
1922 * config/sparc/sparc.c (output_return): Drop seen from call.
1923 (output_sibcall): Likewise.
1924 * config/visium/visium.c (output_branch): Likewise.
1926 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1928 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1931 2018-02-10 Alan Modra <amodra@gmail.com>
1934 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1935 Specify LR as an input.
1937 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1940 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1941 remove_member_access_dummy_vars): New functions.
1942 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1943 lower_omp_1, execute_lower_omp): Use them.
1945 PR rtl-optimization/84308
1946 * shrink-wrap.c (spread_components): Release todo vector.
1948 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1950 PR rtl-optimization/57193
1951 * ira-color.c (struct allocno_color_data): Add member
1952 conflict_allocno_hard_prefs.
1953 (update_conflict_allocno_hard_prefs): New.
1954 (bucket_allocno_compare_func): Add a preference based on
1955 conflict_allocno_hard_prefs.
1956 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1957 (color_allocnos): Remove a dead code. Initiate
1958 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1960 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1963 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1964 constraint from =wa to wa. Avoid a subreg on the output operand,
1965 instead use a pseudo and subreg it in a move.
1966 (p9_xxbrd_<mode>): Changed to ...
1967 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1968 (p9_xxbrd_v2df): New expander.
1969 (p9_xxbrw_<mode>): Changed to ...
1970 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1971 (p9_xxbrw_v4sf): New expander.
1973 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1975 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1977 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1980 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1981 multiply in 32-bit mode.
1982 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1983 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1986 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1988 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1989 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1990 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1991 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1993 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1996 * dwarf2out.c (is_trivial_indirect_ref): New function.
1997 (dwarf2out_late_global_decl): Do not generate a location
1998 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1999 and that are not defined in the current unit.
2001 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2003 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2004 instead of a libcall for UNORDERED.
2006 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2009 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2010 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2012 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2015 * config/s390/s390.c (s390_set_current_function): Invoke
2016 s390_indirect_branch_settings also if fndecl didn't change.
2018 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2020 * config/rs6000/rs6000.md (blockage): Set length to zero.
2022 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2024 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2026 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2029 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2030 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2034 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2035 PARALLEL incoming that failed vt_get_decl_and_offset check.
2038 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2039 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2041 (get_variable_section): For decls in named .bss* sections pass true as
2042 second argument to bss_initializer_p.
2044 2018-02-09 Marek Polacek <polacek@redhat.com>
2045 Jakub Jelinek <jakub@redhat.com>
2048 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2049 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2050 Sync some changes from cxx_fold_indirect_ref.
2052 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2054 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2056 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2057 (BLOCK_INLINE_ENTRY_LABEL): New.
2058 (dwarf2out_var_location): Disregard inline entry markers.
2059 (inline_entry_data): New struct.
2060 (inline_entry_data_hasher): New hashtable type.
2061 (inline_entry_data_hasher::hash): New.
2062 (inline_entry_data_hasher::equal): New.
2063 (inline_entry_data_table): New variable.
2064 (add_high_low_attributes): Add DW_AT_entry_pc and
2065 DW_AT_GNU_entry_view attributes if a pending entry is found
2066 in inline_entry_data_table. Add old entry_pc attribute only
2067 if debug nonbinding markers are disabled.
2068 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2069 markers are enabled.
2070 (block_within_block_p, dwarf2out_inline_entry): New.
2071 (dwarf2out_finish): Check that no entries remained in
2072 inline_entry_data_table.
2073 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2074 (final_scan_insn, notice_source_line): Likewise.
2075 (rest_of_clean_state): Skip inline entry markers.
2076 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2078 * gimple.c (gimple_build_debug_inline_entry): New.
2079 * gimple.h (enum gimple_debug_subcode): Add
2080 GIMPLE_DEBUG_INLINE_ENTRY.
2081 (gimple_build_debug_inline_entry): Declare.
2082 (gimple_debug_inline_entry_p): New.
2083 (gimple_debug_nonbind_marker_p): Adjust.
2084 * insn-notes.def (INLINE_ENTRY): New.
2085 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2086 inline entry marker notes.
2087 (print_insn): Likewise.
2088 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2089 (INSN_DEBUG_MARKER_KIND): Likewise.
2090 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2091 * tree-inline.c (expand_call_inline): Build and insert
2092 debug_inline_entry stmt.
2093 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2094 inline entry blocks early, if nonbind markers are enabled.
2095 (dump_scope_block): Dump fragment info.
2096 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2097 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2098 (gimple_build_debug_inline_entry): New.
2099 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2100 Enable/disable inline entry points too.
2101 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2102 (DEBUG_INSN): Describe inline entry markers.
2104 * common.opt (gvariable-location-views): New.
2105 (gvariable-location-views=incompat5): New.
2106 * config.in: Rebuilt.
2107 * configure: Rebuilt.
2108 * configure.ac: Test assembler for view support.
2109 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2110 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2111 * dwarf2out.c (var_loc_view): New typedef.
2112 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2113 (dwarf2out_locviews_in_attribute): New.
2114 (dwarf2out_locviews_in_loclist): New.
2115 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2116 (enum dw_line_info_opcode): Add LI_adv_address.
2117 (struct dw_line_info_table): Add view.
2118 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2119 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2120 (zero_view_p): New variable.
2121 (ZERO_VIEW_P): New macro.
2122 (output_asm_line_debug_info): New.
2123 (struct var_loc_node): Add view.
2124 (add_AT_view_list, AT_loc_list): New.
2125 (add_var_loc_to_decl): Add view param. Test it against last.
2126 (new_loc_list): Add view params. Record them.
2127 (AT_loc_list_ptr): Handle loc and view lists.
2128 (view_list_to_loc_list_val_node): New.
2129 (print_dw_val): Handle dw_val_class_view_list.
2130 (size_of_die): Likewise.
2131 (value_format): Likewise.
2132 (loc_list_has_views): New.
2133 (gen_llsym): Set vl_symbol too.
2134 (maybe_gen_llsym, skip_loc_list_entry): New.
2135 (dwarf2out_maybe_output_loclist_view_pair): New.
2136 (output_loc_list): Output view list or entries too.
2137 (output_view_list_offset): New.
2138 (output_die): Handle dw_val_class_view_list.
2139 (output_dwarf_version): New.
2140 (output_compilation_unit_header): Use it.
2141 (output_skeleton_debug_sections): Likewise.
2142 (output_rnglists, output_line_info): Likewise.
2143 (output_pubnames, output_aranges): Update version comments.
2144 (output_one_line_info_table): Output view numbers in asm comments.
2145 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2146 Call maybe_gen_llsym.
2147 (loc_list_from_tree_1): Adjust.
2148 (add_AT_location_description): Create view list attribute if
2149 needed, check it's absent otherwise.
2150 (convert_cfa_to_fb_loc_list): Adjust.
2151 (maybe_emit_file): Call output_asm_line_debug_info for test.
2152 (dwarf2out_var_location): Reset views as needed. Precompute
2153 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2154 attribute. Set view.
2155 (new_line_info_table): Reset next view.
2156 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2157 (dwarf2out_source_line): Likewise. Output view resets and labels to
2158 the assembler, or select appropriate line info opcodes.
2159 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2160 (optimize_string_length): Catch it. Adjust.
2161 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2162 dw_val_class_view_list, and remove it if no longer needed.
2163 (hash_loc_list): Hash view numbers.
2164 (loc_list_hasher::equal): Compare them.
2165 (optimize_location_lists): Check whether a view list symbol is
2166 needed, and whether the locview attribute is present, and
2167 whether they match. Remove the locview attribute if no longer
2169 (index_location_lists): Call skip_loc_list_entry for test.
2170 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2171 Use output_dwarf_version.
2172 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2173 (struct dw_val_node): Add val_view_list.
2174 * final.c (SEEN_NEXT_VIEW): New.
2175 (set_next_view_needed): New.
2176 (clear_next_view_needed): New.
2177 (maybe_output_next_view): New.
2178 (final_start_function): Rename to...
2179 (final_start_function_1): ... this. Take pointer to FIRST,
2180 add SEEN parameter. Emit param bindings in the initial view.
2181 (final_start_function): Reintroduce SEEN-less interface.
2182 (final): Rename to...
2183 (final_1): ... this. Take SEEN parameter. Output final pending
2184 next view at the end.
2185 (final): Reintroduce seen-less interface.
2186 (final_scan_insn): Output pending next view before switching
2187 sections or ending a block. Mark the next view as needed when
2188 outputting variable locations. Notify debug backend of section
2189 changes, and of location view changes.
2190 (rest_of_handle_final): Adjust.
2191 * toplev.c (process_options): Autodetect value for debug variable
2192 location views option. Warn on incompat5 without -gdwarf-5.
2193 * doc/invoke.texi (gvariable-location-views): New.
2194 (gvariable-location-views=incompat5): New.
2195 (gno-variable-location-views): New.
2197 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2199 PR tree-optimization/84136
2200 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2201 that the result of find_edge is non-NULL.
2203 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2206 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2207 storing integer register in SImode. Fix cost of 256 and 512
2208 byte aligned SSE register store.
2210 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2212 * config/i386/i386.c (ix86_multiplication_cost): Fix
2213 multiplication cost for TARGET_AVX512DQ.
2215 2018-02-08 Marek Polacek <polacek@redhat.com>
2217 PR tree-optimization/84238
2218 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2221 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2223 PR tree-optimization/84265
2224 * tree-vect-stmts.c (vectorizable_store): Don't treat
2225 VMAT_CONTIGUOUS accesses as grouped.
2226 (vectorizable_load): Likewise.
2228 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2230 PR tree-optimization/81635
2231 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2232 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2233 (test_round_for_mask): New functions.
2234 (wide_int_cc_tests): Call test_round_for_mask.
2235 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2236 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2237 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2238 range returned by get_range_info.
2240 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2243 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2244 * symtab.c: Include builtins.h
2245 (symtab_node::output_to_lto_symbol_table_p): Move here
2246 from lto-streamer-out.c:output_symbol_p.
2247 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2248 (output_symbol_p): Move all logic to symtab.c
2249 (produce_symtab): Update.
2251 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2253 * config/s390/s390-opts.h (enum indirect_branch): Define.
2254 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2255 (s390_indirect_branch_via_thunk)
2256 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2257 (enum s390_indirect_branch_type): Define.
2258 * config/s390/s390.c (struct s390_frame_layout, struct
2259 machine_function): Remove.
2260 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2261 (indirect_branch_table_label_no, indirect_branch_table_name):
2263 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2264 (enum s390_indirect_branch_option): Define.
2265 (s390_return_addr_from_memory): New function.
2266 (s390_handle_string_attribute): New function.
2267 (s390_attribute_table): Add new attribute handler.
2268 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2269 (s390_indirect_branch_via_thunk): New function.
2270 (s390_indirect_branch_via_inline_thunk): New function.
2271 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2272 sibling call optimization for non z10 compiles.
2273 (s390_emit_call): Force indirect branch target to be a single
2274 register. Add r1 clobber for non-z10 compiles.
2275 (s390_emit_epilogue): Emit return jump via return_use expander.
2276 (s390_reorg): Handle JUMP_INSNs as execute targets.
2277 (s390_option_override_internal): Perform validity checks for the
2278 new command line options.
2279 (s390_indirect_branch_attrvalue): New function.
2280 (s390_indirect_branch_settings): New function.
2281 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2282 (s390_output_indirect_thunk_function): New function.
2283 (s390_code_end): Implement target hook.
2284 (s390_case_values_threshold): Implement target hook.
2285 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2287 * config/s390/s390.h (struct s390_frame_layout)
2288 (struct machine_function): Move here from s390.c.
2289 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2290 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2291 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2292 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2293 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2294 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2295 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2296 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2297 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2298 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2299 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2300 (mnemonic attribute): Add values which aren't recognized
2302 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2303 pattern for branch conversion. Fix mnemonic attribute.
2304 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2305 indirect branch via thunk if requested.
2306 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2307 ("*indirect_jump"): Disable for branch conversion using out of
2309 ("indirect_jump_via_thunk<mode>_z10")
2310 ("indirect_jump_via_thunk<mode>")
2311 ("indirect_jump_via_inlinethunk<mode>_z10")
2312 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2313 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2314 ("casesi_jump_via_inlinethunk<mode>_z10")
2315 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2316 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2317 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2318 ("*indirect2_jump"): Disable for branch conversion.
2319 ("casesi_jump"): Turn into expander and expand patterns for branch
2321 ("return_use"): New expander.
2322 ("*return"): Emit return via thunk and rename it to ...
2323 ("*return<mode>"): ... this one.
2324 * config/s390/s390.opt: Add new options and and enum for the
2327 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2329 * lra-constraints.c (match_reload): Unconditionally use
2330 gen_lowpart_SUBREG, rather than selecting between that
2331 and equivalent gen_rtx_SUBREG code.
2333 2018-02-08 Richard Biener <rguenther@suse.de>
2335 PR tree-optimization/84233
2336 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2337 changed flag instead of boguously re-using phi_inserted.
2339 2018-02-08 Martin Jambor <mjambor@suse.cz>
2341 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2342 static local variables.
2344 2018-02-08 Richard Biener <rguenther@suse.de>
2346 PR tree-optimization/84278
2347 * tree-vect-stmts.c (vectorizable_store): When looking for
2348 smaller vector types to perform grouped strided loads/stores
2349 make sure the mode is supported by the target.
2350 (vectorizable_load): Likewise.
2352 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2354 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2355 Increase LDP/STP opportunities by adding adjacent callee-saves.
2357 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2359 PR rtl-optimization/84068
2360 PR rtl-optimization/83459
2361 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2363 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2365 PR tree-optimization/84224
2366 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2367 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2370 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2373 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2374 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2375 count, remove one USE.
2377 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2379 * doc/install.texi (Configuration): Document the
2380 --with-long-double-format={ibm,ieee} PowerPC configuration
2384 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2385 Convert from define_expand to be define_insn_and_split. Rework
2386 float/double/_Float128 conversions to QI/HI/SImode to work with
2387 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2388 conversions to QI/HImode types did a store and then a load to
2389 truncate the value. For conversions to VSX registers, don't split
2390 the insn, instead emit the code directly. Use the code iterator
2391 any_fix to combine signed and unsigned conversions.
2392 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2393 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2394 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2395 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2396 (fix_<mode>di2_hw): Likewise.
2397 (fixuns_<mode>di2_hw): Likewise.
2398 (fix_<mode>si2_hw): Likewise.
2399 (fixuns_<mode>si2_hw): Likewise.
2400 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2401 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2402 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2403 fix<uns>_trunc<SFDF:mode>si2_p8.
2404 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2406 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2407 (fix<uns>_<mode>_mem): Likewise.
2408 (fctiw<u>z_<mode>_mem): Likewise.
2409 (fix<uns>_<mode>_mem): Likewise.
2410 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2411 the register allocator from doing a direct move to the GPRs to do
2412 a store, and instead use the ISA 3.0 store byte/half-word from
2413 vector register instruction. For IEEE 128-bit floating point,
2414 also optimize stores of 32-bit ints.
2415 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2417 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2419 * genextract.c (push_pathstr_operand): New function to support
2421 (walk_rtx): Call push_pathstr_operand.
2422 (print_path): Support [a-zA-Z].
2424 2018-02-07 Richard Biener <rguenther@suse.de>
2426 PR tree-optimization/84037
2427 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2428 (cse_and_gimplify_to_preheader): Declare.
2429 (vect_get_place_in_interleaving_chain): Likewise.
2430 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2432 (_loop_vec_info::~_loop_vec_info): Delete it.
2433 (cse_and_gimplify_to_preheader): New function.
2434 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2435 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2436 (vectorizable_load): Likewise. For grouped stores always base
2437 the IV on the first element.
2438 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2439 condition before gimplifying.
2441 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2443 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2444 *DIV_EXPR and *MOD_EXPR.
2446 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2449 * config/i386/i386.c (ix86_option_override_internal): Mask out
2450 the CF_SET bit when checking -fcf-protection.
2452 2018-02-07 Tom de Vries <tom@codesourcery.com>
2455 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2458 2018-02-07 Richard Biener <rguenther@suse.de>
2460 PR tree-optimization/84204
2461 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2464 PR tree-optimization/84205
2465 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2466 special-case isl_ast_op_zdiv_r.
2468 PR tree-optimization/84223
2469 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2470 Only add conditions from within the region.
2471 (gather_bbs::after_dom_children): Adjust.
2473 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2476 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2477 * config/avr/avr.md: Only post-reload split REG-REG moves if
2478 either register is GENERAL_REG_P.
2480 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2482 PR tree-optimization/84235
2483 * tree-ssa-scopedtables.c
2484 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2485 if the subtraction is performed in floating point type where NaNs are
2486 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2487 build 1. Formatting fix.
2489 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2492 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2493 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2494 and skip it regardless of bb boundaries. Use CALL_P macro,
2495 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2496 unnecessarily, formatting fix.
2498 2018-02-06 Michael Collison <michael.collison@arm.com>
2500 * config/arm/thumb2.md:
2501 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2502 (*thumb_mov_notscc): Ditto.
2504 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2507 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2508 unsigned_fix, not "s".
2510 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2512 * configure.ac (gcc_fn_eh_frame_ro): New function.
2513 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2514 correct .eh_frame permissions.
2515 * configure: Regenerate.
2517 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2519 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2522 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2524 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2525 Display warning message for -mno-speculate-indirect-jumps.
2527 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2529 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2531 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2533 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2535 PR tree-optimization/84225
2536 * tree-eh.c (find_trapping_overflow): Only call
2537 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2539 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2542 * config/i386/i386.c: Reimplement the check of possible options
2543 -mibt/-mshstk conbination. Change error messages.
2544 * doc/invoke.texi: Fix a typo: remove extra '='.
2546 2018-02-06 Marek Polacek <polacek@redhat.com>
2548 PR tree-optimization/84228
2549 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2551 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2554 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2555 emitted arch directives.
2556 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2557 __ARM_FEATURE_COPROC before changing architectures.
2559 2018-02-06 Richard Biener <rguenther@suse.de>
2561 * config/i386/i386.c (print_reg): Fix typo.
2562 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2564 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2566 * configure: Regenerate.
2568 2018-02-05 Martin Sebor <msebor@redhat.com>
2570 PR tree-optimization/83369
2571 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2574 2018-02-05 Martin Liska <mliska@suse.cz>
2576 * doc/invoke.texi: Cherry-pick upstream r323995.
2578 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2580 * ira.c (ira_init_register_move_cost): Adjust comment.
2582 2018-02-05 Martin Liska <mliska@suse.cz>
2584 PR gcov-profile/84137
2585 * doc/gcov.texi: Fix typo in documentation.
2587 2018-02-05 Martin Liska <mliska@suse.cz>
2589 PR gcov-profile/83879
2590 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2591 using dlopen functionality.
2593 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2595 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2596 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2597 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2598 _mm_maskz_range_round_ss): New intrinsics.
2599 (__builtin_ia32_rangesd128_round)
2600 (__builtin_ia32_rangess128_round): Remove.
2601 (__builtin_ia32_rangesd128_mask_round,
2602 __builtin_ia32_rangess128_mask_round): New builtins.
2603 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2604 __builtin_ia32_rangess128_round): Remove.
2605 (__builtin_ia32_rangesd128_mask_round,
2606 __builtin_ia32_rangess128_mask_round): New builtins.
2607 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2608 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2609 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2610 "<round_saeonly_constraint>")): Changed to ...
2611 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2612 "<round_saeonly_scalar_constraint>")): ... this.
2613 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2614 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2615 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2616 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2617 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2619 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2621 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2623 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2624 Remove all values except native, 8540 and 8548.
2626 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2628 * config/i386/i386.c (ix86_output_function_return): Pass
2629 INVALID_REGNUM, instead of -1, as invalid register number to
2630 indirect_thunk_name and output_indirect_thunk.
2632 2018-02-02 Julia Koval <julia.koval@intel.com>
2634 * config.gcc: Add -march=icelake.
2635 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2636 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2637 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2638 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2639 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2640 (processor_target_table): Add icelake.
2641 (ix86_option_override_internal): Handle new PTAs.
2642 (get_builtin_code_for_version): Handle icelake.
2643 (M_INTEL_COREI7_ICELAKE): New.
2644 (fold_builtin_cpu): Handle icelake.
2645 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2646 * doc/invoke.texi: Add -march=icelake.
2648 2018-02-02 Julia Koval <julia.koval@intel.com>
2650 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2651 to wide_int_bitmask.
2652 * wide-int-bitmask.h: New.
2654 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2657 * config/i386/i386.md: Replace Pmode with word_mode in
2658 builtin_setjmp_setup and builtin_longjmp to support x32.
2660 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2664 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2666 (rs6000_supported_cpu_names): New static variable.
2667 (linux_cpu_translation_table): Likewise.
2668 (elf_platform) <cpu>: Define new static variable and use it.
2669 Translate kernel AT_PLATFORM name to canonical name if needed.
2670 Error if platform name is unknown.
2672 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2675 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2677 2018-02-01 Jeff Law <law@redhat.com>
2680 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2681 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2682 the scratch if RELEASE_VIA_POP is false.
2683 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2684 If we have to save a temporary register, decrement SIZE appropriately.
2685 Pass new arguments to release_scratch_register_on_entry.
2686 (ix86_adjust_stack_and_probe): Likewise.
2687 (ix86_emit_probe_stack_range): Pass new arguments to
2688 release_scratch_register_on_entry.
2690 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2692 PR rtl-optimization/84157
2693 * combine.c (change_zero_ext): Use REG_P predicate in
2694 front of HARD_REGISTER_P predicate.
2696 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2698 * config/avr/avr.c (avr_option_override): Move disabling of
2699 -fdelete-null-pointer-checks to...
2700 * common/config/avr/avr-common.c (avr_option_optimization_table):
2703 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2705 PR tree-optimization/81635
2706 * tree-data-ref.c (split_constant_offset_1): For types that
2707 wrap on overflow, try to use range info to prove that wrapping
2710 2018-02-01 Renlin Li <renlin.li@arm.com>
2713 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2715 (aarch64_register_move_cost): Likewise.
2716 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2718 (REG_CLASS_NAMES): Likewise.
2719 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2720 TAILCALL_ADDR_REGS. Remove IP registers.
2721 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2723 2018-02-01 Richard Biener <rguenther@suse.de>
2725 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2726 for specifying RPO order and allow NULL for that.
2727 * domwalk.c (dom_walker::dom_walker): Likewise.
2728 (dom_walker::walk): Handle NULL RPO order.
2729 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2731 (rewrite_update_dom_walker): Likewise.
2732 (mark_def_dom_walker): Likewise.
2734 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2736 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2737 (aarch64_maybe_expand_sve_subreg_move): Declare.
2738 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2739 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2741 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2742 that are semantically a reverse operation.
2743 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2744 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2745 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2747 (aarch64_can_change_mode_class): For big-endian, forbid changes
2748 between two SVE modes if they have different element sizes.
2750 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2752 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2753 the TImode handling for big-endian targets.
2755 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2757 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2758 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2760 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2761 Remove BSWAP handing for big-endian targets and use the form of
2762 LD1RQ appropariate for the mode.
2764 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2766 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2767 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2770 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2773 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2774 check for operands that need to go through aarch64_sve_reload_be.
2776 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2778 PR tree-optimization/81661
2779 PR tree-optimization/84117
2780 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2781 * tree-eh.c: Include gimplify.h.
2782 (find_trapping_overflow, replace_trapping_overflow,
2783 rewrite_to_non_trapping_overflow): New functions.
2784 * tree-vect-loop.c: Include tree-eh.h.
2785 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2786 * tree-data-ref.c: Include tree-eh.h.
2787 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2789 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2791 PR rtl-optimization/84123
2792 * combine.c (change_zero_ext): Check if hard register satisfies
2793 can_change_dest_mode before calling gen_lowpart_SUBREG.
2795 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2798 * ira.c (ira_init_register_move_cost): Remove assert.
2800 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2802 PR rtl-optimization/84071
2803 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2804 * doc/tm.texi: Regenerate.
2806 2018-01-31 Richard Biener <rguenther@suse.de>
2808 PR tree-optimization/84132
2809 * tree-data-ref.c (analyze_miv_subscript): Properly
2810 check whether evolution_function_is_affine_multivariate_p
2811 before calling gcd_of_steps_may_divide_p.
2813 2018-01-31 Julia Koval <julia.koval@intel.com>
2816 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2817 * config/i386/i386.md (rdpid_rex64) New.
2818 (rdpid): Make 32bit only.
2820 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2823 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2824 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2826 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2829 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2831 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2833 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2835 PR rtl-optimization/84071
2836 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2837 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2839 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2841 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2842 (arc_attribute_table): Add 'aux' attribute.
2843 (arc_in_small_data_p): Consider aux like variables.
2844 (arc_is_aux_reg_p): New function.
2845 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2846 (arc_get_aux_arg): New function.
2847 (prepare_move_operands): Handle aux-register access.
2848 (arc_handle_aux_attribute): New function.
2849 * doc/extend.texi (ARC Variable attributes): Add subsection.
2851 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2853 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2854 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2855 (arc_attribute_table): Add 'uncached' attribute.
2856 (arc_print_operand): Print '.di' flag for uncached memory
2858 (arc_in_small_data_p): Do not consider for small data the uncached
2860 (arc_is_uncached_mem_p): New function.
2861 * config/arc/predicates.md (compact_store_memory_operand): Check
2862 for uncached memory accesses.
2863 (nonvol_nonimm_operand): Likewise.
2864 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2866 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2869 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2870 falign-loops=): Add Optimization flag.
2872 2018-01-30 Jeff Law <law@redhat.com>
2875 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2876 INT_REGISTERS_SAVED. Check it prior to calling
2877 get_scratch_register_on_entry.
2878 (ix86_adjust_stack_and_probe): Similarly.
2879 (ix86_emit_probe_stack_range): Similarly.
2880 (ix86_expand_prologue): Corresponding changes.
2882 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2885 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2886 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2888 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2891 * lra-constraints.c (curr_insn_transform): Process AND in the
2894 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2896 PR rtl-optimization/83986
2897 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2898 dependence against last_pending_memory_flush in addition to
2901 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2903 PR tree-optimization/81611
2904 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2907 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2910 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2913 2018-01-30 Richard Biener <rguenther@suse.de>
2914 Jakub Jelinek <jakub@redhat.com>
2916 PR tree-optimization/84111
2917 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2918 inner loops added during recursion, as they don't have up-to-date
2921 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2924 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2925 (can_inline_edge_by_limits_p): ... here.
2926 (can_early_inline_edge_p, check_callers,
2927 update_caller_keys, update_callee_keys, recursive_inlining,
2928 add_new_edges_to_heap, speculation_useful_p,
2929 inline_small_functions,
2930 inline_small_functions, flatten_function,
2931 inline_to_all_callers_1): Update.
2933 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2935 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2938 2018-01-30 Richard Biener <rguenther@suse.de>
2940 PR tree-optimization/83008
2941 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2942 invariant and constant vector uses in stmts when they need
2945 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2948 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2949 * configure: Regenerate.
2951 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2953 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2955 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2956 Use gen_rtx_REG rather than gen_lowpart.
2958 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2960 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2961 rather than 0 when creating partial subregs.
2963 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2965 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2968 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2971 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2972 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2973 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2974 flags. This restores the settings used before the 2017-07-24.
2975 Turning off pre increment/decrement/modify allows IVOPTS to
2976 optimize DF/SF loops where the index is an int.
2978 2018-01-29 Richard Biener <rguenther@suse.de>
2979 Kelvin Nilsen <kelvin@gcc.gnu.org>
2982 * tree-vect-stmts.c (vectorizable_call): Don't call
2983 targetm.vectorize_builtin_md_vectorized_function if callee is
2986 2018-01-22 Carl Love <cel@us.ibm.com>
2988 * doc/extend.tex: Fix typo in second arg in
2989 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2991 2018-01-29 Richard Biener <rguenther@suse.de>
2993 PR tree-optimization/84086
2994 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2995 (flush_ssaname_freelist): When SSA names were released reset
2996 the SCEV hash table.
2998 2018-01-29 Richard Biener <rguenther@suse.de>
3000 PR tree-optimization/84057
3001 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3002 removed paths when removing edges.
3004 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3006 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3007 -mfunction-return=@var{choice}.
3009 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3012 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3014 (layout::print_source_line): Likewise.
3015 (test_get_line_width_without_trailing_whitespace): Add test cases.
3017 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3020 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3023 2018-01-26 Jim Wilson <jimw@sifive.com>
3025 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3027 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3030 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3032 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3033 and CMP + SUB-immediate -> SUBS.
3035 2018-01-26 Martin Sebor <msebor@redhat.com>
3037 PR tree-optimization/83896
3038 * tree-ssa-strlen.c (get_string_len): Rename...
3039 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3040 Avoid assuming length is constant.
3041 (handle_char_store): Use HOST_WIDE_INT for string length.
3043 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3046 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3047 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3049 2018-01-26 Richard Biener <rguenther@suse.de>
3051 PR rtl-optimization/84003
3052 * dse.c (record_store): Only record redundant stores when
3053 the earlier store aliases at least all accesses the later one does.
3055 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3057 PR rtl-optimization/83985
3058 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3059 REG_CFA_RESTORE insns.
3060 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3061 REG_CFA_RESTORE insns here.
3064 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3065 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3067 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3069 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3070 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3071 (arc_init): Likewise.
3072 (arc_override_options): Likewise.
3073 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3075 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3077 * config/arc/arc.h (TARGET_DBNZ): Define.
3078 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3079 properly set the tune attribute.
3080 (dbnz): Use TARGET_DBNZ guard.
3081 * config/arc/arc.opt (mtune): Add core3 option.
3083 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3085 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3086 recognize new pic like addresses.
3087 (arc_delegitimize_address): Clean up.
3089 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3091 * config/arc/arc-arches.def: Option mrf16 valid for all
3093 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3094 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3095 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3096 * config/arc/arc-tables.opt: Regenerate.
3097 * config/arc/arc.c (arc_conditional_register_usage): Handle
3098 reduced register file case.
3099 (arc_file_start): Set must have build attributes.
3100 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3102 * config/arc/arc.opt (mrf16): Add new option.
3103 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3104 * config/arc/genmultilib.awk: Handle new mrf16 option.
3105 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3106 * config/arc/t-multilib: Regenerate.
3107 * doc/invoke.texi (ARC Options): Document mrf16 option.
3109 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3111 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3112 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3113 (arc_attribute_table): Add 'secure_call' attribute.
3114 (arc_print_operand): Print secure call operand.
3115 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3117 (arc_is_secure_call_p): New function. * config/arc/arc.md
3118 (call_i): Add support for sjli instruction.
3119 (call_value_i): Likewise.
3120 * config/arc/constraints.md (Csc): New constraint.
3122 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3123 John Eric Martin <John.Martin@emmicro-us.com>
3125 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3126 * config/arc/arc.c (_arc_jli_section): New struct.
3127 (arc_jli_section): New type.
3128 (rc_jli_sections): New static variable.
3129 (arc_handle_jli_attribute): New function.
3130 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3131 (arc_file_end): New function.
3132 (TARGET_ASM_FILE_END): Define.
3133 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3134 (arc_add_jli_section): New function.
3135 (jli_call_scan): Likewise.
3136 (arc_reorg): Call jli_call_scan.
3137 (arc_output_addsi): Remove 'S' from printing asm operand.
3138 (arc_is_jli_call_p): New function.
3139 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3141 (movhi_insn): Likewise.
3142 (movsi_insn): Likewise.
3143 (movsi_set_cc_insn): Likewise.
3144 (loadqi_update): Likewise.
3145 (load_zeroextendqisi_update): Likewise.
3146 (load_signextendqisi_update): Likewise.
3147 (loadhi_update): Likewise.
3148 (load_zeroextendhisi_update): Likewise.
3149 (load_signextendhisi_update): Likewise.
3150 (loadsi_update): Likewise.
3151 (loadsf_update): Likewise.
3152 (movsicc_insn): Likewise.
3153 (bset_insn): Likewise.
3154 (bxor_insn): Likewise.
3155 (bclr_insn): Likewise.
3156 (bmsk_insn): Likewise.
3157 (bicsi3_insn): Likewise.
3158 (cmpsi_cc_c_insn): Likewise.
3159 (movsi_ne): Likewise.
3160 (movsi_cond_exec): Likewise.
3161 (clrsbsi2): Likewise.
3172 (call_i): Remove 'S' asm letter, add jli instruction.
3173 (call_value_i): Likewise.
3174 * config/arc/arc.op (mjli-always): New option.
3175 * config/arc/constraints.md (Cji): New constraint.
3176 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3178 (subsf3_fpx): Likewise.
3179 (mulsf3_fpx): Likewise.
3180 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3182 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3184 * doc/invoke.texi (ARC): Document mjli-always option.
3186 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3188 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3189 avoid addition with 0 and use incw and decw where possible.
3191 2018-01-26 Richard Biener <rguenther@suse.de>
3193 PR tree-optimization/81082
3194 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3195 association if it requires casting to unsigned.
3196 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3197 from fold_plusminus_mult_expr to catch important cases late when
3198 range info is available.
3200 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3202 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3203 * configure.ac (hidden_linkonce): New test.
3204 * configure: Regenerate.
3205 * config.in: Regenerate.
3207 2018-01-26 Julia Koval <julia.koval@intel.com>
3209 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3210 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3211 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3212 _mm_mask_bitshuffle_epi64_mask): Fix type.
3213 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3214 USI_FTYPE_V4DI_V4DI_USI): Remove.
3215 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3216 __builtin_ia32_vpshufbitqmb256_mask,
3217 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3218 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3219 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3221 2018-01-26 Alan Modra <amodra@gmail.com>
3224 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3225 UNSPEC_VBPERMQ. Sort other unspecs.
3227 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3229 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3231 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3234 * predict.c (drop_profile): Do not push/pop cfun; update also
3236 (handle_missing_profiles): Fix logic looking for zero profiles.
3238 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3241 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3242 on functions with #pragma omp declare simd or functions with simd
3244 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3245 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3246 Remove trailing \n from warning_at calls.
3248 2018-01-25 Tom de Vries <tom@codesourcery.com>
3251 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3252 for neutered workers.
3254 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3257 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3258 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3260 2018-01-24 Jeff Law <law@redhat.com>
3263 * i386.c (get_probe_interval): Move to earlier point.
3264 (ix86_compute_frame_layout): If -fstack-clash-protection and
3265 the frame is larger than the probe interval, then use pushes
3266 to save registers rather than reg->mem moves.
3267 (ix86_expand_prologue): Remove conditional for int_registers_saved
3270 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3273 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3274 min/max for never referenced object.
3276 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3279 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3281 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3282 attributes from DECL_ATTRIBUTES (decl) without affecting
3283 DECL_ATTRIBUTES (current_function_decl).
3284 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3285 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3287 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3289 PR tree-optimization/83979
3290 * fold-const.c (fold_comparison): Use constant_boolean_node
3291 instead of boolean_{true,false}_node.
3293 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3295 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3298 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3300 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3301 Simplify the clause that sets the length attribute.
3302 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3303 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3304 clause that sets the length attribute.
3305 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3307 2018-01-24 Tom de Vries <tom@codesourcery.com>
3310 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3311 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3312 Add strict parameter.
3313 (prevent_branch_around_nothing): Insert dummy insn between branch to
3314 label and label with no ptx insn inbetween.
3315 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3317 2018-01-24 Tom de Vries <tom@codesourcery.com>
3320 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3321 for neutered threads in warp.
3322 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3324 2018-01-24 Richard Biener <rguenther@suse.de>
3326 PR tree-optimization/83176
3327 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3330 2018-01-24 Richard Biener <rguenther@suse.de>
3332 PR tree-optimization/82819
3333 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3334 code generating pluses that are no-ops in the target precision.
3336 2018-01-24 Richard Biener <rguenther@suse.de>
3339 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3341 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3343 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3344 to merge probabilities.
3345 * predict.c (probably_never_executed): Also mark as cold functions
3346 with global 0 profile and guessed local profile.
3347 * profile-count.c (profile_probability::combine_with_count): New
3349 * profile-count.h (profile_probability::operator*,
3350 profile_probability::operator*=, profile_probability::operator/,
3351 profile_probability::operator/=): Reduce precision to adjusted
3352 and set value to guessed on contradictory divisions.
3353 (profile_probability::combine_with_freq): Remove.
3354 (profile_probability::combine_wiht_count): Declare.
3355 (profile_count::force_nonzero):: Set to adjusted.
3356 (profile_count::probability_in):: Set quality to adjusted.
3357 * tree-ssa-tail-merge.c (replace_block_by): Use
3360 2018-01-23 Andrew Waterman <andrew@sifive.com>
3361 Jim Wilson <jimw@sifive.com>
3363 * config/riscv/riscv.c (riscv_stack_boundary): New.
3364 (riscv_option_override): Set riscv_stack_boundary. Handle
3365 riscv_preferred_stack_boundary_arg.
3366 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3367 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3368 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3369 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3370 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3371 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3373 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3376 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3377 of struct ix86_frame.
3378 (ix86_expand_epilogue): Likewise. Add a local variable for
3379 the reg_save_offset field in struct ix86_frame.
3381 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3383 PR tree-optimization/82604
3384 * tree-loop-distribution.c (enum partition_kind): New enum item
3385 PKIND_PARTIAL_MEMSET.
3386 (partition_builtin_p): Support above new enum item.
3387 (generate_code_for_partition): Ditto.
3388 (compute_access_range): Differentiate cases that equality can be
3389 proven at all loops, the innermost loops or no loops.
3390 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3391 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3392 (finalize_partitions, distribute_loop): Don't fuse partition of
3393 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3394 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3397 2018-01-23 Martin Liska <mliska@suse.cz>
3399 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3400 order to ignore the predictor.
3401 (PRED_POLYMORPHIC_CALL): Likewise.
3402 (PRED_RECURSIVE_CALL): Likewise.
3404 2018-01-23 Martin Liska <mliska@suse.cz>
3406 * tree-profile.c (tree_profiling): Print function header to
3407 aware reader which function we are working on.
3408 * value-prof.c (gimple_find_values_to_profile): Do not print
3409 not interesting value histograms.
3411 2018-01-23 Martin Liska <mliska@suse.cz>
3413 * profile-count.h (enum profile_quality): Add
3414 profile_uninitialized as the first value. Do not number values
3415 as they are zero based.
3416 (profile_count::verify): Update sanity check.
3417 (profile_probability::verify): Likewise.
3419 2018-01-23 Nathan Sidwell <nathan@acm.org>
3421 * doc/invoke.texi (ffor-scope): Deprecate.
3423 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3425 PR tree-optimization/83510
3426 * domwalk.c (set_all_edges_as_executable): New function.
3427 (dom_walker::dom_walker): Convert bool param
3428 "skip_unreachable_blocks" to enum reachability. Move setup of
3429 edge flags to set_all_edges_as_executable and only do it when
3430 reachability is REACHABLE_BLOCKS.
3431 * domwalk.h (enum dom_walker::reachability): New enum.
3432 (dom_walker::dom_walker): Convert bool param
3433 "skip_unreachable_blocks" to enum reachability.
3434 (set_all_edges_as_executable): New decl.
3435 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3436 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3438 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3439 but converting true to REACHABLE_BLOCKS.
3440 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3442 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3443 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3444 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3446 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3447 if check_all_array_refs will be called.
3449 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3451 * tree.c (selftest::test_location_wrappers): Add more test
3454 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3456 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3457 (selftest::test_bit_in_range): Likewise.
3459 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3462 * doc/sourcebuild.texi (vect_float): Say that the selector
3463 only describes the situation when -funsafe-math-optimizations is on.
3464 (vect_float_strict): Document.
3466 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3468 PR tree-optimization/83965
3469 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3470 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3471 instead of checking only for a reduction.
3472 (vect_recog_widen_sum_pattern): Likewise.
3474 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3476 * predict.c (probably_never_executed): Only use precise profile info.
3477 (compute_function_frequency): Skip after inlining hack since we now
3478 have quality checking.
3480 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3482 * profile-count.h (profile_probability::very_unlikely,
3483 profile_probability::unlikely, profile_probability::even): Set
3484 precision to guessed.
3486 2018-01-23 Richard Biener <rguenther@suse.de>
3488 PR tree-optimization/83963
3489 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3490 Properly terminate dominator walk when crossing the exit edge not
3491 when visiting its source block.
3493 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3496 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3497 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3499 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3501 PR tree-optimization/83957
3502 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3503 semicolon after for body surrounded by braces.
3505 PR tree-optimization/83081
3506 * profile-count.h (profile_probability::split): New method.
3507 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3508 Use profile_probability::split.
3509 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3510 when splitting a single conditional jump into 2.
3512 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3514 PR tree-optimization/69452
3515 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3518 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3520 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3521 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3522 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3524 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3526 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3527 * config/rl78/rl78.md (movdi): New define_expand.
3528 * config/rl78/rl78.c (rl78_split_movdi): New function.
3530 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3533 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3535 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3536 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3537 128-bit to produce an UNSPEC move to get the double word with the
3538 signbit and then a shift directly to do signbit.
3539 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3540 implementation with a new version that just does either a direct
3541 move or a regular move. Move memory interface to separate insns.
3542 Move insns so they are next to the expander.
3543 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3544 with signbit move. Split big and little endian case.
3545 (signbit<mode>2_dm_mem_le): Likewise.
3546 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3547 (signbit<mode>2_dm2): Likewise.
3549 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3551 * config/rl78/rl78.md (anddi3): New define_expand.
3553 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3555 * config/rl78/rl78.md (umindi3): New define_expand.
3557 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3559 * config/rl78/rl78.md (smindi3): New define_expand.
3561 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3563 * config/rl78/rl78.md (smaxdi3): New define_expand.
3565 2018-01-22 Carl Love <cel@us.ibm.com>
3567 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3568 LVX_V1TI): Add macro expansion.
3569 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3570 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3571 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3572 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3573 Change check to determine if the instruction is a byte reversing
3574 entry. Fix typo in comment.
3575 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3576 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3577 Add def_builtin calls for new builtins.
3578 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3579 Add define_insn expansion.
3581 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3583 * config/rl78/rl78.md (umaxdi3): New define_expand.
3585 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3587 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3588 for non-QImode registers.
3590 2018-01-22 Richard Biener <rguenther@suse.de>
3592 PR tree-optimization/83963
3593 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3594 including the loop exit block.
3595 (scop_detection::merge_sese): Likewise.
3596 (scop_detection::add_scop): Do it here instead.
3598 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3600 * doc/sourcebuild.texi (arm_softfloat): Document.
3602 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3605 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3606 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3607 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3609 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3610 David Edelsohn <dje.gcc@gmail.com>
3613 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3614 Change "crset eq" to "crset 2".
3615 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3616 (*call_indirect_aix<mode>_nospec): Likewise.
3617 (*call_value_indirect_aix<mode>_nospec): Likewise.
3618 (*call_indirect_elfv2<mode>_nospec): Likewise.
3619 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3620 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3621 change assembly output from . to $.
3622 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3623 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3624 (*tablejump<mode>_internal1_nospec): Likewise.
3626 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3629 * config/sh/sh_optimize_sett_clrt.cc:
3630 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3632 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3634 PR tree-optimization/83940
3635 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3636 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3637 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3638 use it to pass back the definition type.
3639 (vect_check_store_rhs): Likewise rhs_dt_out.
3640 (vect_build_gather_load_calls): Add a mask_dt argument and use
3641 it instead of a call to vect_is_simple_use.
3642 (vectorizable_store): Update calls to vect_check_load_store_mask
3643 and vect_check_store_rhs. Use the dt returned by the latter instead
3644 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3645 instead of calls to vect_is_simple_use. Pass the scalar rather
3646 than the vector operand to vect_is_simple_use when handling
3647 second and subsequent copies of an rhs value.
3648 (vectorizable_load): Update calls to vect_check_load_store_mask
3649 and vect_build_gather_load_calls. Use the cached mask_dt and
3650 gs_info.offset_dt instead of calls to vect_is_simple_use.
3652 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3655 * tree-emutls.c: Include gimplify.h.
3656 (lower_emutls_2): New function.
3657 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3658 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3659 it before further processing.
3662 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3663 UINTVAL (trueop1) instead of INTVAL (op1).
3665 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3669 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3670 INCOMING_FRAME_SP_OFFSET if not defined.
3671 (scan_trace): Add ENTRY argument. If true and
3672 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3673 emit a note to adjust the CFA offset.
3674 (create_cfi_notes): Adjust scan_trace callers.
3675 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3676 INCOMING_FRAME_SP_OFFSET in the CIE.
3677 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3678 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3680 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3681 * doc/tm.texi: Regenerated.
3683 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3685 PR rtl-optimization/83147
3686 * lra-constraints.c (remove_inheritance_pseudos): Use
3687 lra_substitute_pseudo_within_insn.
3689 2018-01-19 Tom de Vries <tom@codesourcery.com>
3690 Cesar Philippidis <cesar@codesourcery.com>
3693 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3695 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3698 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3699 spaces for function labels.
3701 2018-01-19 Martin Liska <mliska@suse.cz>
3703 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3704 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3705 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3706 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3707 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3708 (PRED_CONST_RETURN): Change from 69 to 65.
3709 (PRED_NULL_RETURN): Change from 91 to 71.
3710 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3711 (PRED_LOOP_GUARD): Change from 66 to 73.
3713 2018-01-19 Martin Liska <mliska@suse.cz>
3715 * predict.c (predict_insn_def): Add new assert.
3716 (struct branch_predictor): Change type to signed integer.
3717 (test_prediction_value_range): Amend test to cover
3719 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3720 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3721 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3722 (PRED_LOOP_IV_COMPARE): Likewise.
3723 * predict.h (PROB_UNINITIALIZED): Define new constant.
3725 2018-01-19 Martin Liska <mliska@suse.cz>
3727 * predict.c (dump_prediction): Add new format for
3728 analyze_brprob.py script which is enabled with -details
3730 * profile-count.h (precise_p): New function.
3732 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3734 PR tree-optimization/83922
3735 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3736 there are no statements that need masking.
3737 (vect_active_double_reduction_p): New function.
3738 (vect_analyze_loop_operations): Use it when handling phis that
3739 are not in the loop header.
3741 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3743 PR tree-optimization/83914
3744 * tree-vect-loop.c (vectorizable_induction): Don't convert
3745 init_expr or apply the peeling adjustment for inductions
3746 that are nested within the vectorized loop.
3748 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3750 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3753 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3757 * function.h (gimplify_parameters): Add gimple_seq * argument.
3758 * function.c: Include gimple.h and options.h.
3759 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3760 for the added local temporaries if needed.
3761 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3762 if there are any parameter cleanups, wrap whole body into a
3763 try/finally with the cleanups.
3765 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3768 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3769 Use GET_MODE_CLASS for scalar floating point.
3771 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3775 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3776 Fix call of call_cgraph_insertion_hooks.
3778 2018-01-18 Martin Sebor <msebor@redhat.com>
3780 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3782 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3785 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3788 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3791 * common.opt: (-ffile-prefix-map): New option.
3792 * opts.c (common_handle_option): Defer it.
3793 * opts-global.c (handle_common_deferred_options): Handle it.
3794 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3795 * file-prefix-map.h: New file.
3796 (remap_debug_filename, add_debug_prefix_map): ...here.
3797 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3798 * final.c (debug_prefix_map, add_debug_prefix_map
3799 remap_debug_filename): Move to...
3800 * file-prefix-map.c: New file.
3801 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3802 generalize, get rid of alloca(), use strrchr() instead of strchr().
3803 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3804 Implement in terms of add_prefix_map().
3805 (remap_macro_filename, remap_debug_filename): Implement in term of
3807 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3808 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3809 * dbxout.c: Include file-prefix-map.h.
3810 * varasm.c: Likewise.
3811 * vmsdbgout.c: Likewise.
3812 * xcoffout.c: Likewise.
3813 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3814 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3815 * doc/invoke.texi (-ffile-prefix-map): Document.
3816 (-fdebug-prefix-map): Update description.
3818 2018-01-18 Martin Liska <mliska@suse.cz>
3820 * config/i386/i386.c (indirect_thunk_name): Document that also
3822 (output_indirect_thunk): Document why both instructions
3823 (pause and lfence) are generated.
3825 2018-01-18 Richard Biener <rguenther@suse.de>
3827 PR tree-optimization/83887
3828 * graphite-scop-detection.c
3829 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3830 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3831 (scop_detection::merge_sese): Re-implement with a flood-fill
3832 algorithm that properly finds a SESE region if it exists.
3834 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3837 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3838 pointer_diff optimizations use view_convert instead of convert.
3840 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3842 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3843 Generate different code for -mno-speculate-indirect-jumps.
3844 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3845 (*call_indirect_aix<mode>): Disable for
3846 -mno-speculate-indirect-jumps.
3847 (*call_indirect_aix<mode>_nospec): New define_insn.
3848 (*call_value_indirect_aix<mode>): Disable for
3849 -mno-speculate-indirect-jumps.
3850 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3851 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3852 -mno-speculate-indirect-jumps.
3853 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3855 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3857 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3858 long double type, set the flags for noting the default long double
3859 type, even if we don't pass or return a long double type.
3861 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3864 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3867 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3869 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3870 support for merge[hl].
3871 (fold_mergehl_helper): New helper function.
3872 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3873 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3874 (altivec_vmrglw_direct): Add xxmrglw insn.
3876 2018-01-17 Andrew Waterman <andrew@sifive.com>
3878 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3879 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3881 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3884 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3885 call the lto_location_cache before reading the
3886 DECL_SOURCE_LOCATION of the types.
3888 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3889 Richard Sandiford <richard.sandiford@linaro.org>
3891 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3892 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3893 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3894 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3895 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3897 * config/aarch64/constraints.md (aarch64_movti_operand):
3899 * config/aarch64/predicates.md (Uti): Add new constraint.
3901 2018-01-17 Carl Love <cel@us.ibm.com>
3903 * config/rs6000/vsx.md (define_expand xl_len_r,
3904 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3905 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3907 (define_expand, define_insn): Move the shift left from the
3908 define_insn to the define_expand for lxvl and stxvl instructions.
3909 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3910 and XL_LEN_R definitions to PURE.
3912 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3914 * config/i386/i386.c (indirect_thunk_name): Declare regno
3915 as unsigned int. Compare regno with INVALID_REGNUM.
3916 (output_indirect_thunk): Ditto.
3917 (output_indirect_thunk_function): Ditto.
3918 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3919 in the call to output_indirect_thunk_function.
3921 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3924 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3925 rather than the size of inner_type to determine the stack slot size
3926 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3928 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3931 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3934 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3936 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3937 endian Linux systems to optionally enable multilibs for selecting
3938 the long double type if the user configured an explicit type.
3939 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3940 have no long double multilibs if not defined.
3941 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3942 warn if the user used -mabi={ieee,ibm}longdouble and we built
3943 multilibs for long double.
3944 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3945 appropriate multilib option.
3946 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3948 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3949 for building long double multilibs.
3950 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3952 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3954 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3957 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3959 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3962 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3965 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3968 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3970 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3971 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3973 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3975 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3976 different rtl trees depending on TARGET_64BIT.
3977 (rs6000_gen_lvx): Likewise.
3979 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3981 * config/visium/visium.md (nop): Tweak comment.
3982 (hazard_nop): Likewise.
3984 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3986 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3987 -mspeculate-indirect-jumps.
3988 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3989 for -mno-speculate-indirect-jumps.
3990 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3991 (*call_value_indirect_elfv2<mode>): Disable for
3992 -mno-speculate-indirect-jumps.
3993 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3994 (indirect_jump): Emit different RTL for
3995 -mno-speculate-indirect-jumps.
3996 (*indirect_jump<mode>): Disable for
3997 -mno-speculate-indirect-jumps.
3998 (*indirect_jump<mode>_nospec): New define_insn.
3999 (tablejump): Emit different RTL for
4000 -mno-speculate-indirect-jumps.
4001 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4002 (tablejumpsi_nospec): New define_expand.
4003 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4004 (tablejumpdi_nospec): New define_expand.
4005 (*tablejump<mode>_internal1): Disable for
4006 -mno-speculate-indirect-jumps.
4007 (*tablejump<mode>_internal1_nospec): New define_insn.
4008 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4011 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4013 * caller-save.c (insert_save): Drop unnecessary parameter. All
4016 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4017 Richard Biener <rguenth@suse.de>
4020 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4021 return early, inline manually is_gimple_sizepos. Make sure if we
4022 call gimplify_expr we don't end up with a gimple constant.
4023 * tree.c (variably_modified_type_p): Don't return true for
4024 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4025 * gimplify.h (is_gimple_sizepos): Remove.
4027 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4029 PR tree-optimization/83857
4030 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4031 vectorizable_live_operation for pure SLP statements.
4032 (vectorizable_live_operation): Handle PHIs.
4034 2018-01-16 Richard Biener <rguenther@suse.de>
4036 PR tree-optimization/83867
4037 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4038 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4040 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4043 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4044 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4045 If off is not INTEGER_CST, issue a may not be aligned warning
4046 rather than isn't aligned. Use isn%'t rather than isn't.
4047 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4049 <case MULT_EXPR>: Improve the case when bottom and one of the
4050 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4051 operand, in that case check if the other operand is multiple of
4052 bottom divided by the INTEGER_CST operand.
4054 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4057 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4058 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4059 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4060 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4061 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4062 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4063 (pa_function_arg_size): New function.
4065 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4067 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4068 in a separate statement.
4070 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4072 PR tree-optimization/83847
4073 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4074 group gathers and scatters.
4076 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4078 PR rtl-optimization/86620
4079 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4081 PR rtl-optimization/83213
4082 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4083 to last if both are JUMP_INSNs.
4085 PR tree-optimization/83843
4086 * gimple-ssa-store-merging.c
4087 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4088 store_immediate_info for bswap/nop orig_stores.
4090 2018-01-15 Andrew Waterman <andrew@sifive.com>
4092 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4094 <UDIV>: Increase cost if !TARGET_DIV.
4096 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4098 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4099 (define_attr "cr_logical_3op"): New.
4100 (cceq_ior_compare): Adjust.
4101 (cceq_ior_compare_complement): Adjust.
4102 (*cceq_rev_compare): Adjust.
4103 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4104 (is_cracked_insn): Adjust.
4105 (insn_must_be_first_in_group): Adjust.
4106 * config/rs6000/40x.md: Adjust.
4107 * config/rs6000/440.md: Adjust.
4108 * config/rs6000/476.md: Adjust.
4109 * config/rs6000/601.md: Adjust.
4110 * config/rs6000/603.md: Adjust.
4111 * config/rs6000/6xx.md: Adjust.
4112 * config/rs6000/7450.md: Adjust.
4113 * config/rs6000/7xx.md: Adjust.
4114 * config/rs6000/8540.md: Adjust.
4115 * config/rs6000/cell.md: Adjust.
4116 * config/rs6000/e300c2c3.md: Adjust.
4117 * config/rs6000/e500mc.md: Adjust.
4118 * config/rs6000/e500mc64.md: Adjust.
4119 * config/rs6000/e5500.md: Adjust.
4120 * config/rs6000/e6500.md: Adjust.
4121 * config/rs6000/mpc.md: Adjust.
4122 * config/rs6000/power4.md: Adjust.
4123 * config/rs6000/power5.md: Adjust.
4124 * config/rs6000/power6.md: Adjust.
4125 * config/rs6000/power7.md: Adjust.
4126 * config/rs6000/power8.md: Adjust.
4127 * config/rs6000/power9.md: Adjust.
4128 * config/rs6000/rs64.md: Adjust.
4129 * config/rs6000/titan.md: Adjust.
4131 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4133 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4134 ix86_indirect_branch_register logic.
4136 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4138 * config/i386/constraints.md (Bs): Update
4139 ix86_indirect_branch_register check. Don't check
4140 ix86_indirect_branch_register with GOT_memory_operand.
4142 * config/i386/predicates.md (GOT_memory_operand): Don't check
4143 ix86_indirect_branch_register here.
4144 (GOT32_symbol_operand): Likewise.
4146 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4148 * config/i386/predicates.md (constant_call_address_operand):
4149 Rewrite ix86_indirect_branch_register logic.
4150 (sibcall_insn_operand): Likewise.
4152 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4154 * config/i386/constraints.md (Bs): Replace
4155 ix86_indirect_branch_thunk_register with
4156 ix86_indirect_branch_register.
4158 * config/i386/i386.md (indirect_jump): Likewise.
4159 (tablejump): Likewise.
4160 (*sibcall_memory): Likewise.
4161 (*sibcall_value_memory): Likewise.
4162 Peepholes of indirect call and jump via memory: Likewise.
4163 * config/i386/i386.opt: Likewise.
4164 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4165 (GOT_memory_operand): Likewise.
4166 (call_insn_operand): Likewise.
4167 (sibcall_insn_operand): Likewise.
4168 (GOT32_symbol_operand): Likewise.
4170 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4173 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4174 type rather than type addr's type points to.
4175 (expand_omp_atomic_mutex): Likewise.
4176 (expand_omp_atomic): Likewise.
4178 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4181 * config/i386/i386.c (output_indirect_thunk_function): Use
4182 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4183 for __x86_return_thunk.
4185 2018-01-15 Richard Biener <rguenther@suse.de>
4188 * expmed.c (extract_bit_field_1): Fix typo.
4190 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4193 * config/arm/iterators.md (VF): New mode iterator.
4194 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4195 Remove integer-related logic from pattern.
4196 (neon_vabd<mode>_3): Likewise.
4198 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4201 * common.opt (fstrict-overflow): No longer an alias.
4202 (fwrapv-pointer): New option.
4203 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4204 also for pointer types based on flag_wrapv_pointer.
4205 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4206 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4207 opts->x_flag_wrapv got set.
4208 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4209 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4210 POINTER_TYPE_OVERFLOW_UNDEFINED.
4211 * match.pd: Likewise in address comparison pattern.
4212 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4214 2018-01-15 Richard Biener <rguenther@suse.de>
4217 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4218 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4219 Reset type names to their identifier if their TYPE_DECL doesn't
4220 have linkage (and thus is used for ODR and devirt).
4221 (save_debug_info_for_decl): Remove.
4222 (save_debug_info_for_type): Likewise.
4223 (add_tree_to_fld_list): Adjust.
4224 * tree-pretty-print.c (dump_generic_node): Make dumping of
4225 type names more robust.
4227 2018-01-15 Richard Biener <rguenther@suse.de>
4229 * BASE-VER: Bump to 8.0.1.
4231 2018-01-14 Martin Sebor <msebor@redhat.com>
4234 * builtins.c (check_access): Avoid warning when the no-warning bit
4237 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4239 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4240 * ira-color (allocno_hard_regs_compare): Likewise.
4242 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4245 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4246 Use .pushsection/.popsection.
4248 2018-01-14 Martin Sebor <msebor@redhat.com>
4251 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4253 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4255 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4256 entry from extra_headers.
4257 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4258 extra_headers, make the list bitwise identical to the i?86-*-* one.
4260 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4262 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4263 -mcmodel=large with -mindirect-branch=thunk,
4264 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4265 -mfunction-return=thunk-extern.
4266 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4267 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4268 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4270 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4272 * config/i386/i386.c (print_reg): Print the name of the full
4273 integer register without '%'.
4274 (ix86_print_operand): Handle 'V'.
4275 * doc/extend.texi: Document 'V' modifier.
4277 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4279 * config/i386/constraints.md (Bs): Disallow memory operand for
4280 -mindirect-branch-register.
4282 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4283 (GOT_memory_operand): Likewise.
4284 (call_insn_operand): Likewise.
4285 (sibcall_insn_operand): Likewise.
4286 (GOT32_symbol_operand): Likewise.
4287 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4288 for -mindirect-branch-register.
4289 (tablejump): Likewise.
4290 (*sibcall_memory): Likewise.
4291 (*sibcall_value_memory): Likewise.
4292 Disallow peepholes of indirect call and jump via memory for
4293 -mindirect-branch-register.
4294 (*call_pop): Replace m with Bw.
4295 (*call_value_pop): Likewise.
4296 (*sibcall_pop_memory): Replace m with Bs.
4297 * config/i386/i386.opt (mindirect-branch-register): New option.
4298 * doc/invoke.texi: Document -mindirect-branch-register option.
4300 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4302 * config/i386/i386-protos.h (ix86_output_function_return): New.
4303 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4304 set function_return_type.
4305 (indirect_thunk_name): Add ret_p to indicate thunk for function
4307 (output_indirect_thunk_function): Pass false to
4308 indirect_thunk_name.
4309 (ix86_output_indirect_branch_via_reg): Likewise.
4310 (ix86_output_indirect_branch_via_push): Likewise.
4311 (output_indirect_thunk_function): Create alias for function
4312 return thunk if regno < 0.
4313 (ix86_output_function_return): New function.
4314 (ix86_handle_fndecl_attribute): Handle function_return.
4315 (ix86_attribute_table): Add function_return.
4316 * config/i386/i386.h (machine_function): Add
4317 function_return_type.
4318 * config/i386/i386.md (simple_return_internal): Use
4319 ix86_output_function_return.
4320 (simple_return_internal_long): Likewise.
4321 * config/i386/i386.opt (mfunction-return=): New option.
4322 (indirect_branch): Mention -mfunction-return=.
4323 * doc/extend.texi: Document function_return function attribute.
4324 * doc/invoke.texi: Document -mfunction-return= option.
4326 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4328 * config/i386/i386-opts.h (indirect_branch): New.
4329 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4330 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4331 with local indirect jump when converting indirect call and jump.
4332 (ix86_set_indirect_branch_type): New.
4333 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4334 (indirectlabelno): New.
4335 (indirect_thunk_needed): Likewise.
4336 (indirect_thunk_bnd_needed): Likewise.
4337 (indirect_thunks_used): Likewise.
4338 (indirect_thunks_bnd_used): Likewise.
4339 (INDIRECT_LABEL): Likewise.
4340 (indirect_thunk_name): Likewise.
4341 (output_indirect_thunk): Likewise.
4342 (output_indirect_thunk_function): Likewise.
4343 (ix86_output_indirect_branch_via_reg): Likewise.
4344 (ix86_output_indirect_branch_via_push): Likewise.
4345 (ix86_output_indirect_branch): Likewise.
4346 (ix86_output_indirect_jmp): Likewise.
4347 (ix86_code_end): Call output_indirect_thunk_function if needed.
4348 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4350 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4351 (ix86_attribute_table): Add indirect_branch.
4352 * config/i386/i386.h (machine_function): Add indirect_branch_type
4353 and has_local_indirect_jump.
4354 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4356 (tablejump): Likewise.
4357 (*indirect_jump): Use ix86_output_indirect_jmp.
4358 (*tablejump_1): Likewise.
4359 (simple_return_indirect_internal): Likewise.
4360 * config/i386/i386.opt (mindirect-branch=): New option.
4361 (indirect_branch): New.
4364 (thunk-inline): Likewise.
4365 (thunk-extern): Likewise.
4366 * doc/extend.texi: Document indirect_branch function attribute.
4367 * doc/invoke.texi: Document -mindirect-branch= option.
4369 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4372 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4374 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4376 * ipa-inline.c (want_inline_small_function_p): Return false if
4377 inlining has already failed with CIF_FINAL_ERROR.
4378 (update_caller_keys): Call want_inline_small_function_p before
4380 (update_callee_keys): Likewise.
4382 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4384 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4386 (rs6000_quadword_masked_address_p): Likewise.
4387 (quad_aligned_load_p): Likewise.
4388 (quad_aligned_store_p): Likewise.
4389 (const_load_sequence_p): Add comment to describe the outer-most loop.
4390 (mimic_memory_attributes_and_flags): New function.
4391 (rs6000_gen_stvx): Likewise.
4392 (replace_swapped_aligned_store): Likewise.
4393 (rs6000_gen_lvx): Likewise.
4394 (replace_swapped_aligned_load): Likewise.
4395 (replace_swapped_load_constant): Capitalize argument name in
4396 comment describing this function.
4397 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4398 and stores that access quad-word aligned addresses and replace
4399 with stvx or lvx instructions when appropriate.
4400 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4401 New function prototype.
4402 (rs6000_quadword_masked_address_p): Likewise.
4403 (rs6000_gen_lvx): Likewise.
4404 (rs6000_gen_stvx): Likewise.
4405 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4406 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4407 when memory address is aligned.
4408 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4409 this split to select lvx instruction when memory address is aligned.
4410 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4411 instruction when memory address is aligned.
4412 (*vsx_le_perm_load_v16qi): Likewise.
4413 (four unnamed splitters): Modify to select the stvx instruction
4414 when memory is aligned.
4416 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4418 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4419 which appears in the queue multiple times.
4421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4422 Alan Hayward <alan.hayward@arm.com>
4423 David Sherwood <david.sherwood@arm.com>
4425 * tree-vectorizer.h (vec_lower_bound): New structure.
4426 (_loop_vec_info): Add check_nonzero and lower_bounds.
4427 (LOOP_VINFO_CHECK_NONZERO): New macro.
4428 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4429 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4430 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4431 fields. Make seg_len the distance travelled, not including the
4433 (dr_direction_indicator): Declare.
4434 (dr_zero_step_indicator): Likewise.
4435 (dr_known_forward_stride_p): Likewise.
4436 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4438 (runtime_alias_check_p): Allow runtime alias checks with
4440 (operator ==): Compare access_size and align.
4441 (prune_runtime_alias_test_list): Rework for new distinction between
4442 the access_size and seg_len.
4443 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4445 (get_segment_min_max): New function.
4446 (create_intersect_range_checks): Use it.
4447 (dr_step_indicator): New function.
4448 (dr_direction_indicator): Likewise.
4449 (dr_zero_step_indicator): Likewise.
4450 (dr_known_forward_stride_p): Likewise.
4451 * tree-loop-distribution.c (data_ref_segment_size): Return
4452 DR_STEP * (niters - 1).
4453 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4455 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4456 (vect_preserves_scalar_order_p): New function, split out from...
4457 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4458 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4459 (vect_vfa_access_size): New function.
4460 (vect_vfa_align): Likewise.
4461 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4462 (dump_lower_bound): New function.
4463 (vect_check_lower_bound): Likewise.
4464 (vect_small_gap_p): Likewise.
4465 (vectorizable_with_step_bound_p): Likewise.
4466 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4467 depencies if the vectorization factor is 1. Convert the checks
4468 for nonzero steps into checks on the bounds of DR_STEP. Try using
4469 a bunds check for variable steps if the minimum required step is
4470 relatively small. Update calls to the dr_with_seg_len
4471 constructor and to vect_compile_time_alias.
4472 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4474 (vect_loop_versioning): Call it.
4475 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4477 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4479 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4480 Alan Hayward <alan.hayward@arm.com>
4481 David Sherwood <david.sherwood@arm.com>
4483 * doc/sourcebuild.texi (vect_scatter_store): Document.
4484 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4486 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4488 * genopinit.c (main): Add supports_vec_scatter_store and
4489 supports_vec_scatter_store_cached to target_optabs.
4490 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4491 IFN_MASK_SCATTER_STORE.
4492 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4494 * internal-fn.h (internal_store_fn_p): Declare.
4495 (internal_fn_stored_value_index): Likewise.
4496 * internal-fn.c (scatter_store_direct): New macro.
4497 (expand_scatter_store_optab_fn): New function.
4498 (direct_scatter_store_optab_supported_p): New macro.
4499 (internal_store_fn_p): New function.
4500 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4501 IFN_MASK_SCATTER_STORE.
4502 (internal_fn_mask_index): Likewise.
4503 (internal_fn_stored_value_index): New function.
4504 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4506 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4507 * optabs-query.c (supports_vec_scatter_store_p): New function.
4508 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4509 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4510 true for scatter stores.
4511 (vect_gather_scatter_fn_p): Handle scatter stores too.
4512 (vect_check_gather_scatter): Consider using scatter stores if
4513 supports_vec_scatter_store_p.
4514 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4516 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4517 internal_fn_stored_value_index.
4518 (check_load_store_masking): Handle scatter stores too.
4519 (vect_get_store_rhs): Make public.
4520 (vectorizable_call): Use internal_store_fn_p.
4521 (vectorizable_store): Handle scatter store internal functions.
4522 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4523 when deciding whether the end of the group has been reached.
4524 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4525 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4526 (mask_scatter_store<mode>): New insns.
4528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4529 Alan Hayward <alan.hayward@arm.com>
4530 David Sherwood <david.sherwood@arm.com>
4532 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4533 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4534 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4536 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4537 Use vect_truncate_gather_scatter_offset if we can't treat the
4538 operation as a normal gather load or scatter store.
4539 (get_group_load_store_type): Take the gather_scatter_info
4540 as argument. Try using a gather load or scatter store for
4541 single-element groups.
4542 (get_load_store_type): Update calls to get_group_load_store_type
4543 and vect_use_strided_gather_scatters_p.
4545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4546 Alan Hayward <alan.hayward@arm.com>
4547 David Sherwood <david.sherwood@arm.com>
4549 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4550 optional tree argument.
4551 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4553 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4554 but continue to use the current value as a fallback.
4555 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4556 to compare the updates.
4557 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4558 (get_load_store_type): Use it when handling a strided access.
4559 (vect_get_strided_load_store_ops): New function.
4560 (vect_get_data_ptr_increment): Likewise.
4561 (vectorizable_load): Handle strided gather loads. Always pass
4562 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4564 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4565 Alan Hayward <alan.hayward@arm.com>
4566 David Sherwood <david.sherwood@arm.com>
4568 * doc/md.texi (gather_load@var{m}): Document.
4569 (mask_gather_load@var{m}): Likewise.
4570 * genopinit.c (main): Add supports_vec_gather_load and
4571 supports_vec_gather_load_cached to target_optabs.
4572 * optabs-tree.c (init_tree_optimization_optabs): Use
4573 ggc_cleared_alloc to allocate target_optabs.
4574 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4575 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4577 * internal-fn.h (internal_load_fn_p): Declare.
4578 (internal_gather_scatter_fn_p): Likewise.
4579 (internal_fn_mask_index): Likewise.
4580 (internal_gather_scatter_fn_supported_p): Likewise.
4581 * internal-fn.c (gather_load_direct): New macro.
4582 (expand_gather_load_optab_fn): New function.
4583 (direct_gather_load_optab_supported_p): New macro.
4584 (direct_internal_fn_optab): New function.
4585 (internal_load_fn_p): Likewise.
4586 (internal_gather_scatter_fn_p): Likewise.
4587 (internal_fn_mask_index): Likewise.
4588 (internal_gather_scatter_fn_supported_p): Likewise.
4589 * optabs-query.c (supports_at_least_one_mode_p): New function.
4590 (supports_vec_gather_load_p): Likewise.
4591 * optabs-query.h (supports_vec_gather_load_p): Declare.
4592 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4593 and memory_type field.
4594 (NUM_PATTERNS): Bump to 15.
4595 * tree-vect-data-refs.c: Include internal-fn.h.
4596 (vect_gather_scatter_fn_p): New function.
4597 (vect_describe_gather_scatter_call): Likewise.
4598 (vect_check_gather_scatter): Try using internal functions for
4599 gather loads. Recognize existing calls to a gather load function.
4600 (vect_analyze_data_refs): Consider using gather loads if
4601 supports_vec_gather_load_p.
4602 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4603 (vect_get_gather_scatter_offset_type): Likewise.
4604 (vect_convert_mask_for_vectype): Likewise.
4605 (vect_add_conversion_to_patterm): Likewise.
4606 (vect_try_gather_scatter_pattern): Likewise.
4607 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4608 (vect_vect_recog_func_ptrs): Add it.
4609 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4610 internal_fn_mask_index and internal_gather_scatter_fn_p.
4611 (check_load_store_masking): Take the gather_scatter_info as an
4612 argument and handle gather loads.
4613 (vect_get_gather_scatter_ops): New function.
4614 (vectorizable_call): Check internal_load_fn_p.
4615 (vectorizable_load): Likewise. Handle gather load internal
4617 (vectorizable_store): Update call to check_load_store_masking.
4618 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4619 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4620 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4621 (aarch64_gather_scale_operand_d): New predicates.
4622 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4623 (mask_gather_load<mode>): New insns.
4625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4626 Alan Hayward <alan.hayward@arm.com>
4627 David Sherwood <david.sherwood@arm.com>
4629 * optabs.def (fold_left_plus_optab): New optab.
4630 * doc/md.texi (fold_left_plus_@var{m}): Document.
4631 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4632 * internal-fn.c (fold_left_direct): Define.
4633 (expand_fold_left_optab_fn): Likewise.
4634 (direct_fold_left_optab_supported_p): Likewise.
4635 * fold-const-call.c (fold_const_fold_left): New function.
4636 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4637 * tree-parloops.c (valid_reduction_p): New function.
4638 (gather_scalar_reductions): Use it.
4639 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4640 (vect_finish_replace_stmt): Declare.
4641 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4642 (needs_fold_left_reduction_p): New function, split out from...
4643 (vect_is_simple_reduction): ...here. Accept reductions that
4644 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4645 (vect_force_simple_reduction): Also store the reduction type in
4646 the assignment's STMT_VINFO_REDUC_TYPE.
4647 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4648 (merge_with_identity): New function.
4649 (vect_expand_fold_left): Likewise.
4650 (vectorize_fold_left_reduction): Likewise.
4651 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4652 scalar phi in place for it. Check for target support and reject
4653 cases that would reassociate the operation. Defer the transform
4654 phase to vectorize_fold_left_reduction.
4655 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4656 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4657 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4659 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4661 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4662 call to ifc_temp_var.
4664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4665 Alan Hayward <alan.hayward@arm.com>
4666 David Sherwood <david.sherwood@arm.com>
4668 * target.def (legitimize_address_displacement): Take the original
4669 offset as a poly_int.
4670 * targhooks.h (default_legitimize_address_displacement): Update
4672 * targhooks.c (default_legitimize_address_displacement): Likewise.
4673 * doc/tm.texi: Regenerate.
4674 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4675 as an argument, moving assert of ad->disp == ad->disp_term to...
4676 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4677 Try calling targetm.legitimize_address_displacement before expanding
4678 the address rather than afterwards, and adjust for the new interface.
4679 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4680 Match the new hook interface. Handle SVE addresses.
4681 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4686 * Makefile.in (OBJS): Add early-remat.o.
4687 * target.def (select_early_remat_modes): New hook.
4688 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4689 * doc/tm.texi: Regenerate.
4690 * targhooks.h (default_select_early_remat_modes): Declare.
4691 * targhooks.c (default_select_early_remat_modes): New function.
4692 * timevar.def (TV_EARLY_REMAT): New timevar.
4693 * passes.def (pass_early_remat): New pass.
4694 * tree-pass.h (make_pass_early_remat): Declare.
4695 * early-remat.c: New file.
4696 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4698 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4701 Alan Hayward <alan.hayward@arm.com>
4702 David Sherwood <david.sherwood@arm.com>
4704 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4705 vfm1 with a bound_epilog parameter.
4706 (vect_do_peeling): Update calls accordingly, and move the prologue
4707 call earlier in the function. Treat the base bound_epilog as 0 for
4708 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4709 this base when peeling for gaps.
4710 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4711 with fully-masked loops.
4712 (vect_estimate_min_profitable_iters): Handle the single peeled
4713 iteration in that case.
4715 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4716 Alan Hayward <alan.hayward@arm.com>
4717 David Sherwood <david.sherwood@arm.com>
4719 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4720 single-element interleaving even if the size is not a power of 2.
4721 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4722 accesses for single-element interleaving if the group size is
4725 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4726 Alan Hayward <alan.hayward@arm.com>
4727 David Sherwood <david.sherwood@arm.com>
4729 * doc/md.texi (fold_extract_last_@var{m}): Document.
4730 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4731 * optabs.def (fold_extract_last_optab): New optab.
4732 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4733 * internal-fn.c (fold_extract_direct): New macro.
4734 (expand_fold_extract_optab_fn): Likewise.
4735 (direct_fold_extract_optab_supported_p): Likewise.
4736 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4737 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4738 EXTRACT_LAST_REDUCTION.
4739 (get_initial_def_for_reduction): Do not create an initial vector
4740 for EXTRACT_LAST_REDUCTION reductions.
4741 (vectorizable_reduction): Leave the scalar phi in place for
4742 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4743 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4744 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4745 transform phase to vectorizable_condition.
4746 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4748 (vect_finish_stmt_generation): ...here.
4749 (vect_finish_replace_stmt): New function.
4750 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4751 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4753 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4756 Alan Hayward <alan.hayward@arm.com>
4757 David Sherwood <david.sherwood@arm.com>
4759 * doc/md.texi (extract_last_@var{m}): Document.
4760 * optabs.def (extract_last_optab): New optab.
4761 * internal-fn.def (EXTRACT_LAST): New internal function.
4762 * internal-fn.c (cond_unary_direct): New macro.
4763 (expand_cond_unary_optab_fn): Likewise.
4764 (direct_cond_unary_optab_supported_p): Likewise.
4765 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4766 loops using EXTRACT_LAST.
4767 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4768 (extract_last_<mode>): ...this optab.
4769 (vec_extract<mode><Vel>): Update accordingly.
4771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4772 Alan Hayward <alan.hayward@arm.com>
4773 David Sherwood <david.sherwood@arm.com>
4775 * target.def (empty_mask_is_expensive): New hook.
4776 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4777 * doc/tm.texi: Regenerate.
4778 * targhooks.h (default_empty_mask_is_expensive): Declare.
4779 * targhooks.c (default_empty_mask_is_expensive): New function.
4780 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4781 if the target says that empty masks are expensive.
4782 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4784 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4787 Alan Hayward <alan.hayward@arm.com>
4788 David Sherwood <david.sherwood@arm.com>
4790 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4791 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4792 (vect_use_loop_mask_for_alignment_p): New function.
4793 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4794 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4795 niters_skip argument. Make sure that the first niters_skip elements
4796 of the first iteration are inactive.
4797 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4798 Update call to vect_set_loop_masks_directly.
4799 (get_misalign_in_elems): New function, split out from...
4800 (vect_gen_prolog_loop_niters): ...here.
4801 (vect_update_init_of_dr): Take a code argument that specifies whether
4802 the adjustment should be added or subtracted.
4803 (vect_update_init_of_drs): Likewise.
4804 (vect_prepare_for_masked_peels): New function.
4805 (vect_do_peeling): Skip prologue peeling if we're using a mask
4806 instead. Update call to vect_update_inits_of_drs.
4807 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4809 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4810 alignment. Do not include the number of peeled iterations in
4811 the minimum threshold in that case.
4812 (vectorizable_induction): Adjust the start value down by
4813 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4814 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4815 Take the number of skipped iterations into account when calculating
4817 * tree-vect-stmts.c (vect_gen_while_not): New function.
4819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4820 Alan Hayward <alan.hayward@arm.com>
4821 David Sherwood <david.sherwood@arm.com>
4823 * doc/sourcebuild.texi (vect_fully_masked): Document.
4824 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4826 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4828 (vect_analyze_loop_2): ...here. Don't check the vectorization
4829 factor against the number of loop iterations if the loop is
4832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4833 Alan Hayward <alan.hayward@arm.com>
4834 David Sherwood <david.sherwood@arm.com>
4836 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4837 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4838 (dump_groups): Update accordingly.
4839 (iv_use::mem_type): New member variable.
4840 (address_p): New function.
4841 (record_use): Add a mem_type argument and initialize the new
4843 (record_group_use): Add a mem_type argument. Use address_p.
4844 Remove obsolete null checks of base_object. Update call to record_use.
4845 (find_interesting_uses_op): Update call to record_group_use.
4846 (find_interesting_uses_cond): Likewise.
4847 (find_interesting_uses_address): Likewise.
4848 (get_mem_type_for_internal_fn): New function.
4849 (find_address_like_use): Likewise.
4850 (find_interesting_uses_stmt): Try find_address_like_use before
4851 calling find_interesting_uses_op.
4852 (addr_offset_valid_p): Use the iv mem_type field as the type
4853 of the addressed memory.
4854 (add_autoinc_candidates): Likewise.
4855 (get_address_cost): Likewise.
4856 (split_small_address_groups_p): Use address_p.
4857 (split_address_groups): Likewise.
4858 (add_iv_candidate_for_use): Likewise.
4859 (autoinc_possible_for_pair): Likewise.
4860 (rewrite_groups): Likewise.
4861 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4862 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4863 (get_alias_ptr_type_for_ptr_address): New function.
4864 (rewrite_use_address): Rewrite address uses in calls that were
4865 identified by find_address_like_use.
4867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4868 Alan Hayward <alan.hayward@arm.com>
4869 David Sherwood <david.sherwood@arm.com>
4871 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4873 * gimple-expr.h (is_gimple_addressable: Likewise.
4874 * gimple-expr.c (is_gimple_address): Likewise.
4875 * internal-fn.c (expand_call_mem_ref): New function.
4876 (expand_mask_load_optab_fn): Use it.
4877 (expand_mask_store_optab_fn): Likewise.
4879 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4880 Alan Hayward <alan.hayward@arm.com>
4881 David Sherwood <david.sherwood@arm.com>
4883 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4884 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4885 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4886 (cond_umax@var{mode}): Document.
4887 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4888 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4889 (cond_umin_optab, cond_umax_optab): New optabs.
4890 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4891 (COND_IOR, COND_XOR): New internal functions.
4892 * internal-fn.h (get_conditional_internal_fn): Declare.
4893 * internal-fn.c (cond_binary_direct): New macro.
4894 (expand_cond_binary_optab_fn): Likewise.
4895 (direct_cond_binary_optab_supported_p): Likewise.
4896 (get_conditional_internal_fn): New function.
4897 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4898 Cope with reduction statements that are vectorized as calls rather
4900 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4901 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4902 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4903 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4904 (UNSPEC_COND_EOR): New unspecs.
4905 (optab): Add mappings for them.
4906 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4907 (sve_int_op, sve_fp_op): New int attributes.
4909 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4910 Alan Hayward <alan.hayward@arm.com>
4911 David Sherwood <david.sherwood@arm.com>
4913 * optabs.def (while_ult_optab): New optab.
4914 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4915 * internal-fn.def (WHILE_ULT): New internal function.
4916 * internal-fn.h (direct_internal_fn_supported_p): New override
4917 that takes two types as argument.
4918 * internal-fn.c (while_direct): New macro.
4919 (expand_while_optab_fn): New function.
4920 (convert_optab_supported_p): Likewise.
4921 (direct_while_optab_supported_p): New macro.
4922 * wide-int.h (wi::udiv_ceil): New function.
4923 * tree-vectorizer.h (rgroup_masks): New structure.
4924 (vec_loop_masks): New typedef.
4925 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4927 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4928 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4929 (vect_max_vf): New function.
4930 (slpeel_make_loop_iterate_ntimes): Delete.
4931 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4932 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4933 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4934 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4935 internal-fn.h, stor-layout.h and optabs-query.h.
4936 (vect_set_loop_mask): New function.
4937 (add_preheader_seq): Likewise.
4938 (add_header_seq): Likewise.
4939 (interleave_supported_p): Likewise.
4940 (vect_maybe_permute_loop_masks): Likewise.
4941 (vect_set_loop_masks_directly): Likewise.
4942 (vect_set_loop_condition_masked): Likewise.
4943 (vect_set_loop_condition_unmasked): New function, split out from
4944 slpeel_make_loop_iterate_ntimes.
4945 (slpeel_make_loop_iterate_ntimes): Rename to..
4946 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4947 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4948 (vect_do_peeling): Update call accordingly.
4949 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4951 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4952 mask_compare_type, can_fully_mask_p and fully_masked_p.
4953 (release_vec_loop_masks): New function.
4954 (_loop_vec_info): Use it to free the loop masks.
4955 (can_produce_all_loop_masks_p): New function.
4956 (vect_get_max_nscalars_per_iter): Likewise.
4957 (vect_verify_full_masking): Likewise.
4958 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4959 retries, and free the mask rgroups before retrying. Check loop-wide
4960 reasons for disallowing fully-masked loops. Make the final decision
4961 about whether use a fully-masked loop or not.
4962 (vect_estimate_min_profitable_iters): Do not assume that peeling
4963 for the number of iterations will be needed for fully-masked loops.
4964 (vectorizable_reduction): Disable fully-masked loops.
4965 (vectorizable_live_operation): Likewise.
4966 (vect_halve_mask_nunits): New function.
4967 (vect_double_mask_nunits): Likewise.
4968 (vect_record_loop_mask): Likewise.
4969 (vect_get_loop_mask): Likewise.
4970 (vect_transform_loop): Handle the case in which the final loop
4971 iteration might handle a partial vector. Call vect_set_loop_condition
4972 instead of slpeel_make_loop_iterate_ntimes.
4973 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4974 (check_load_store_masking): New function.
4975 (prepare_load_store_mask): Likewise.
4976 (vectorizable_store): Handle fully-masked loops.
4977 (vectorizable_load): Likewise.
4978 (supportable_widening_operation): Use vect_halve_mask_nunits for
4980 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4981 (vect_gen_while): New function.
4982 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4983 (aarch64_uqdec<mode>): New insn.
4985 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4986 Alan Hayward <alan.hayward@arm.com>
4987 David Sherwood <david.sherwood@arm.com>
4989 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4990 (reduc_xor_scal_optab): New optabs.
4991 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4992 (reduc_xor_scal_@var{m}): Document.
4993 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4994 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4996 * fold-const-call.c (fold_const_call): Handle them.
4997 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4998 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4999 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5000 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5001 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5002 (UNSPEC_XORV): New unspecs.
5003 (optab): Add entries for them.
5004 (BITWISEV): New int iterator.
5005 (bit_reduc_op): New int attributes.
5007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5008 Alan Hayward <alan.hayward@arm.com>
5009 David Sherwood <david.sherwood@arm.com>
5011 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5012 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5013 * optabs.def (vec_shl_insert_optab): New optab.
5014 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5015 (duplicate_and_interleave): Likewise.
5016 * tree-vect-loop.c: Include internal-fn.h.
5017 (neutral_op_for_slp_reduction): New function, split out from
5018 get_initial_defs_for_reduction.
5019 (get_initial_def_for_reduction): Handle option 2 for variable-length
5020 vectors by loading the neutral value into a vector and then shifting
5021 the initial value into element 0.
5022 (get_initial_defs_for_reduction): Replace the code argument with
5023 the neutral value calculated by neutral_op_for_slp_reduction.
5024 Use gimple_build_vector for constant-length vectors.
5025 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5026 but the first group_size elements have a neutral value.
5027 Use duplicate_and_interleave otherwise.
5028 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5029 Update call to get_initial_defs_for_reduction. Handle SLP
5030 reductions for variable-length vectors by creating one vector
5031 result for each scalar result, with the elements associated
5032 with other scalar results stubbed out with the neutral value.
5033 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5034 Require IFN_VEC_SHL_INSERT for double reductions on
5035 variable-length vectors, or SLP reductions that have
5036 a neutral value. Require can_duplicate_and_interleave_p
5037 support for variable-length unchained SLP reductions if there
5038 is no neutral value, such as for MIN/MAX reductions. Also require
5039 the number of vector elements to be a multiple of the number of
5040 SLP statements when doing variable-length unchained SLP reductions.
5041 Update call to vect_create_epilog_for_reduction.
5042 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5043 and remove initial values.
5044 (duplicate_and_interleave): Make public.
5045 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5046 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5048 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5049 Alan Hayward <alan.hayward@arm.com>
5050 David Sherwood <david.sherwood@arm.com>
5052 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5053 (can_duplicate_and_interleave_p): New function.
5054 (vect_get_and_check_slp_defs): Take the vector of statements
5055 rather than just the current one. Remove excess parentheses.
5056 Restriction rejectinon of vect_constant_def and vect_external_def
5057 for variable-length vectors to boolean types, or types for which
5058 can_duplicate_and_interleave_p is false.
5059 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5060 (duplicate_and_interleave): New function.
5061 (vect_get_constant_vectors): Use gimple_build_vector for
5062 constant-length vectors and suitable variable-length constant
5063 vectors. Use duplicate_and_interleave for other variable-length
5064 vectors. Don't defer the update when inserting new statements.
5066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5067 Alan Hayward <alan.hayward@arm.com>
5068 David Sherwood <david.sherwood@arm.com>
5070 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5071 min_profitable_iters doesn't go negative.
5073 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5074 Alan Hayward <alan.hayward@arm.com>
5075 David Sherwood <david.sherwood@arm.com>
5077 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5078 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5079 * optabs.def (vec_mask_load_lanes_optab): New optab.
5080 (vec_mask_store_lanes_optab): Likewise.
5081 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5082 (MASK_STORE_LANES): Likewise.
5083 * internal-fn.c (mask_load_lanes_direct): New macro.
5084 (mask_store_lanes_direct): Likewise.
5085 (expand_mask_load_optab_fn): Handle masked operations.
5086 (expand_mask_load_lanes_optab_fn): New macro.
5087 (expand_mask_store_optab_fn): Handle masked operations.
5088 (expand_mask_store_lanes_optab_fn): New macro.
5089 (direct_mask_load_lanes_optab_supported_p): Likewise.
5090 (direct_mask_store_lanes_optab_supported_p): Likewise.
5091 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5093 (vect_load_lanes_supported): Likewise.
5094 * tree-vect-data-refs.c (strip_conversion): New function.
5095 (can_group_stmts_p): Likewise.
5096 (vect_analyze_data_ref_accesses): Use it instead of checking
5097 for a pair of assignments.
5098 (vect_store_lanes_supported): Take a masked_p parameter.
5099 (vect_load_lanes_supported): Likewise.
5100 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5101 vect_store_lanes_supported and vect_load_lanes_supported.
5102 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5103 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5104 parameter. Don't allow gaps for masked accesses.
5105 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5106 and vect_load_lanes_supported.
5107 (get_load_store_type): Take a masked_p parameter and update
5108 call to get_group_load_store_type.
5109 (vectorizable_store): Update call to get_load_store_type.
5110 Handle IFN_MASK_STORE_LANES.
5111 (vectorizable_load): Update call to get_load_store_type.
5112 Handle IFN_MASK_LOAD_LANES.
5114 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5115 Alan Hayward <alan.hayward@arm.com>
5116 David Sherwood <david.sherwood@arm.com>
5118 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5120 * config/aarch64/aarch64-protos.h
5121 (aarch64_sve_struct_memory_operand_p): Declare.
5122 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5123 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5124 (VPRED, vpred): Handle SVE structure modes.
5125 * config/aarch64/constraints.md (Utx): New constraint.
5126 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5127 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5128 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5129 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5130 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5131 structure modes. Split into pieces after RA.
5132 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5133 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5135 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5136 SVE structure modes.
5137 (aarch64_classify_address): Likewise.
5138 (sizetochar): Move earlier in file.
5139 (aarch64_print_operand): Handle SVE register lists.
5140 (aarch64_array_mode): New function.
5141 (aarch64_sve_struct_memory_operand_p): Likewise.
5142 (TARGET_ARRAY_MODE): Redefine.
5144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5145 Alan Hayward <alan.hayward@arm.com>
5146 David Sherwood <david.sherwood@arm.com>
5148 * target.def (array_mode): New target hook.
5149 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5150 * doc/tm.texi: Regenerate.
5151 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5152 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5153 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5155 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5159 Alan Hayward <alan.hayward@arm.com>
5160 David Sherwood <david.sherwood@arm.com>
5162 * fold-const.c (fold_binary_loc): Check the argument types
5163 rather than the result type when testing for a vector operation.
5165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5167 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5168 * doc/tm.texi: Regenerate.
5170 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5171 Alan Hayward <alan.hayward@arm.com>
5172 David Sherwood <david.sherwood@arm.com>
5174 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5175 (sve): Document new AArch64 extension.
5176 * doc/md.texi (w): Extend the description of the AArch64
5177 constraint to include SVE vectors.
5178 (Upl, Upa): Document new AArch64 predicate constraints.
5179 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5181 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5182 (msve-vector-bits=): New option.
5183 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5184 SVE when these are disabled.
5185 (sve): New extension.
5186 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5187 modes. Adjust their number of units based on aarch64_sve_vg.
5188 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5189 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5190 aarch64_addr_query_type.
5191 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5192 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5193 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5194 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5195 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5196 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5197 (aarch64_simd_imm_zero_p): Delete.
5198 (aarch64_check_zero_based_sve_index_immediate): Declare.
5199 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5200 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5201 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5202 (aarch64_sve_float_mul_immediate_p): Likewise.
5203 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5205 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5206 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5207 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5208 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5209 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5210 (aarch64_regmode_natural_size): Likewise.
5211 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5212 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5214 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5215 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5216 for VG and the SVE predicate registers.
5217 (V_ALIASES): Add a "z"-prefixed alias.
5218 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5219 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5220 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5221 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5222 (REG_CLASS_NAMES): Add entries for them.
5223 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5224 and the predicate registers.
5225 (aarch64_sve_vg): Declare.
5226 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5227 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5228 (REGMODE_NATURAL_SIZE): Define.
5229 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5231 * config/aarch64/aarch64.c: Include cfgrtl.h.
5232 (simd_immediate_info): Add a constructor for series vectors,
5233 and an associated step field.
5234 (aarch64_sve_vg): New variable.
5235 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5236 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5237 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5238 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5239 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5240 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5241 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5242 (aarch64_get_mask_mode): New functions.
5243 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5244 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5245 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5246 predicate modes and predicate registers. Explicitly restrict
5247 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5248 to store a vector mode if it is recognized by
5249 aarch64_classify_vector_mode.
5250 (aarch64_regmode_natural_size): New function.
5251 (aarch64_hard_regno_caller_save_mode): Return the original mode
5253 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5254 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5255 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5256 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5258 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5259 does not overlap dest if the function is frame-related. Handle
5261 (aarch64_split_add_offset): New function.
5262 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5263 them aarch64_add_offset.
5264 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5265 and update call to aarch64_sub_sp.
5266 (aarch64_add_cfa_expression): New function.
5267 (aarch64_expand_prologue): Pass extra temporary registers to the
5268 functions above. Handle the case in which we need to emit new
5269 DW_CFA_expressions for registers that were originally saved
5270 relative to the stack pointer, but now have to be expressed
5271 relative to the frame pointer.
5272 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5274 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5275 IP0 and IP1 values for SVE frames.
5276 (aarch64_expand_vec_series): New function.
5277 (aarch64_expand_sve_widened_duplicate): Likewise.
5278 (aarch64_expand_sve_const_vector): Likewise.
5279 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5280 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5281 into the register, rather than emitting a SET directly.
5282 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5283 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5284 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5285 (offset_9bit_signed_scaled_p): New functions.
5286 (aarch64_replicate_bitmask_imm): New function.
5287 (aarch64_bitmask_imm): Use it.
5288 (aarch64_cannot_force_const_mem): Reject expressions involving
5289 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5290 (aarch64_classify_index): Handle SVE indices, by requiring
5291 a plain register index with a scale that matches the element size.
5292 (aarch64_classify_address): Handle SVE addresses. Assert that
5293 the mode of the address is VOIDmode or an integer mode.
5294 Update call to aarch64_classify_symbol.
5295 (aarch64_classify_symbolic_expression): Update call to
5296 aarch64_classify_symbol.
5297 (aarch64_const_vec_all_in_range_p): New function.
5298 (aarch64_print_vector_float_operand): Likewise.
5299 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5300 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5301 and the FP immediates 1.0 and 0.5.
5302 (aarch64_print_address_internal): Handle SVE addresses.
5303 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5304 (aarch64_regno_regclass): Handle predicate registers.
5305 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5307 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5308 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5309 (aarch64_convert_sve_vector_bits): New function.
5310 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5311 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5313 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5314 Handle SVE vector and predicate modes. Accept VL-based constants
5315 that need only one temporary register, and VL offsets that require
5316 no temporary registers.
5317 (aarch64_conditional_register_usage): Mark the predicate registers
5318 as fixed if SVE isn't available.
5319 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5320 Return true for SVE vector and predicate modes.
5321 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5322 rather than an unsigned int. Handle SVE modes.
5323 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5325 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5327 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5328 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5329 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5330 (aarch64_sve_float_mul_immediate_p): New functions.
5331 (aarch64_sve_valid_immediate): New function.
5332 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5333 Explicitly reject structure modes. Check for INDEX constants.
5334 Handle PTRUE and PFALSE constants.
5335 (aarch64_check_zero_based_sve_index_immediate): New function.
5336 (aarch64_simd_imm_zero_p): Delete.
5337 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5338 vector modes. Accept constants in the range of CNT[BHWD].
5339 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5340 ask for an Advanced SIMD mode.
5341 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5342 (aarch64_simd_vector_alignment): Handle SVE predicates.
5343 (aarch64_vectorize_preferred_vector_alignment): New function.
5344 (aarch64_simd_vector_alignment_reachable): Use it instead of
5346 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5347 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5349 (MAX_VECT_LEN): Delete.
5350 (expand_vec_perm_d): Add a vec_flags field.
5351 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5352 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5353 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5355 (aarch64_evpc_rev): Rename to...
5356 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5357 (aarch64_evpc_rev_global): New function.
5358 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5359 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5361 (aarch64_evpc_sve_tbl): New function.
5362 (aarch64_expand_vec_perm_const_1): Update after rename of
5363 aarch64_evpc_rev. Handle SVE permutes too, trying
5364 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5365 than aarch64_evpc_tbl.
5366 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5367 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5368 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5369 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5370 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5371 (aarch64_expand_sve_vcond): New functions.
5372 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5373 of aarch64_vector_mode_p.
5374 (aarch64_dwarf_poly_indeterminate_value): New function.
5375 (aarch64_compute_pressure_classes): Likewise.
5376 (aarch64_can_change_mode_class): Likewise.
5377 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5378 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5379 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5380 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5381 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5382 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5383 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5384 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5386 (Dn, Dl, Dr): Accept const as well as const_vector.
5387 (Dz): Likewise. Compare against CONST0_RTX.
5388 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5389 of "vector" where appropriate.
5390 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5391 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5392 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5393 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5394 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5395 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5396 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5397 (v_int_equiv): Extend to SVE modes.
5398 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5400 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5401 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5402 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5403 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5404 (SVE_COND_FP_CMP): New int iterators.
5405 (perm_hilo): Handle the new unpack unspecs.
5406 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5408 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5409 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5410 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5411 (aarch64_equality_operator, aarch64_constant_vector_operand)
5412 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5413 (aarch64_sve_nonimmediate_operand): Likewise.
5414 (aarch64_sve_general_operand): Likewise.
5415 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5416 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5417 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5418 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5419 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5420 (aarch64_sve_float_arith_immediate): Likewise.
5421 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5422 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5423 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5424 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5425 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5426 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5427 (aarch64_sve_float_arith_operand): Likewise.
5428 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5429 (aarch64_sve_float_mul_operand): Likewise.
5430 (aarch64_sve_vec_perm_operand): Likewise.
5431 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5432 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5433 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5434 as well as const_vector.
5435 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5436 in file. Use CONST0_RTX and CONSTM1_RTX.
5437 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5438 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5439 Use aarch64_simd_imm_zero.
5440 * config/aarch64/aarch64-sve.md: New file.
5441 * config/aarch64/aarch64.md: Include it.
5442 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5443 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5444 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5445 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5446 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5447 (sve): New attribute.
5448 (enabled): Disable instructions with the sve attribute unless
5450 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5451 aarch64_expand_mov_immediate.
5452 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5453 CNT[BHSD] immediates.
5454 (movti): Split CONST_POLY_INT moves into two halves.
5455 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5456 Split additions that need a temporary here if the destination
5457 is the stack pointer.
5458 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5459 (*add<mode>3_poly_1): New instruction.
5460 (set_clobber_cc): New expander.
5462 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5464 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5465 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5466 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5467 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5468 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5469 Change innermode from fixed_mode_size to machine_mode.
5470 (simplify_subreg): Update call accordingly. Handle a constant-sized
5471 subreg of a variable-length CONST_VECTOR.
5473 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5474 Alan Hayward <alan.hayward@arm.com>
5475 David Sherwood <david.sherwood@arm.com>
5477 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5478 (add_offset_to_base): New function, split out from...
5479 (create_mem_ref): ...here. When handling a scale other than 1,
5480 check first whether the address is valid without the offset.
5481 Add it into the base if so, leaving the index and scale as-is.
5483 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5486 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5487 fold_for_warn before checking if arg2 is INTEGER_CST.
5489 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5491 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5492 (store_multiple_operation): Delete.
5493 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5494 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5495 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5496 guarded by TARGET_STRING.
5497 (rs6000_output_load_multiple): Delete.
5498 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5499 OPTION_MASK_STRING / TARGET_STRING handling.
5500 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5501 (const rs6000_opt_masks) <"string">: Change mask to 0.
5502 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5503 (MASK_STRING): Delete.
5504 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5506 (load_multiple): Delete.
5513 (store_multiple): Delete.
5520 (movmemsi_8reg): Delete.
5521 (corresponding unnamed define_insn): Delete.
5522 (movmemsi_6reg): Delete.
5523 (corresponding unnamed define_insn): Delete.
5524 (movmemsi_4reg): Delete.
5525 (corresponding unnamed define_insn): Delete.
5526 (movmemsi_2reg): Delete.
5527 (corresponding unnamed define_insn): Delete.
5528 (movmemsi_1reg): Delete.
5529 (corresponding unnamed define_insn): Delete.
5530 * config/rs6000/rs6000.opt (mno-string): New.
5531 (mstring): Replace by deprecation warning stub.
5532 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5534 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5536 * regrename.c (regrename_do_replace): If replacing the same
5537 reg multiple times, try to reuse last created gen_raw_REG.
5540 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5541 main to workaround a bug in GDB.
5543 2018-01-12 Tom de Vries <tom@codesourcery.com>
5546 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5548 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5550 PR rtl-optimization/80481
5551 * ira-color.c (get_cap_member): New function.
5552 (allocnos_conflict_by_live_ranges_p): Use it.
5553 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5554 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5556 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5559 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5560 (*saddl_se_1): Ditto.
5562 (*ssubl_se_1): Ditto.
5564 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5566 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5567 rather than wi::to_widest for DR_INITs.
5568 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5569 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5570 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5572 (vect_analyze_group_access_1): Note that here.
5574 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5576 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5577 polynomial type sizes.
5579 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5581 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5582 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5583 (gimple_add_tmp_var): Likewise.
5585 2018-01-12 Martin Liska <mliska@suse.cz>
5587 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5588 (gimple_alloc_sizes): Likewise.
5589 (dump_gimple_statistics): Use PRIu64 in printf format.
5590 * gimple.h: Change uint64_t to int.
5592 2018-01-12 Martin Liska <mliska@suse.cz>
5594 * tree-core.h: Use uint64_t instead of int.
5595 * tree.c (tree_node_counts): Likewise.
5596 (tree_node_sizes): Likewise.
5597 (dump_tree_statistics): Use PRIu64 in printf format.
5599 2018-01-12 Martin Liska <mliska@suse.cz>
5601 * Makefile.in: As qsort_chk is implemented in vec.c, add
5602 vec.o to linkage of gencfn-macros.
5603 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5604 passing the info to record_node_allocation_statistics.
5605 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5607 * ggc-common.c (struct ggc_usage): Add operator== and use
5608 it in operator< and compare function.
5609 * mem-stats.h (struct mem_usage): Likewise.
5610 * vec.c (struct vec_usage): Remove operator< and compare
5611 function. Can be simply inherited.
5613 2018-01-12 Martin Jambor <mjambor@suse.cz>
5616 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5617 * tree-ssa-math-opts.c: Include domwalk.h.
5618 (convert_mult_to_fma_1): New function.
5619 (fma_transformation_info): New type.
5620 (fma_deferring_state): Likewise.
5621 (cancel_fma_deferring): New function.
5622 (result_of_phi): Likewise.
5623 (last_fma_candidate_feeds_initial_phi): Likewise.
5624 (convert_mult_to_fma): Added deferring logic, split actual
5625 transformation to convert_mult_to_fma_1.
5626 (math_opts_dom_walker): New type.
5627 (math_opts_dom_walker::after_dom_children): New method, body moved
5628 here from pass_optimize_widening_mul::execute, added deferring logic
5630 (pass_optimize_widening_mul::execute): Moved most of code to
5631 math_opts_dom_walker::after_dom_children.
5632 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5633 * config/i386/i386.c (ix86_option_override_internal): Added
5634 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5636 2018-01-12 Richard Biener <rguenther@suse.de>
5639 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5640 inline instance vars.
5642 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5645 * config/rx/rx.c (rx_is_restricted_memory_address):
5648 2018-01-12 Richard Biener <rguenther@suse.de>
5650 PR tree-optimization/80846
5651 * target.def (split_reduction): New target hook.
5652 * targhooks.c (default_split_reduction): New function.
5653 * targhooks.h (default_split_reduction): Declare.
5654 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5655 target requests first reduce vectors by combining low and high
5657 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5658 (get_vectype_for_scalar_type_and_size): Export.
5659 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5660 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5661 * doc/tm.texi: Regenerate.
5662 * config/i386/i386.c (ix86_split_reduction): Implement
5663 TARGET_VECTORIZE_SPLIT_REDUCTION.
5665 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5668 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5669 in PIC mode except for TARGET_VXWORKS_RTP.
5670 * config/sparc/sparc.c: Include cfgrtl.h.
5671 (TARGET_INIT_PIC_REG): Define.
5672 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5673 (sparc_pic_register_p): New predicate.
5674 (sparc_legitimate_address_p): Use it.
5675 (sparc_legitimize_pic_address): Likewise.
5676 (sparc_delegitimize_address): Likewise.
5677 (sparc_mode_dependent_address_p): Likewise.
5678 (gen_load_pcrel_sym): Remove 4th parameter.
5679 (load_got_register): Adjust call to above. Remove obsolete stuff.
5680 (sparc_expand_prologue): Do not call load_got_register here.
5681 (sparc_flat_expand_prologue): Likewise.
5682 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5683 (sparc_use_pseudo_pic_reg): New function.
5684 (sparc_init_pic_reg): Likewise.
5685 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5686 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5688 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5690 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5691 Add item for branch_cost.
5693 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5695 PR rtl-optimization/83565
5696 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5697 not extend the result to a larger mode for rotate operations.
5698 (num_sign_bit_copies1): Likewise.
5700 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5703 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5705 Use values-Xc.o for -pedantic.
5706 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5708 2018-01-12 Martin Liska <mliska@suse.cz>
5711 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5713 (possible_polymorphic_call_targets): Use it.
5714 (ipa_devirt): Likewise.
5716 2018-01-12 Martin Liska <mliska@suse.cz>
5718 * profile-count.h (enum profile_quality): Use 0 as invalid
5719 enum value of profile_quality.
5721 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5723 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5724 -mext-string options.
5726 2018-01-12 Richard Biener <rguenther@suse.de>
5728 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5729 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5730 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5732 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5734 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5736 * configure.ac (--with-long-double-format): Add support for the
5737 configuration option to change the default long double format on
5739 * config.gcc (powerpc*-linux*-*): Likewise.
5740 * configure: Regenerate.
5741 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5742 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5743 used without modification.
5745 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5747 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5748 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5749 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5750 MISC_BUILTIN_SPEC_BARRIER.
5751 (rs6000_init_builtins): Likewise.
5752 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5754 (speculation_barrier): New define_insn.
5755 * doc/extend.texi: Document __builtin_speculation_barrier.
5757 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5760 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5761 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5762 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5764 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5765 integral modes instead of "ss" and "sd".
5766 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5767 vectors with 32-bit and 64-bit elements.
5768 (vecdupssescalarmodesuffix): New mode attribute.
5769 (vec_dup<mode>): Use it.
5771 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5774 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5775 frame if argument is passed on stack.
5777 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5780 * ree.c (combine_reaching_defs): Optimize also
5781 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5782 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5784 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5787 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5789 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5792 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5793 after they are computed.
5795 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5797 PR tree-optimization/83695
5798 * gimple-loop-linterchange.cc
5799 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5800 reset cached scev information after interchange.
5801 (pass_linterchange::execute): Remove call to scev_reset_htab.
5803 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5805 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5806 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5807 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5808 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5809 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5810 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5811 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5812 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5813 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5814 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5815 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5816 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5817 (V_lane_reg): Likewise.
5818 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5820 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5821 (vfmal_lane_low<mode>_intrinsic,
5822 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5823 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5824 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5825 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5826 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5827 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5829 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5831 * config/arm/arm-cpus.in (fp16fml): New feature.
5832 (ALL_SIMD): Add fp16fml.
5833 (armv8.2-a): Add fp16fml as an option.
5834 (armv8.3-a): Likewise.
5835 (armv8.4-a): Add fp16fml as part of fp16.
5836 * config/arm/arm.h (TARGET_FP16FML): Define.
5837 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5839 * config/arm/arm-modes.def (V2HF): Define.
5840 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5841 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5842 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5843 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5844 vfmsl_low, vfmsl_high): New set of builtins.
5845 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5846 (vfml_op): New code attribute.
5847 (VFMLHALVES): New int iterator.
5848 (VFML, VFMLSEL): New mode attributes.
5849 (V_reg): Define mapping for V2HF.
5850 (V_hi, V_lo): New mode attributes.
5851 (VF_constraint): Likewise.
5852 (vfml_half, vfml_half_selector): New int attributes.
5853 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5855 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5856 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5858 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5859 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5860 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5861 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5863 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5864 Document new effective target and option set.
5866 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5868 * config/arm/arm-cpus.in (armv8_4): New feature.
5869 (ARMv8_4a): New fgroup.
5870 (armv8.4-a): New arch.
5871 * config/arm/arm-tables.opt: Regenerate.
5872 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5873 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5874 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5875 Add matching rules for -march=armv8.4-a and extensions.
5876 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5878 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5881 * config/rx/rx.md (BW): New mode attribute.
5882 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5884 2018-01-11 Richard Biener <rguenther@suse.de>
5886 PR tree-optimization/83435
5887 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5888 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5889 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5891 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5892 Alan Hayward <alan.hayward@arm.com>
5893 David Sherwood <david.sherwood@arm.com>
5895 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5897 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5898 (aarch64_print_address_internal): Use it to check for a zero offset.
5900 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5901 Alan Hayward <alan.hayward@arm.com>
5902 David Sherwood <david.sherwood@arm.com>
5904 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5905 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5906 Return a poly_int64 rather than a HOST_WIDE_INT.
5907 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5908 rather than a HOST_WIDE_INT.
5909 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5910 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5911 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5912 final_offset from HOST_WIDE_INT to poly_int64.
5913 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5914 to_constant when getting the number of units in an Advanced SIMD
5916 (aarch64_builtin_vectorized_function): Check for a constant number
5918 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5920 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5921 attribute instead of GET_MODE_NUNITS.
5922 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5923 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5924 GET_MODE_SIZE for fixed-size registers.
5925 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5926 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5927 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5928 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5929 (aarch64_print_operand, aarch64_print_address_internal)
5930 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5931 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5932 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5933 Handle polynomial GET_MODE_SIZE.
5934 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5935 wider than SImode without modification.
5936 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5937 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5938 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5939 passing and returning SVE modes.
5940 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5941 rather than GEN_INT.
5942 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5943 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5944 (aarch64_allocate_and_probe_stack_space): Likewise.
5945 (aarch64_layout_frame): Cope with polynomial offsets.
5946 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5947 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5949 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5950 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5951 poly_int64 rather than a HOST_WIDE_INT.
5952 (aarch64_get_separate_components, aarch64_process_components)
5953 (aarch64_expand_prologue, aarch64_expand_epilogue)
5954 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5955 (aarch64_anchor_offset): New function, split out from...
5956 (aarch64_legitimize_address): ...here.
5957 (aarch64_builtin_vectorization_cost): Handle polynomial
5958 TYPE_VECTOR_SUBPARTS.
5959 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5961 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5962 number of elements from the PARALLEL rather than the mode.
5963 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5964 rather than GET_MODE_BITSIZE.
5965 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5966 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5967 (aarch64_expand_vec_perm_const_1): Handle polynomial
5968 d->perm.length () and d->perm elements.
5969 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5970 Apply to_constant to d->perm elements.
5971 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5972 polynomial CONST_VECTOR_NUNITS.
5973 (aarch64_move_pointer): Take amount as a poly_int64 rather
5975 (aarch64_progress_pointer): Avoid temporary variable.
5976 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5977 the mode attribute instead of GET_MODE.
5979 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5980 Alan Hayward <alan.hayward@arm.com>
5981 David Sherwood <david.sherwood@arm.com>
5983 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5984 x exists before using it.
5985 (aarch64_add_constant_internal): Rename to...
5986 (aarch64_add_offset_1): ...this. Replace regnum with separate
5987 src and dest rtxes. Handle the case in which they're different,
5988 including when the offset is zero. Replace scratchreg with an rtx.
5989 Use 2 additions if there is no spare register into which we can
5990 move a 16-bit constant.
5991 (aarch64_add_constant): Delete.
5992 (aarch64_add_offset): Replace reg with separate src and dest
5993 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5994 Use aarch64_add_offset_1.
5995 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5996 an rtx rather than an int. Take the delta as a poly_int64
5997 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5998 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5999 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6000 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6001 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6003 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6004 aarch64_add_constant.
6006 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6008 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6009 Use scalar_float_mode.
6011 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6013 * config/aarch64/aarch64-simd.md
6014 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6015 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6016 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6017 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6018 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6019 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6020 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6021 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6022 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6023 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6025 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6028 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6029 targ_options->x_arm_arch_string is non NULL.
6031 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6033 * config/aarch64/aarch64.h
6034 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6036 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6039 * expmed.c (emit_store_flag_force): Swap if const op0
6040 and change VOIDmode to mode of op0.
6042 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6044 PR rtl-optimization/83761
6045 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6046 than bytes to mode_for_size.
6048 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6051 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6052 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6055 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6058 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6059 when in layout mode.
6060 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6061 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6064 2018-01-10 Michael Collison <michael.collison@arm.com>
6066 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6067 * config/aarch64/aarch64-option-extension.def: Add
6068 AARCH64_OPT_EXTENSION of 'fp16fml'.
6069 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6070 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6071 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6072 * config/aarch64/constraints.md (Ui7): New constraint.
6073 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6074 (VFMLA_SEL_W): Ditto.
6077 (VFMLA16_LOW): New int iterator.
6078 (VFMLA16_HIGH): Ditto.
6079 (UNSPEC_FMLAL): New unspec.
6080 (UNSPEC_FMLSL): Ditto.
6081 (UNSPEC_FMLAL2): Ditto.
6082 (UNSPEC_FMLSL2): Ditto.
6083 (f16mac): New code attribute.
6084 * config/aarch64/aarch64-simd-builtins.def
6085 (aarch64_fmlal_lowv2sf): Ditto.
6086 (aarch64_fmlsl_lowv2sf): Ditto.
6087 (aarch64_fmlalq_lowv4sf): Ditto.
6088 (aarch64_fmlslq_lowv4sf): Ditto.
6089 (aarch64_fmlal_highv2sf): Ditto.
6090 (aarch64_fmlsl_highv2sf): Ditto.
6091 (aarch64_fmlalq_highv4sf): Ditto.
6092 (aarch64_fmlslq_highv4sf): Ditto.
6093 (aarch64_fmlal_lane_lowv2sf): Ditto.
6094 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6095 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6096 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6097 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6098 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6099 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6100 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6101 (aarch64_fmlal_lane_highv2sf): Ditto.
6102 (aarch64_fmlsl_lane_highv2sf): Ditto.
6103 (aarch64_fmlal_laneq_highv2sf): Ditto.
6104 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6105 (aarch64_fmlalq_lane_highv4sf): Ditto.
6106 (aarch64_fmlsl_lane_highv4sf): Ditto.
6107 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6108 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6109 * config/aarch64/aarch64-simd.md:
6110 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6111 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6112 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6113 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6114 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6115 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6116 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6117 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6118 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6119 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6120 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6121 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6122 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6123 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6124 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6125 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6126 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6127 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6128 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6129 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6130 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6131 (vfmlsl_low_u32): Ditto.
6132 (vfmlalq_low_u32): Ditto.
6133 (vfmlslq_low_u32): Ditto.
6134 (vfmlal_high_u32): Ditto.
6135 (vfmlsl_high_u32): Ditto.
6136 (vfmlalq_high_u32): Ditto.
6137 (vfmlslq_high_u32): Ditto.
6138 (vfmlal_lane_low_u32): Ditto.
6139 (vfmlsl_lane_low_u32): Ditto.
6140 (vfmlal_laneq_low_u32): Ditto.
6141 (vfmlsl_laneq_low_u32): Ditto.
6142 (vfmlalq_lane_low_u32): Ditto.
6143 (vfmlslq_lane_low_u32): Ditto.
6144 (vfmlalq_laneq_low_u32): Ditto.
6145 (vfmlslq_laneq_low_u32): Ditto.
6146 (vfmlal_lane_high_u32): Ditto.
6147 (vfmlsl_lane_high_u32): Ditto.
6148 (vfmlal_laneq_high_u32): Ditto.
6149 (vfmlsl_laneq_high_u32): Ditto.
6150 (vfmlalq_lane_high_u32): Ditto.
6151 (vfmlslq_lane_high_u32): Ditto.
6152 (vfmlalq_laneq_high_u32): Ditto.
6153 (vfmlslq_laneq_high_u32): Ditto.
6154 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6155 (AARCH64_FL_FOR_ARCH8_4): New.
6156 (AARCH64_ISA_F16FML): New ISA flag.
6157 (TARGET_F16FML): New feature flag for fp16fml.
6158 (doc/invoke.texi): Document new fp16fml option.
6160 2018-01-10 Michael Collison <michael.collison@arm.com>
6162 * config/aarch64/aarch64-builtins.c:
6163 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6164 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6165 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6166 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6167 (AARCH64_ISA_SHA3): New ISA flag.
6168 (TARGET_SHA3): New feature flag for sha3.
6169 * config/aarch64/iterators.md (sha512_op): New int attribute.
6170 (CRYPTO_SHA512): New int iterator.
6171 (UNSPEC_SHA512H): New unspec.
6172 (UNSPEC_SHA512H2): Ditto.
6173 (UNSPEC_SHA512SU0): Ditto.
6174 (UNSPEC_SHA512SU1): Ditto.
6175 * config/aarch64/aarch64-simd-builtins.def
6176 (aarch64_crypto_sha512hqv2di): New builtin.
6177 (aarch64_crypto_sha512h2qv2di): Ditto.
6178 (aarch64_crypto_sha512su0qv2di): Ditto.
6179 (aarch64_crypto_sha512su1qv2di): Ditto.
6180 (aarch64_eor3qv8hi): Ditto.
6181 (aarch64_rax1qv2di): Ditto.
6182 (aarch64_xarqv2di): Ditto.
6183 (aarch64_bcaxqv8hi): Ditto.
6184 * config/aarch64/aarch64-simd.md:
6185 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6186 (aarch64_crypto_sha512su0qv2di): Ditto.
6187 (aarch64_crypto_sha512su1qv2di): Ditto.
6188 (aarch64_eor3qv8hi): Ditto.
6189 (aarch64_rax1qv2di): Ditto.
6190 (aarch64_xarqv2di): Ditto.
6191 (aarch64_bcaxqv8hi): Ditto.
6192 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6193 (vsha512h2q_u64): Ditto.
6194 (vsha512su0q_u64): Ditto.
6195 (vsha512su1q_u64): Ditto.
6196 (veor3q_u16): Ditto.
6197 (vrax1q_u64): Ditto.
6199 (vbcaxq_u16): Ditto.
6200 * config/arm/types.md (crypto_sha512): New type attribute.
6201 (crypto_sha3): Ditto.
6202 (doc/invoke.texi): Document new sha3 option.
6204 2018-01-10 Michael Collison <michael.collison@arm.com>
6206 * config/aarch64/aarch64-builtins.c:
6207 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6208 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6209 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6210 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6211 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6212 (AARCH64_ISA_SM4): New ISA flag.
6213 (TARGET_SM4): New feature flag for sm4.
6214 * config/aarch64/aarch64-simd-builtins.def
6215 (aarch64_sm3ss1qv4si): Ditto.
6216 (aarch64_sm3tt1aq4si): Ditto.
6217 (aarch64_sm3tt1bq4si): Ditto.
6218 (aarch64_sm3tt2aq4si): Ditto.
6219 (aarch64_sm3tt2bq4si): Ditto.
6220 (aarch64_sm3partw1qv4si): Ditto.
6221 (aarch64_sm3partw2qv4si): Ditto.
6222 (aarch64_sm4eqv4si): Ditto.
6223 (aarch64_sm4ekeyqv4si): Ditto.
6224 * config/aarch64/aarch64-simd.md:
6225 (aarch64_sm3ss1qv4si): Ditto.
6226 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6227 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6228 (aarch64_sm4eqv4si): Ditto.
6229 (aarch64_sm4ekeyqv4si): Ditto.
6230 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6231 (sm3part_op): Ditto.
6232 (CRYPTO_SM3TT): Ditto.
6233 (CRYPTO_SM3PART): Ditto.
6234 (UNSPEC_SM3SS1): New unspec.
6235 (UNSPEC_SM3TT1A): Ditto.
6236 (UNSPEC_SM3TT1B): Ditto.
6237 (UNSPEC_SM3TT2A): Ditto.
6238 (UNSPEC_SM3TT2B): Ditto.
6239 (UNSPEC_SM3PARTW1): Ditto.
6240 (UNSPEC_SM3PARTW2): Ditto.
6241 (UNSPEC_SM4E): Ditto.
6242 (UNSPEC_SM4EKEY): Ditto.
6243 * config/aarch64/constraints.md (Ui2): New constraint.
6244 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6245 * config/arm/types.md (crypto_sm3): New type attribute.
6246 (crypto_sm4): Ditto.
6247 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6248 (vsm3tt1aq_u32): Ditto.
6249 (vsm3tt1bq_u32): Ditto.
6250 (vsm3tt2aq_u32): Ditto.
6251 (vsm3tt2bq_u32): Ditto.
6252 (vsm3partw1q_u32): Ditto.
6253 (vsm3partw2q_u32): Ditto.
6254 (vsm4eq_u32): Ditto.
6255 (vsm4ekeyq_u32): Ditto.
6256 (doc/invoke.texi): Document new sm4 option.
6258 2018-01-10 Michael Collison <michael.collison@arm.com>
6260 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6261 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6262 (AARCH64_FL_FOR_ARCH8_4): New.
6263 (AARCH64_FL_V8_4): New flag.
6264 (doc/invoke.texi): Document new armv8.4-a option.
6266 2018-01-10 Michael Collison <michael.collison@arm.com>
6268 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6269 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6270 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6271 * config/aarch64/aarch64-option-extension.def: Add
6272 AARCH64_OPT_EXTENSION of 'sha2'.
6273 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6274 (crypto): Disable sha2 and aes if crypto disabled.
6275 (crypto): Enable aes and sha2 if enabled.
6276 (simd): Disable sha2 and aes if simd disabled.
6277 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6279 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6280 (TARGET_SHA2): New feature flag for sha2.
6281 (TARGET_AES): New feature flag for aes.
6282 * config/aarch64/aarch64-simd.md:
6283 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6284 conditional on TARGET_AES.
6285 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6286 (aarch64_crypto_sha1hsi): Make pattern conditional
6288 (aarch64_crypto_sha1hv4si): Ditto.
6289 (aarch64_be_crypto_sha1hv4si): Ditto.
6290 (aarch64_crypto_sha1su1v4si): Ditto.
6291 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6292 (aarch64_crypto_sha1su0v4si): Ditto.
6293 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6294 (aarch64_crypto_sha256su0v4si): Ditto.
6295 (aarch64_crypto_sha256su1v4si): Ditto.
6296 (doc/invoke.texi): Document new aes and sha2 options.
6298 2018-01-10 Martin Sebor <msebor@redhat.com>
6300 PR tree-optimization/83781
6301 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6304 2018-01-11 Martin Sebor <msebor@gmail.com>
6305 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6307 PR tree-optimization/83501
6308 PR tree-optimization/81703
6310 * tree-ssa-strlen.c (get_string_cst): Rename...
6311 (get_string_len): ...to this. Handle global constants.
6312 (handle_char_store): Adjust.
6314 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6315 Jim Wilson <jimw@sifive.com>
6317 * config/riscv/riscv-protos.h (riscv_output_return): New.
6318 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6319 (riscv_attribute_table, riscv_output_return),
6320 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6321 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6322 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6324 (riscv_expand_prologue): Add early return for naked function.
6325 (riscv_expand_epilogue): Likewise.
6326 (riscv_function_ok_for_sibcall): Return false for naked function.
6327 (riscv_set_current_function): New.
6328 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6329 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6330 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6331 * doc/extend.texi (RISC-V Function Attributes): New.
6333 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6335 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6336 check for 128-bit long double before checking TCmode.
6337 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6338 128-bit long doubles before checking TFmode or TCmode.
6339 (FLOAT128_IBM_P): Likewise.
6341 2018-01-10 Martin Sebor <msebor@redhat.com>
6343 PR tree-optimization/83671
6344 * builtins.c (c_strlen): Unconditionally return zero for the empty
6346 Use -Warray-bounds for warnings.
6347 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6348 for non-constant array indices with COMPONENT_REF, arrays of
6349 arrays, and pointers to arrays.
6350 (gimple_fold_builtin_strlen): Determine and set length range for
6351 non-constant character arrays.
6353 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6356 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6359 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6361 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6363 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6366 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6367 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6368 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6369 indexed_or_indirect_operand predicate.
6370 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6371 (*vsx_le_perm_load_v8hi): Likewise.
6372 (*vsx_le_perm_load_v16qi): Likewise.
6373 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6374 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6375 (*vsx_le_perm_store_v8hi): Likewise.
6376 (*vsx_le_perm_store_v16qi): Likewise.
6377 (eight unnamed splitters): Likewise.
6379 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6381 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6382 * config/rs6000/emmintrin.h: Likewise.
6383 * config/rs6000/mmintrin.h: Likewise.
6384 * config/rs6000/xmmintrin.h: Likewise.
6386 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6389 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6391 * tree.c (tree_nop_conversion): Return true for location wrapper
6393 (maybe_wrap_with_location): New function.
6394 (selftest::check_strip_nops): New function.
6395 (selftest::test_location_wrappers): New function.
6396 (selftest::tree_c_tests): Call it.
6397 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6398 (maybe_wrap_with_location): New decl.
6399 (EXPR_LOCATION_WRAPPER_P): New macro.
6400 (location_wrapper_p): New inline function.
6401 (tree_strip_any_location_wrapper): New inline function.
6403 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6406 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6407 stack_realign_offset for the largest alignment of stack slot
6409 (ix86_find_max_used_stack_alignment): New function.
6410 (ix86_finalize_stack_frame_flags): Use it. Set
6411 max_used_stack_alignment if we don't realign stack.
6412 * config/i386/i386.h (machine_function): Add
6413 max_used_stack_alignment.
6415 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6417 * config/arm/arm.opt (-mbranch-cost): New option.
6418 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6421 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6424 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6425 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6427 2018-01-10 Richard Biener <rguenther@suse.de>
6430 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6431 early out so it also covers the case where we have a non-NULL
6434 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6436 PR tree-optimization/83753
6437 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6438 for non-strided grouped accesses if the number of elements is 1.
6440 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6443 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6444 * i386.h (TARGET_USE_GATHER): Define.
6445 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6447 2018-01-10 Martin Liska <mliska@suse.cz>
6450 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6451 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6453 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6454 CLEANUP_NO_PARTITIONING is not set.
6456 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6458 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6459 for vectors, as a partial revert of r254296.
6460 * rtl.h (const_vec_p): Delete.
6461 (const_vec_duplicate_p): Don't test for vector CONSTs.
6462 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6463 * expmed.c (make_tree): Likewise.
6466 * common.md (E, F): Use CONSTANT_P instead of checking for
6468 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6469 checking for CONST_VECTOR.
6471 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6474 * predict.c (force_edge_cold): Handle in more sane way edges
6477 2018-01-09 Carl Love <cel@us.ibm.com>
6479 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6481 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6482 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6483 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6484 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6485 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6486 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6487 * config/rs6000/rs6000-protos.h: Add extern defition for
6488 rs6000_generate_float2_double_code.
6489 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6491 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6492 (float2_v2df): Add define_expand.
6494 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6497 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6498 op_mode in the force_to_mode call.
6500 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6502 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6503 instead of checking each element individually.
6504 (aarch64_evpc_uzp): Likewise.
6505 (aarch64_evpc_zip): Likewise.
6506 (aarch64_evpc_ext): Likewise.
6507 (aarch64_evpc_rev): Likewise.
6508 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6509 instead of checking each element individually. Return true without
6511 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6512 whether all selected elements come from the same input, instead of
6513 checking each element individually. Remove calls to gen_rtx_REG,
6514 start_sequence and end_sequence and instead assert that no rtl is
6517 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6519 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6520 order of HIGH and CONST checks.
6522 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6524 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6525 if the destination isn't an SSA_NAME.
6527 2018-01-09 Richard Biener <rguenther@suse.de>
6529 PR tree-optimization/83668
6530 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6532 (canonicalize_loop_form): ... here, renamed from ...
6533 (canonicalize_loop_closed_ssa_form): ... this and amended to
6534 swap successor edges for loop exit blocks to make us use
6535 the RPO order we need for initial schedule generation.
6537 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6539 PR tree-optimization/64811
6540 * match.pd: When optimizing comparisons with Inf, avoid
6541 introducing or losing exceptions from comparisons with NaN.
6543 2018-01-09 Martin Liska <mliska@suse.cz>
6546 * asan.c (shadow_mem_size): Add gcc_assert.
6548 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6550 Don't save registers in main().
6553 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6554 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6555 * config/avr/avr.c (avr_set_current_function): Don't error if
6556 naked, OS_task or OS_main are specified at the same time.
6557 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6559 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6561 * common/config/avr/avr-common.c (avr_option_optimization_table):
6562 Switch on -mmain-is-OS_task for optimizing compilations.
6564 2018-01-09 Richard Biener <rguenther@suse.de>
6566 PR tree-optimization/83572
6567 * graphite.c: Include cfganal.h.
6568 (graphite_transform_loops): Connect infinite loops to exit
6569 and remove fake edges at the end.
6571 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6573 * ipa-inline.c (edge_badness): Revert accidental checkin.
6575 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6578 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6579 symbols; not inline clones.
6581 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6584 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6585 hard registers. Formatting fixes.
6587 PR preprocessor/83722
6588 * gcc.c (try_generate_repro): Pass
6589 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6590 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6593 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6594 Kito Cheng <kito.cheng@gmail.com>
6596 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6597 (riscv_leaf_function_p): Delete.
6598 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6600 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6602 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6604 (do_ifelse): New function.
6605 (do_isel): New function.
6606 (do_sub3): New function.
6607 (do_add3): New function.
6608 (do_load_mask_compare): New function.
6609 (do_overlap_load_compare): New function.
6610 (expand_compare_loop): New function.
6611 (expand_block_compare): Call expand_compare_loop() when appropriate.
6612 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6614 (-mblock-compare-inline-loop-limit): New option.
6616 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6619 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6620 Reverse order of second and third operands in first alternative.
6621 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6622 of first and second elements in UNSPEC_VPERMR vector.
6623 (altivec_expand_vec_perm_le): Likewise.
6625 2018-01-08 Jeff Law <law@redhat.com>
6627 PR rtl-optimizatin/81308
6628 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6629 (process_switch): If group_case_labels makes a change, then set
6631 (pass_convert_switch::execute): If a switch is converted, then
6632 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6634 PR rtl-optimization/81308
6635 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6638 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6640 PR target/83663 - Revert r255946
6641 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6642 generation for cases where splatting a value is not useful.
6643 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6644 across a vec_duplicate and a paradoxical subreg forming a vector
6645 mode to a vec_concat.
6647 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6649 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6650 -march=armv8.3-a variants.
6651 * config/arm/t-multilib: Likewise.
6652 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6654 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6656 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6658 (cceq_ior_compare_complement): Give it a name so I can use it, and
6659 change boolean_or_operator predicate to boolean_operator so it can
6660 be used to generate a crand.
6661 (eqne): New code iterator.
6662 (bd/bd_neg): New code_attrs.
6663 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6664 a single define_insn.
6665 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6666 decrement (bdnzt/bdnzf/bdzt/bdzf).
6667 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6668 with the new names of the branch decrement patterns, and added the
6669 names of the branch decrement conditional patterns.
6671 2018-01-08 Richard Biener <rguenther@suse.de>
6673 PR tree-optimization/83563
6674 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6677 2018-01-08 Richard Biener <rguenther@suse.de>
6680 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6682 2018-01-08 Richard Biener <rguenther@suse.de>
6684 PR tree-optimization/83685
6685 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6686 references to abnormals.
6688 2018-01-08 Richard Biener <rguenther@suse.de>
6691 * dwarf2out.c (output_indirect_strings): Handle empty
6692 skeleton_debug_str_hash.
6693 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6695 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6697 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6698 (emit_store_direct): Likewise.
6699 (arc_trampoline_adjust_address): Likewise.
6700 (arc_asm_trampoline_template): New function.
6701 (arc_initialize_trampoline): Use asm_trampoline_template.
6702 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6703 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6704 * config/arc/arc.md (flush_icache): Delete pattern.
6706 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6708 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6709 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6712 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6715 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6716 by not USED_FOR_TARGET.
6717 (make_pass_resolve_sw_modes): Likewise.
6719 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6721 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6724 2018-01-08 Richard Biener <rguenther@suse.de>
6727 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6729 2018-01-08 Richard Biener <rguenther@suse.de>
6732 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6734 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6737 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6738 basic blocks with a small number of successors.
6739 (convert_control_dep_chain_into_preds): Improve handling of
6741 (dump_predicates): Split apart into...
6742 (dump_pred_chain): ...here...
6743 (dump_pred_info): ...and here.
6744 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6745 (can_chain_union_be_invalidated_p): Improve check for invalidation
6747 (uninit_uses_cannot_happen): Avoid unnecessary if
6748 convert_control_dep_chain_into_preds yielded nothing.
6750 2018-01-06 Martin Sebor <msebor@redhat.com>
6752 PR tree-optimization/83640
6753 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6754 subtracting negative offset from size.
6755 (builtin_access::overlap): Adjust offset bounds of the access to fall
6756 within the size of the object if possible.
6758 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6760 PR rtl-optimization/83699
6761 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6762 extract_bit_field_as_subreg to cases in which the extracted
6763 value is also a vector.
6765 * lra-constraints.c (process_alt_operands): Test for the equivalence
6766 substitutions when detecting a possible reload cycle.
6768 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6771 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6772 by default if flag_selective_schedling{,2}. Formatting fixes.
6774 PR rtl-optimization/83682
6775 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6776 if it has non-VECTOR_MODE element mode.
6777 (vec_duplicate_p): Likewise.
6780 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6781 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6783 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6786 * config/i386/i386-builtin.def
6787 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6788 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6789 Require also OPTION_MASK_ISA_AVX512F in addition to
6790 OPTION_MASK_ISA_GFNI.
6791 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6792 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6793 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6794 to OPTION_MASK_ISA_GFNI.
6795 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6796 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6797 OPTION_MASK_ISA_AVX512BW.
6798 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6799 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6800 addition to OPTION_MASK_ISA_GFNI.
6801 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6802 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6803 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6804 to OPTION_MASK_ISA_GFNI.
6805 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6806 a requirement for all ISAs rather than any of them with a few
6808 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6810 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6811 bitmasks to be enabled with 3 exceptions, instead of requiring any
6812 enabled ISA with lots of exceptions.
6813 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6814 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6815 Change avx512bw in isa attribute to avx512f.
6816 * config/i386/sgxintrin.h: Add license boilerplate.
6817 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6818 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6819 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6820 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6822 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6823 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6824 temporarily sse2 rather than sse if not enabled already.
6827 * config/i386/sse.md (VI248_VLBW): Rename to ...
6828 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6829 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6830 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6831 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6832 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6833 mode iterator instead of VI248_VLBW.
6835 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6837 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6838 (record_modified): Skip clobbers; add debug output.
6839 (param_change_prob): Use sreal frequencies.
6841 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6843 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6844 punt for user-aligned variables.
6846 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6848 * tree-chrec.c (chrec_contains_symbols): Return true for
6851 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6854 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6855 of (x|y) == x for BICS pattern.
6857 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6859 PR tree-optimization/83605
6860 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6861 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6864 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6866 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6867 * config/epiphany/rtems.h: New file.
6869 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6870 Uros Bizjak <ubizjak@gmail.com>
6873 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6874 QIreg_operand instead of register_operand predicate.
6875 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6876 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6877 comments instead of -fmitigate[-_]rop.
6879 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6882 * cgraphunit.c (symbol_table::compile): Switch to text_section
6883 before calling assembly_start debug hook.
6884 * run-rtl-passes.c (run_rtl_passes): Likewise.
6887 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6889 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6890 range_int_cst_p rather than !symbolic_range_p before calling
6891 extract_range_from_multiplicative_op_1.
6893 2018-01-04 Jeff Law <law@redhat.com>
6895 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6896 redundant test in assertion.
6898 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6900 * doc/rtl.texi: Document machine_mode wrapper classes.
6902 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6904 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6907 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6909 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6910 the VEC_PERM_EXPR fold to fail.
6912 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6915 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6916 to switched_sections.
6918 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6921 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6924 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6927 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6928 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6930 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6933 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6934 is BLKmode and bitpos not zero or mode change is needed.
6936 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6939 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6942 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6945 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6946 instead of MULT rtx. Update all corresponding splitters.
6948 (*ssub<modesuffix>): Ditto.
6950 (*cmp_sadd_di): Update split patterns.
6951 (*cmp_sadd_si): Ditto.
6952 (*cmp_sadd_sidi): Ditto.
6953 (*cmp_ssub_di): Ditto.
6954 (*cmp_ssub_si): Ditto.
6955 (*cmp_ssub_sidi): Ditto.
6956 * config/alpha/predicates.md (const23_operand): New predicate.
6957 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6958 Look for ASHIFT, not MULT inner operand.
6959 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6961 2018-01-04 Martin Liska <mliska@suse.cz>
6963 PR gcov-profile/83669
6964 * gcov.c (output_intermediate_file): Add version to intermediate
6966 * doc/gcov.texi: Document new field 'version' in intermediate
6967 file format. Fix location of '-k' option of gcov command.
6969 2018-01-04 Martin Liska <mliska@suse.cz>
6972 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6974 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6976 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6978 2018-01-03 Martin Sebor <msebor@redhat.com>
6980 PR tree-optimization/83655
6981 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6982 checking calls with invalid arguments.
6984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6986 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6987 (vectorizable_mask_load_store): Delete.
6988 (vectorizable_call): Return false for masked loads and stores.
6989 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6990 instead of gimple_assign_rhs1.
6991 (vectorizable_load): Handle IFN_MASK_LOAD.
6992 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6996 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6998 (vectorizable_mask_load_store): ...here.
6999 (vectorizable_load): ...and here.
7001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7003 * tree-vect-stmts.c (vect_build_all_ones_mask)
7004 (vect_build_zero_merge_argument): New functions, split out from...
7005 (vectorizable_load): ...here.
7007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7009 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7011 (vectorizable_mask_load_store): ...here.
7012 (vectorizable_store): ...and here.
7014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7016 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7018 (vectorizable_mask_load_store): ...here.
7020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7022 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7023 (vect_model_store_cost): Take a vec_load_store_type instead of a
7025 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7026 (vect_model_store_cost): Take a vec_load_store_type instead of a
7028 (vectorizable_mask_load_store): Update accordingly.
7029 (vectorizable_store): Likewise.
7030 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7034 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7035 IFN_MASK_LOAD calls here rather than...
7036 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7039 Alan Hayward <alan.hayward@arm.com>
7040 David Sherwood <david.sherwood@arm.com>
7042 * expmed.c (extract_bit_field_1): For vector extracts,
7043 fall back to extract_bit_field_as_subreg if vec_extract
7046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7047 Alan Hayward <alan.hayward@arm.com>
7048 David Sherwood <david.sherwood@arm.com>
7050 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7051 they are variable or constant sized.
7052 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7053 slots for constant-sized data.
7055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7056 Alan Hayward <alan.hayward@arm.com>
7057 David Sherwood <david.sherwood@arm.com>
7059 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7060 handling COND_EXPRs with boolean comparisons, try to find a better
7061 basis for the mask type than the boolean itself.
7063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7065 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7066 is calculated and how it can be overridden.
7067 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7068 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7070 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7074 Alan Hayward <alan.hayward@arm.com>
7075 David Sherwood <david.sherwood@arm.com>
7077 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7078 Remove the mode argument.
7079 (aarch64_simd_valid_immediate): Remove the mode and inverse
7081 * config/aarch64/iterators.md (bitsize): New iterator.
7082 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7083 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7084 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7085 aarch64_simd_valid_immediate.
7086 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7087 (aarch64_reg_or_bic_imm): Likewise.
7088 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7089 with an insn_type enum and msl with a modifier_type enum.
7090 Replace element_width with a scalar_mode. Change the shift
7091 to unsigned int. Add constructors for scalar_float_mode and
7092 scalar_int_mode elements.
7093 (aarch64_vect_float_const_representable_p): Delete.
7094 (aarch64_can_const_movi_rtx_p)
7095 (aarch64_simd_scalar_immediate_valid_for_move)
7096 (aarch64_simd_make_constant): Update call to
7097 aarch64_simd_valid_immediate.
7098 (aarch64_advsimd_valid_immediate_hs): New function.
7099 (aarch64_advsimd_valid_immediate): Likewise.
7100 (aarch64_simd_valid_immediate): Remove mode and inverse
7101 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7102 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7103 and aarch64_float_const_representable_p on the result.
7104 (aarch64_output_simd_mov_immediate): Remove mode argument.
7105 Update call to aarch64_simd_valid_immediate and use of
7106 simd_immediate_info.
7107 (aarch64_output_scalar_simd_mov_immediate): Update call
7110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7111 Alan Hayward <alan.hayward@arm.com>
7112 David Sherwood <david.sherwood@arm.com>
7114 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7115 (mode_nunits): Likewise CONST_MODE_NUNITS.
7116 * machmode.def (ADJUST_NUNITS): Document.
7117 * genmodes.c (mode_data::need_nunits_adj): New field.
7118 (blank_mode): Update accordingly.
7119 (adj_nunits): New variable.
7120 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7122 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7123 listed in adj_nunits.
7124 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7125 listed in adj_nunits. Don't emit case statements for such modes.
7126 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7127 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7128 nothing if adj_nunits is nonnull.
7129 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7130 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7131 (emit_mode_fbit): Update use of print_maybe_const_decl.
7132 (emit_move_size): Likewise. Treat the array as non-const
7134 (emit_mode_adjustments): Handle adj_nunits.
7136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7138 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7139 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7140 (VECTOR_MODES): Use it.
7141 (make_vector_modes): Take the prefix as an argument.
7143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7144 Alan Hayward <alan.hayward@arm.com>
7145 David Sherwood <david.sherwood@arm.com>
7147 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7148 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7149 for MODE_VECTOR_BOOL.
7150 * machmode.def (VECTOR_BOOL_MODE): Document.
7151 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7152 (make_vector_bool_mode): New function.
7153 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7155 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7156 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7158 * stor-layout.c (int_mode_for_mode): Likewise.
7159 * tree.c (build_vector_type_for_mode): Likewise.
7160 * varasm.c (output_constant_pool_2): Likewise.
7161 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7162 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7163 for MODE_VECTOR_BOOL.
7164 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7165 of mode class checks.
7166 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7167 instead of a list of mode class checks.
7168 (expand_vector_scalar_condition): Likewise.
7169 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7172 Alan Hayward <alan.hayward@arm.com>
7173 David Sherwood <david.sherwood@arm.com>
7175 * machmode.h (mode_size): Change from unsigned short to
7177 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7178 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7179 or if measurement_type is not polynomial.
7180 (fixed_size_mode::includes_p): Check for constant-sized modes.
7181 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7182 return a poly_uint16 rather than an unsigned short.
7183 (emit_mode_size): Change the type of mode_size from unsigned short
7184 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7185 (emit_mode_adjustments): Cope with polynomial vector sizes.
7186 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7188 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7190 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7191 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7192 * caller-save.c (setup_save_areas): Likewise.
7193 (replace_reg_with_saved_mem): Likewise.
7194 * calls.c (emit_library_call_value_1): Likewise.
7195 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7196 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7197 (gen_lowpart_for_combine): Likewise.
7198 * convert.c (convert_to_integer_1): Likewise.
7199 * cse.c (equiv_constant, cse_insn): Likewise.
7200 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7201 (cselib_subst_to_values): Likewise.
7202 * dce.c (word_dce_process_block): Likewise.
7203 * df-problems.c (df_word_lr_mark_ref): Likewise.
7204 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7205 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7206 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7207 (rtl_for_decl_location): Likewise.
7208 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7209 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7210 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7211 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7212 (expand_expr_real_1): Likewise.
7213 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7214 (pad_below): Likewise.
7215 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7216 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7217 * ira.c (get_subreg_tracking_sizes): Likewise.
7218 * ira-build.c (ira_create_allocno_objects): Likewise.
7219 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7220 (ira_sort_regnos_for_alter_reg): Likewise.
7221 * ira-costs.c (record_operand_costs): Likewise.
7222 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7223 (resolve_simple_move): Likewise.
7224 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7225 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7226 (lra_constraints): Likewise.
7227 (CONST_POOL_OK_P): Reject variable-sized modes.
7228 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7229 (add_pseudo_to_slot, lra_spill): Likewise.
7230 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7231 * optabs-query.c (get_best_extraction_insn): Likewise.
7232 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7233 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7234 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7235 * recog.c (offsettable_address_addr_space_p): Likewise.
7236 * regcprop.c (maybe_mode_change): Likewise.
7237 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7238 * regrename.c (build_def_use): Likewise.
7239 * regstat.c (dump_reg_info): Likewise.
7240 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7241 (find_reloads, find_reloads_subreg_address): Likewise.
7242 * reload1.c (eliminate_regs_1): Likewise.
7243 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7244 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7245 (simplify_binary_operation_1, simplify_subreg): Likewise.
7246 * targhooks.c (default_function_arg_padding): Likewise.
7247 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7248 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7249 (verify_gimple_assign_ternary): Likewise.
7250 * tree-inline.c (estimate_move_cost): Likewise.
7251 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7252 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7253 (get_address_cost_ainc): Likewise.
7254 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7255 (vect_supportable_dr_alignment): Likewise.
7256 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7257 (vectorizable_reduction): Likewise.
7258 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7259 (vectorizable_operation, vectorizable_load): Likewise.
7260 * tree.c (build_same_sized_truth_vector_type): Likewise.
7261 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7262 * var-tracking.c (emit_note_insn_var_location): Likewise.
7263 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7264 (ADDR_VEC_ALIGN): Likewise.
7266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7267 Alan Hayward <alan.hayward@arm.com>
7268 David Sherwood <david.sherwood@arm.com>
7270 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7272 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7273 or if measurement_type is polynomial.
7274 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7275 * combine.c (make_extraction): Likewise.
7276 * dse.c (find_shift_sequence): Likewise.
7277 * dwarf2out.c (mem_loc_descriptor): Likewise.
7278 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7279 (extract_bit_field, extract_low_bits): Likewise.
7280 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7281 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7282 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7283 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7284 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7285 * reload.c (find_reloads): Likewise.
7286 * reload1.c (alter_reg): Likewise.
7287 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7288 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7289 * tree-if-conv.c (predicate_mem_writes): Likewise.
7290 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7291 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7292 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7293 * valtrack.c (dead_debug_insert_temp): Likewise.
7294 * varasm.c (mergeable_constant_section): Likewise.
7295 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7298 Alan Hayward <alan.hayward@arm.com>
7299 David Sherwood <david.sherwood@arm.com>
7301 * expr.c (expand_assignment): Cope with polynomial mode sizes
7302 when assigning to a CONCAT.
7304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7305 Alan Hayward <alan.hayward@arm.com>
7306 David Sherwood <david.sherwood@arm.com>
7308 * machmode.h (mode_precision): Change from unsigned short to
7310 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7312 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7313 or if measurement_type is not polynomial.
7314 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7315 in which the mode is already known to be a scalar_int_mode.
7316 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7317 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7319 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7320 for GET_MODE_PRECISION.
7321 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7322 for GET_MODE_PRECISION.
7323 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7325 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7326 (expand_field_assignment, make_extraction): Likewise.
7327 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7328 (get_last_value): Likewise.
7329 * convert.c (convert_to_integer_1): Likewise.
7330 * cse.c (cse_insn): Likewise.
7331 * expr.c (expand_expr_real_1): Likewise.
7332 * lra-constraints.c (simplify_operand_subreg): Likewise.
7333 * optabs-query.c (can_atomic_load_p): Likewise.
7334 * optabs.c (expand_atomic_load): Likewise.
7335 (expand_atomic_store): Likewise.
7336 * ree.c (combine_reaching_defs): Likewise.
7337 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7338 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7339 * tree.h (type_has_mode_precision_p): Likewise.
7340 * ubsan.c (instrument_si_overflow): Likewise.
7342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7343 Alan Hayward <alan.hayward@arm.com>
7344 David Sherwood <david.sherwood@arm.com>
7346 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7347 polynomial numbers of units.
7348 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7349 (valid_vector_subparts_p): New function.
7350 (build_vector_type): Remove temporary shim and take the number
7351 of units as a poly_uint64 rather than an int.
7352 (build_opaque_vector_type): Take the number of units as a
7353 poly_uint64 rather than an int.
7354 * tree.c (build_vector_from_ctor): Handle polynomial
7355 TYPE_VECTOR_SUBPARTS.
7356 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7357 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7358 (build_vector_from_val): If the number of units is variable,
7359 use build_vec_duplicate_cst for constant operands and
7360 VEC_DUPLICATE_EXPR otherwise.
7361 (make_vector_type): Remove temporary is_constant ().
7362 (build_vector_type, build_opaque_vector_type): Take the number of
7363 units as a poly_uint64 rather than an int.
7364 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7366 * cfgexpand.c (expand_debug_expr): Likewise.
7367 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7368 (store_constructor, expand_expr_real_1): Likewise.
7369 (const_scalar_mask_from_tree): Likewise.
7370 * fold-const-call.c (fold_const_reduction): Likewise.
7371 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7372 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7373 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7374 (fold_relational_const): Likewise.
7375 (native_interpret_vector): Likewise. Change the size from an
7376 int to an unsigned int.
7377 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7378 TYPE_VECTOR_SUBPARTS.
7379 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7380 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7381 duplicating a non-constant operand into a variable-length vector.
7382 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7383 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7384 * ipa-icf.c (sem_variable::equals): Likewise.
7385 * match.pd: Likewise.
7386 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7387 * print-tree.c (print_node): Likewise.
7388 * stor-layout.c (layout_type): Likewise.
7389 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7390 * tree-cfg.c (verify_gimple_comparison): Likewise.
7391 (verify_gimple_assign_binary): Likewise.
7392 (verify_gimple_assign_ternary): Likewise.
7393 (verify_gimple_assign_single): Likewise.
7394 * tree-pretty-print.c (dump_generic_node): Likewise.
7395 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7396 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7397 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7398 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7399 (vect_shift_permute_load_chain): Likewise.
7400 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7401 (expand_vector_condition, optimize_vector_constructor): Likewise.
7402 (lower_vec_perm, get_compute_type): Likewise.
7403 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7404 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7405 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7406 (vect_recog_mask_conversion_pattern): Likewise.
7407 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7408 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7409 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7410 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7411 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7412 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7413 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7414 (supportable_widening_operation): Likewise.
7415 (supportable_narrowing_operation): Likewise.
7416 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7418 * varasm.c (output_constant): Likewise.
7420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7421 Alan Hayward <alan.hayward@arm.com>
7422 David Sherwood <david.sherwood@arm.com>
7424 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7425 so that both the length == 3 and length != 3 cases set up their
7426 own permute vectors. Add comments explaining why we know the
7427 number of elements is constant.
7428 (vect_permute_load_chain): Likewise.
7430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7431 Alan Hayward <alan.hayward@arm.com>
7432 David Sherwood <david.sherwood@arm.com>
7434 * machmode.h (mode_nunits): Change from unsigned char to
7436 (ONLY_FIXED_SIZE_MODES): New macro.
7437 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7438 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7439 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7441 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7442 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7443 or if measurement_type is not polynomial.
7444 * genmodes.c (ZERO_COEFFS): New macro.
7445 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7447 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7448 Use ZERO_COEFFS when emitting initializers.
7449 * data-streamer.h (bp_pack_poly_value): New function.
7450 (bp_unpack_poly_value): Likewise.
7451 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7452 for GET_MODE_NUNITS.
7453 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7454 for GET_MODE_NUNITS.
7455 * tree.c (make_vector_type): Remove temporary shim and make
7456 the real function take the number of units as a poly_uint64
7458 (build_vector_type_for_mode): Handle polynomial nunits.
7459 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7460 * emit-rtl.c (const_vec_series_p_1): Likewise.
7461 (gen_rtx_CONST_VECTOR): Likewise.
7462 * fold-const.c (test_vec_duplicate_folding): Likewise.
7463 * genrecog.c (validate_pattern): Likewise.
7464 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7465 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7466 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7467 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7468 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7469 * rtlanal.c (subreg_get_info): Likewise.
7470 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7471 (vect_grouped_load_supported): Likewise.
7472 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7473 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7474 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7475 (simplify_const_unary_operation, simplify_binary_operation_1)
7476 (simplify_const_binary_operation, simplify_ternary_operation)
7477 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7478 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7479 instead of CONST_VECTOR_NUNITS.
7480 * varasm.c (output_constant_pool_2): Likewise.
7481 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7482 explicit-encoded elements in the XVEC for variable-length vectors.
7484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7486 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7489 Alan Hayward <alan.hayward@arm.com>
7490 David Sherwood <david.sherwood@arm.com>
7492 * coretypes.h (fixed_size_mode): Declare.
7493 (fixed_size_mode_pod): New typedef.
7494 * builtins.h (target_builtins::x_apply_args_mode)
7495 (target_builtins::x_apply_result_mode): Change type to
7496 fixed_size_mode_pod.
7497 * builtins.c (apply_args_size, apply_result_size, result_vector)
7498 (expand_builtin_apply_args_1, expand_builtin_apply)
7499 (expand_builtin_return): Update accordingly.
7501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7503 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7504 * cselib.c (cselib_hash_rtx): Likewise.
7505 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7506 CONST_VECTOR encoding.
7508 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7509 Jeff Law <law@redhat.com>
7512 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7513 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7514 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7515 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7518 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7519 explicitly probe *sp in a noreturn function if there were any callee
7520 register saves or frame pointer is needed.
7522 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7525 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7526 BLKmode for ternary, binary or unary expressions.
7529 * var-tracking.c (delete_vta_debug_insn): New inline function.
7530 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7531 insns from get_insns () to NULL instead of each bb separately.
7532 Use delete_vta_debug_insn. No longer static.
7533 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7534 delete_vta_debug_insns callers.
7535 * rtl.h (delete_vta_debug_insns): Declare.
7536 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7537 instead of variable_tracking_main.
7539 2018-01-03 Martin Sebor <msebor@redhat.com>
7541 PR tree-optimization/83603
7542 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7543 arguments past the endof the argument list in functions declared
7544 without a prototype.
7545 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7546 Avoid checking when arguments are null.
7548 2018-01-03 Martin Sebor <msebor@redhat.com>
7551 * doc/extend.texi (attribute const): Fix a typo.
7552 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7553 issuing -Wsuggest-attribute for void functions.
7555 2018-01-03 Martin Sebor <msebor@redhat.com>
7557 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7558 offset_int::from instead of wide_int::to_shwi.
7559 (maybe_diag_overlap): Remove assertion.
7560 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7561 * gimple-ssa-sprintf.c (format_directive): Same.
7562 (parse_directive): Same.
7563 (sprintf_dom_walker::compute_format_length): Same.
7564 (try_substitute_return_value): Same.
7566 2018-01-03 Jeff Law <law@redhat.com>
7569 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7570 non-constant residual for zero at runtime and avoid probing in
7571 that case. Reorganize code for trailing problem to mirror handling
7574 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7576 PR tree-optimization/83501
7577 * tree-ssa-strlen.c (get_string_cst): New.
7578 (handle_char_store): Call get_string_cst.
7580 2018-01-03 Martin Liska <mliska@suse.cz>
7582 PR tree-optimization/83593
7583 * tree-ssa-strlen.c: Include tree-cfg.h.
7584 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7585 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7586 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7588 (strlen_dom_walker::before_dom_children): Call
7589 gimple_purge_dead_eh_edges. Dump tranformation with details
7591 (strlen_dom_walker::before_dom_children): Update call by adding
7592 new argument cleanup_eh.
7593 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7595 2018-01-03 Martin Liska <mliska@suse.cz>
7598 * cif-code.def (VARIADIC_THUNK): New enum value.
7599 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7602 2018-01-03 Jan Beulich <jbeulich@suse.com>
7604 * sse.md (mov<mode>_internal): Tighten condition for when to use
7605 vmovdqu<ssescalarsize> for TI and OI modes.
7607 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7609 Update copyright years.
7611 2018-01-03 Martin Liska <mliska@suse.cz>
7614 * ipa-visibility.c (function_and_variable_visibility): Skip
7615 functions with noipa attribure.
7617 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7619 * gcc.c (process_command): Update copyright notice dates.
7620 * gcov-dump.c (print_version): Ditto.
7621 * gcov.c (print_version): Ditto.
7622 * gcov-tool.c (print_version): Ditto.
7623 * gengtype.c (create_file): Ditto.
7624 * doc/cpp.texi: Bump @copying's copyright year.
7625 * doc/cppinternals.texi: Ditto.
7626 * doc/gcc.texi: Ditto.
7627 * doc/gccint.texi: Ditto.
7628 * doc/gcov.texi: Ditto.
7629 * doc/install.texi: Ditto.
7630 * doc/invoke.texi: Ditto.
7632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7634 * vector-builder.h (vector_builder::m_full_nelts): Change from
7635 unsigned int to poly_uint64.
7636 (vector_builder::full_nelts): Update prototype accordingly.
7637 (vector_builder::new_vector): Likewise.
7638 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7639 (vector_builder::operator ==): Likewise.
7640 (vector_builder::finalize): Likewise.
7641 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7642 Take the number of elements as a poly_uint64 rather than an
7644 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7645 from unsigned int to poly_uint64.
7646 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7647 (vec_perm_indices::new_vector): Likewise.
7648 (vec_perm_indices::length): Likewise.
7649 (vec_perm_indices::nelts_per_input): Likewise.
7650 (vec_perm_indices::input_nelts): Likewise.
7651 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7652 number of elements per input as a poly_uint64 rather than an
7653 unsigned int. Use the original encoding for variable-length
7654 vectors, rather than clamping each individual element.
7655 For the second and subsequent elements in each pattern,
7656 clamp the step and base before clamping their sum.
7657 (vec_perm_indices::series_p): Handle polynomial element counts.
7658 (vec_perm_indices::all_in_range_p): Likewise.
7659 (vec_perm_indices_to_tree): Likewise.
7660 (vec_perm_indices_to_rtx): Likewise.
7661 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7662 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7663 (tree_vector_builder::new_binary_operation): Handle polynomial
7664 element counts. Return false if we need to know the number
7665 of elements at compile time.
7666 * fold-const.c (fold_vec_perm): Punt if the number of elements
7667 isn't known at compile time.
7669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7671 * vec-perm-indices.h (vec_perm_builder): Change element type
7672 from HOST_WIDE_INT to poly_int64.
7673 (vec_perm_indices::element_type): Update accordingly.
7674 (vec_perm_indices::clamp): Handle polynomial element_types.
7675 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7676 (vec_perm_indices::all_in_range_p): Likewise.
7677 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7679 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7680 polynomial vec_perm_indices element types.
7681 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7682 * fold-const.c (fold_vec_perm): Likewise.
7683 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7684 * tree-vect-generic.c (lower_vec_perm): Likewise.
7685 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7686 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7687 element type to HOST_WIDE_INT.
7689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7690 Alan Hayward <alan.hayward@arm.com>
7691 David Sherwood <david.sherwood@arm.com>
7693 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7694 rather than an int. Use plus_constant.
7695 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7696 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7699 Alan Hayward <alan.hayward@arm.com>
7700 David Sherwood <david.sherwood@arm.com>
7702 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7703 a HOST_WIDE_INT to a poly_int64.
7705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7706 Alan Hayward <alan.hayward@arm.com>
7707 David Sherwood <david.sherwood@arm.com>
7709 * calls.c (load_register_parameters): Cope with polynomial
7710 mode sizes. Require a constant size for BLKmode parameters
7711 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7712 forces a parameter to be padded at the lsb end in order to
7713 fill a complete number of words, require the parameter size
7714 to be ordered wrt UNITS_PER_WORD.
7716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7717 Alan Hayward <alan.hayward@arm.com>
7718 David Sherwood <david.sherwood@arm.com>
7720 * reload1.c (spill_stack_slot_width): Change element type
7721 from unsigned int to poly_uint64_pod.
7722 (alter_reg): Treat mode sizes as polynomial.
7724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7725 Alan Hayward <alan.hayward@arm.com>
7726 David Sherwood <david.sherwood@arm.com>
7728 * reload.c (complex_word_subreg_p): New function.
7729 (reload_inner_reg_of_subreg, push_reload): Use it.
7731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7732 Alan Hayward <alan.hayward@arm.com>
7733 David Sherwood <david.sherwood@arm.com>
7735 * lra-constraints.c (process_alt_operands): Reject matched
7736 operands whose sizes aren't ordered.
7737 (match_reload): Refer to this check here.
7739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7740 Alan Hayward <alan.hayward@arm.com>
7741 David Sherwood <david.sherwood@arm.com>
7743 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7744 that the mode size is in the set {1, 2, 4, 8, 16}.
7746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7747 Alan Hayward <alan.hayward@arm.com>
7748 David Sherwood <david.sherwood@arm.com>
7750 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7751 Use plus_constant instead of gen_rtx_PLUS.
7753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7754 Alan Hayward <alan.hayward@arm.com>
7755 David Sherwood <david.sherwood@arm.com>
7757 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7758 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7759 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7760 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7761 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7762 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7763 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7764 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7765 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7766 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7768 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7769 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7770 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7771 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7772 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7773 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7774 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7775 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7776 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7777 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7779 * expr.c (emit_move_resolve_push): Treat the input and result
7780 of PUSH_ROUNDING as a poly_int64.
7781 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7782 (emit_push_insn): Likewise.
7783 * lra-eliminations.c (mark_not_eliminable): Likewise.
7784 * recog.c (push_operand): Likewise.
7785 * reload1.c (elimination_effects): Likewise.
7786 * rtlanal.c (nonzero_bits1): Likewise.
7787 * calls.c (store_one_arg): Likewise. Require the padding to be
7788 known at compile time.
7790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7791 Alan Hayward <alan.hayward@arm.com>
7792 David Sherwood <david.sherwood@arm.com>
7794 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7795 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
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 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7809 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7810 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
7817 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7818 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7819 as a poly_uint64 too.
7821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7822 Alan Hayward <alan.hayward@arm.com>
7823 David Sherwood <david.sherwood@arm.com>
7825 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7828 Alan Hayward <alan.hayward@arm.com>
7829 David Sherwood <david.sherwood@arm.com>
7831 * combine.c (can_change_dest_mode): Handle polynomial
7832 REGMODE_NATURAL_SIZE.
7833 * expmed.c (store_bit_field_1): Likewise.
7834 * expr.c (store_constructor): Likewise.
7835 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7836 and polynomial REGMODE_NATURAL_SIZE.
7837 (gen_lowpart_common): Likewise.
7838 * reginfo.c (record_subregs_of_mode): Likewise.
7839 * rtlanal.c (read_modify_subreg_p): Likewise.
7841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7842 Alan Hayward <alan.hayward@arm.com>
7843 David Sherwood <david.sherwood@arm.com>
7845 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7846 numbers of elements.
7848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7849 Alan Hayward <alan.hayward@arm.com>
7850 David Sherwood <david.sherwood@arm.com>
7852 * match.pd: Cope with polynomial numbers of vector elements.
7854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7855 Alan Hayward <alan.hayward@arm.com>
7856 David Sherwood <david.sherwood@arm.com>
7858 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7859 in a POINTER_PLUS_EXPR.
7861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7862 Alan Hayward <alan.hayward@arm.com>
7863 David Sherwood <david.sherwood@arm.com>
7865 * omp-simd-clone.c (simd_clone_subparts): New function.
7866 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7867 (ipa_simd_modify_function_body): Likewise.
7869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7870 Alan Hayward <alan.hayward@arm.com>
7871 David Sherwood <david.sherwood@arm.com>
7873 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7874 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7875 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7876 (expand_vector_condition, vector_element): Likewise.
7877 (subparts_gt): New function.
7878 (get_compute_type): Use subparts_gt.
7879 (count_type_subparts): Delete.
7880 (expand_vector_operations_1): Use subparts_gt instead of
7881 count_type_subparts.
7883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7884 Alan Hayward <alan.hayward@arm.com>
7885 David Sherwood <david.sherwood@arm.com>
7887 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7888 (vect_compile_time_alias): ...this new function. Do the calculation
7889 on poly_ints rather than trees.
7890 (vect_prune_runtime_alias_test_list): Update call accordingly.
7892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7893 Alan Hayward <alan.hayward@arm.com>
7894 David Sherwood <david.sherwood@arm.com>
7896 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7898 (vect_schedule_slp_instance): Likewise.
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 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7905 constant and extern definitions for variable-length vectors.
7906 (vect_get_constant_vectors): Note that the number of units
7907 is known to be constant.
7909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7910 Alan Hayward <alan.hayward@arm.com>
7911 David Sherwood <david.sherwood@arm.com>
7913 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7914 of units as polynomial. Choose between WIDE and NARROW based
7917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7918 Alan Hayward <alan.hayward@arm.com>
7919 David Sherwood <david.sherwood@arm.com>
7921 * tree-vect-stmts.c (simd_clone_subparts): New function.
7922 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7925 Alan Hayward <alan.hayward@arm.com>
7926 David Sherwood <david.sherwood@arm.com>
7928 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7929 vectors as polynomial. Use build_index_vector for
7932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7933 Alan Hayward <alan.hayward@arm.com>
7934 David Sherwood <david.sherwood@arm.com>
7936 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7937 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7938 for variable-length vectors.
7939 (vectorizable_mask_load_store): Treat the number of units as
7940 polynomial, asserting that it is constant if the condition has
7941 already been enforced.
7942 (vectorizable_store, vectorizable_load): Likewise.
7944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7945 Alan Hayward <alan.hayward@arm.com>
7946 David Sherwood <david.sherwood@arm.com>
7948 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7949 of units as polynomial. Punt if we can't tell at compile time
7950 which vector contains the final result.
7952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7953 Alan Hayward <alan.hayward@arm.com>
7954 David Sherwood <david.sherwood@arm.com>
7956 * tree-vect-loop.c (vectorizable_induction): Treat the number
7957 of units as polynomial. Punt on SLP inductions. Use an integer
7958 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7959 cast of such a series for variable-length floating-point
7962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7963 Alan Hayward <alan.hayward@arm.com>
7964 David Sherwood <david.sherwood@arm.com>
7966 * tree.h (build_index_vector): Declare.
7967 * tree.c (build_index_vector): New function.
7968 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7969 of units as polynomial, forcibly converting it to a constant if
7970 vectorizable_reduction has already enforced the condition.
7971 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7972 to create a {1,2,3,...} vector.
7973 (vectorizable_reduction): Treat the number of units as polynomial.
7974 Choose vectype_in based on the largest scalar element size rather
7975 than the smallest number of units. Enforce the restrictions
7978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7979 Alan Hayward <alan.hayward@arm.com>
7980 David Sherwood <david.sherwood@arm.com>
7982 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7983 number of units as polynomial.
7985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7986 Alan Hayward <alan.hayward@arm.com>
7987 David Sherwood <david.sherwood@arm.com>
7989 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7990 * target.def (autovectorize_vector_sizes): Return the vector sizes
7991 by pointer, using vector_sizes rather than a bitmask.
7992 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7993 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7994 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7996 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7997 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7998 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7999 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8000 * omp-general.c (omp_max_vf): Likewise.
8001 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8002 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8003 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8004 * tree-vect-slp.c (vect_slp_bb): Likewise.
8005 * doc/tm.texi: Regenerate.
8006 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8008 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8009 the vector size as a poly_uint64 rather than an unsigned int.
8010 (current_vector_size): Change from an unsigned int to a poly_uint64.
8011 (get_vectype_for_scalar_type): Update accordingly.
8012 * tree.h (build_truth_vector_type): Take the size and number of
8013 units as a poly_uint64 rather than an unsigned int.
8014 (build_vector_type): Add a temporary overload that takes
8015 the number of units as a poly_uint64 rather than an unsigned int.
8016 * tree.c (make_vector_type): Likewise.
8017 (build_truth_vector_type): Take the number of units as a poly_uint64
8018 rather than an unsigned int.
8020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8021 Alan Hayward <alan.hayward@arm.com>
8022 David Sherwood <david.sherwood@arm.com>
8024 * target.def (get_mask_mode): Take the number of units and length
8025 as poly_uint64s rather than unsigned ints.
8026 * targhooks.h (default_get_mask_mode): Update accordingly.
8027 * targhooks.c (default_get_mask_mode): Likewise.
8028 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8029 * doc/tm.texi: Regenerate.
8031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8032 Alan Hayward <alan.hayward@arm.com>
8033 David Sherwood <david.sherwood@arm.com>
8035 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8036 * omp-general.c (omp_max_vf): Likewise.
8037 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8038 (expand_omp_simd): Handle polynomial safelen.
8039 * omp-low.c (omplow_simd_context): Add a default constructor.
8040 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8041 (lower_rec_simd_input_clauses): Update accordingly.
8042 (lower_rec_input_clauses): Likewise.
8044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8045 Alan Hayward <alan.hayward@arm.com>
8046 David Sherwood <david.sherwood@arm.com>
8048 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8049 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8050 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8051 (vect_analyze_slp_cost): Likewise.
8052 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8053 (vect_model_load_cost): Likewise.
8055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8056 Alan Hayward <alan.hayward@arm.com>
8057 David Sherwood <david.sherwood@arm.com>
8059 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8060 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8061 from an unsigned int * to a poly_uint64_pod *.
8062 (calculate_unrolling_factor): New function.
8063 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8066 Alan Hayward <alan.hayward@arm.com>
8067 David Sherwood <david.sherwood@arm.com>
8069 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8070 from an unsigned int to a poly_uint64.
8071 (_loop_vec_info::slp_unrolling_factor): Likewise.
8072 (_loop_vec_info::vectorization_factor): Change from an int
8074 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8075 (vect_get_num_vectors): New function.
8076 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8077 (vect_get_num_copies): Use vect_get_num_vectors.
8078 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8079 to an unsigned int *.
8080 (vect_analyze_data_refs): Change min_vf from an int * to a
8082 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8083 than an unsigned HOST_WIDE_INT.
8084 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8085 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8086 to an unsigned int *.
8087 (vect_analyze_data_ref_dependences): Likewise.
8088 (vect_compute_data_ref_alignment): Handle polynomial vf.
8089 (vect_enhance_data_refs_alignment): Likewise.
8090 (vect_prune_runtime_alias_test_list): Likewise.
8091 (vect_shift_permute_load_chain): Likewise.
8092 (vect_supportable_dr_alignment): Likewise.
8093 (dependence_distance_ge_vf): Take the vectorization factor as a
8094 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8095 (vect_analyze_data_refs): Change min_vf from an int * to a
8097 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8098 vfm1 as a poly_uint64 rather than an int. Make the same change
8099 for the returned bound_scalar.
8100 (vect_gen_vector_loop_niters): Handle polynomial vf.
8101 (vect_do_peeling): Likewise. Update call to
8102 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8103 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8105 * tree-vect-loop.c (vect_determine_vectorization_factor)
8106 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8107 (vect_get_known_peeling_cost): Likewise.
8108 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8109 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8110 (vect_transform_loop): Likewise. Use the lowest possible VF when
8111 updating the upper bounds of the loop.
8112 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8114 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8115 polynomial unroll factors.
8116 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8117 (vect_make_slp_decision): Likewise.
8118 (vect_supported_load_permutation_p): Likewise, and polynomial
8120 (vect_analyze_slp_cost): Handle polynomial vf.
8121 (vect_slp_analyze_node_operations): Likewise.
8122 (vect_slp_analyze_bb_1): Likewise.
8123 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8124 than an unsigned HOST_WIDE_INT.
8125 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8126 (vectorizable_load): Handle polynomial vf.
8127 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8129 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8132 Alan Hayward <alan.hayward@arm.com>
8133 David Sherwood <david.sherwood@arm.com>
8135 * match.pd: Handle bit operations involving three constants
8136 and try to fold one pair.
8138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8140 * tree-vect-loop-manip.c: Include gimple-fold.h.
8141 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8142 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8143 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8144 Add a path that uses a step of VF instead of 1, but disable it
8146 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8147 and niters_no_overflow parameters. Update calls to
8148 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8149 Create a new SSA name if the latter choses to use a ste other
8150 than zero, and return it via niters_vector_mult_vf_var.
8151 * tree-vect-loop.c (vect_transform_loop): Update calls to
8152 vect_do_peeling, vect_gen_vector_loop_niters and
8153 slpeel_make_loop_iterate_ntimes.
8154 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8155 (vect_gen_vector_loop_niters): Update declarations after above changes.
8157 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8159 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8160 128-bit round to integer instructions.
8161 (ceil<mode>2): Likewise.
8162 (btrunc<mode>2): Likewise.
8163 (round<mode>2): Likewise.
8165 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8167 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8168 unaligned VSX load/store on P8/P9.
8169 (expand_block_clear): Allow the use of unaligned VSX
8170 load/store on P8/P9.
8172 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8174 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8176 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8177 swap associated with both a load and a store.
8179 2018-01-02 Andrew Waterman <andrew@sifive.com>
8181 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8182 * config/riscv/riscv.md (clear_cache): Use it.
8184 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8186 * web.c: Remove out-of-date comment.
8188 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8190 * expr.c (fixup_args_size_notes): Check that any existing
8191 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8192 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8193 (emit_single_push_insn): ...here.
8195 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8197 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8198 (const_vector_encoded_nelts): New function.
8199 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8200 (const_vector_int_elt, const_vector_elt): Declare.
8201 * emit-rtl.c (const_vector_int_elt_1): New function.
8202 (const_vector_elt): Likewise.
8203 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8204 of CONST_VECTOR_ELT.
8206 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8208 * expr.c: Include rtx-vector-builder.h.
8209 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8210 directly on the tree encoding.
8211 (const_vector_from_tree): Likewise.
8212 * optabs.c: Include rtx-vector-builder.h.
8213 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8214 sequence of "u" values.
8215 * vec-perm-indices.c: Include rtx-vector-builder.h.
8216 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8217 directly on the vec_perm_indices encoding.
8219 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8221 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8222 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8223 * rtx-vector-builder.h: New file.
8224 * rtx-vector-builder.c: Likewise.
8225 * rtl.h (rtx_def::u2): Add a const_vector field.
8226 (CONST_VECTOR_NPATTERNS): New macro.
8227 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8228 (CONST_VECTOR_DUPLICATE_P): Likewise.
8229 (CONST_VECTOR_STEPPED_P): Likewise.
8230 (CONST_VECTOR_ENCODED_ELT): Likewise.
8231 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8232 (unwrap_const_vec_duplicate): Likewise.
8233 (const_vec_series_p): Check for a non-duplicated vector encoding.
8234 Say that the function only returns true for integer vectors.
8235 * emit-rtl.c: Include rtx-vector-builder.h.
8236 (gen_const_vec_duplicate_1): Delete.
8237 (gen_const_vector): Call gen_const_vec_duplicate instead of
8238 gen_const_vec_duplicate_1.
8239 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8240 (gen_const_vec_duplicate): Use rtx_vector_builder.
8241 (gen_const_vec_series): Likewise.
8242 (gen_rtx_CONST_VECTOR): Likewise.
8243 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8244 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8245 Build a new vector rather than modifying a CONST_VECTOR in-place.
8246 (handle_special_swappables): Update call accordingly.
8247 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8248 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8249 Build a new vector rather than modifying a CONST_VECTOR in-place.
8250 (handle_special_swappables): Update call accordingly.
8252 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8254 * simplify-rtx.c (simplify_const_binary_operation): Use
8255 CONST_VECTOR_ELT instead of XVECEXP.
8257 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8259 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8260 the selector elements to be different from the data elements
8261 if the selector is a VECTOR_CST.
8262 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8263 ssizetype for the selector.
8265 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8267 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8268 before testing each element individually.
8269 * tree-vect-generic.c (lower_vec_perm): Likewise.
8271 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8273 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8274 * selftest-run-tests.c (selftest::run_tests): Call it.
8275 * vector-builder.h (vector_builder::operator ==): New function.
8276 (vector_builder::operator !=): Likewise.
8277 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8278 (vec_perm_indices::all_from_input_p): New function.
8279 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8280 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8281 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8282 instead of reading the VECTOR_CST directly. Detect whether both
8283 vector inputs are the same before constructing the vec_perm_indices,
8284 and update the number of inputs argument accordingly. Use the
8285 utility functions added above. Only construct sel2 if we need to.
8287 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8289 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8290 the broadcast of the low byte.
8291 (expand_mult_highpart): Use an explicit encoding for the permutes.
8292 * optabs-query.c (can_mult_highpart_p): Likewise.
8293 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8294 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8295 (vectorizable_bswap): Likewise.
8296 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8297 explicit encoding for the power-of-2 permutes.
8298 (vect_permute_store_chain): Likewise.
8299 (vect_grouped_load_supported): Likewise.
8300 (vect_permute_load_chain): Likewise.
8302 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8304 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8305 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8306 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8307 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8308 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8309 (vect_gen_perm_mask_any): Likewise.
8311 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8313 * int-vector-builder.h: New file.
8314 * vec-perm-indices.h: Include int-vector-builder.h.
8315 (vec_perm_indices): Redefine as an int_vector_builder.
8316 (auto_vec_perm_indices): Delete.
8317 (vec_perm_builder): Redefine as a stand-alone class.
8318 (vec_perm_indices::vec_perm_indices): New function.
8319 (vec_perm_indices::clamp): Likewise.
8320 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8321 (vec_perm_indices::new_vector): New function.
8322 (vec_perm_indices::new_expanded_vector): Update for new
8323 vec_perm_indices class.
8324 (vec_perm_indices::rotate_inputs): New function.
8325 (vec_perm_indices::all_in_range_p): Operate directly on the
8326 encoded form, without computing elided elements.
8327 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8328 encoding. Update for new vec_perm_indices class.
8329 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8330 the given vec_perm_builder.
8331 (expand_vec_perm_var): Update vec_perm_builder constructor.
8332 (expand_mult_highpart): Use vec_perm_builder instead of
8333 auto_vec_perm_indices.
8334 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8335 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8336 or double series encoding as appropriate.
8337 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8338 vec_perm_indices instead of auto_vec_perm_indices.
8339 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8340 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8341 (vect_permute_store_chain): Likewise.
8342 (vect_grouped_load_supported): Likewise.
8343 (vect_permute_load_chain): Likewise.
8344 (vect_shift_permute_load_chain): Likewise.
8345 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8346 (vect_transform_slp_perm_load): Likewise.
8347 (vect_schedule_slp_instance): Likewise.
8348 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8349 (vectorizable_mask_load_store): Likewise.
8350 (vectorizable_bswap): Likewise.
8351 (vectorizable_store): Likewise.
8352 (vectorizable_load): Likewise.
8353 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8354 vec_perm_indices instead of auto_vec_perm_indices. Use
8355 tree_to_vec_perm_builder to read the vector from a tree.
8356 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8357 vec_perm_builder instead of a vec_perm_indices.
8358 (have_whole_vector_shift): Use vec_perm_builder and
8359 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8360 truncation to calc_vec_perm_mask_for_shift.
8361 (vect_create_epilog_for_reduction): Likewise.
8362 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8363 from auto_vec_perm_indices to vec_perm_indices.
8364 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8365 instead of changing individual elements.
8366 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8367 the vector in d.perm.
8368 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8369 from auto_vec_perm_indices to vec_perm_indices.
8370 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8371 instead of changing individual elements.
8372 (arm_vectorize_vec_perm_const): Use new_vector to install
8373 the vector in d.perm.
8374 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8375 Update vec_perm_builder constructor.
8376 (rs6000_expand_interleave): Likewise.
8377 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8378 (rs6000_expand_interleave): Likewise.
8380 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8382 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8383 to qimode could truncate the indices.
8384 * optabs.c (expand_vec_perm_var): Likewise.
8386 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8388 * Makefile.in (OBJS): Add vec-perm-indices.o.
8389 * vec-perm-indices.h: New file.
8390 * vec-perm-indices.c: Likewise.
8391 * target.h (vec_perm_indices): Replace with a forward class
8393 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8394 * optabs.h: Include vec-perm-indices.h.
8395 (expand_vec_perm): Delete.
8396 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8397 (expand_vec_perm_const): Declare.
8398 * target.def (vec_perm_const_ok): Replace with...
8399 (vec_perm_const): ...this new hook.
8400 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8401 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8402 * doc/tm.texi: Regenerate.
8403 * optabs.def (vec_perm_const): Delete.
8404 * doc/md.texi (vec_perm_const): Likewise.
8405 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8406 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8407 expand_vec_perm for constant permutation vectors. Assert that
8408 the mode of variable permutation vectors is the integer equivalent
8409 of the mode that is being permuted.
8410 * optabs-query.h (selector_fits_mode_p): Declare.
8411 * optabs-query.c: Include vec-perm-indices.h.
8412 (selector_fits_mode_p): New function.
8413 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8414 is defined, instead of checking whether the vec_perm_const_optab
8415 exists. Use targetm.vectorize.vec_perm_const instead of
8416 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8417 fit in the vector mode before using a variable permute.
8418 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8419 vec_perm_indices instead of an rtx.
8420 (expand_vec_perm): Replace with...
8421 (expand_vec_perm_const): ...this new function. Take the selector
8422 as a vec_perm_indices rather than an rtx. Also take the mode of
8423 the selector. Update call to shift_amt_for_vec_perm_mask.
8424 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8425 Use vec_perm_indices::new_expanded_vector to expand the original
8426 selector into bytes. Check whether the indices fit in the vector
8427 mode before using a variable permute.
8428 (expand_vec_perm_var): Make global.
8429 (expand_mult_highpart): Use expand_vec_perm_const.
8430 * fold-const.c: Includes vec-perm-indices.h.
8431 * tree-ssa-forwprop.c: Likewise.
8432 * tree-vect-data-refs.c: Likewise.
8433 * tree-vect-generic.c: Likewise.
8434 * tree-vect-loop.c: Likewise.
8435 * tree-vect-slp.c: Likewise.
8436 * tree-vect-stmts.c: Likewise.
8437 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8439 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8440 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8441 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8442 (aarch64_vectorize_vec_perm_const): ...this new function.
8443 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8444 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8445 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8446 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8447 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8448 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8449 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8451 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8452 check for NEON modes.
8453 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8454 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8455 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8456 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8458 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8459 the old VEC_PERM_CONST conditions.
8460 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8461 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8462 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8463 (ia64_vectorize_vec_perm_const_ok): Merge into...
8464 (ia64_vectorize_vec_perm_const): ...this new function.
8465 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8466 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8467 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8468 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8469 * config/mips/mips.c (mips_expand_vec_perm_const)
8470 (mips_vectorize_vec_perm_const_ok): Merge into...
8471 (mips_vectorize_vec_perm_const): ...this new function.
8472 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8473 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8474 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8475 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8476 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8477 (rs6000_expand_vec_perm_const): Delete.
8478 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8480 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8481 (altivec_expand_vec_perm_const_le): Take each operand individually.
8482 Operate on constant selectors rather than rtxes.
8483 (altivec_expand_vec_perm_const): Likewise. Update call to
8484 altivec_expand_vec_perm_const_le.
8485 (rs6000_expand_vec_perm_const): Delete.
8486 (rs6000_vectorize_vec_perm_const_ok): Delete.
8487 (rs6000_vectorize_vec_perm_const): New function.
8488 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8489 an element count and rtx array.
8490 (rs6000_expand_extract_even): Update call accordingly.
8491 (rs6000_expand_interleave): Likewise.
8492 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8493 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8494 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8495 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8496 (rs6000_expand_vec_perm_const): Delete.
8497 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8498 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8499 (altivec_expand_vec_perm_const_le): Take each operand individually.
8500 Operate on constant selectors rather than rtxes.
8501 (altivec_expand_vec_perm_const): Likewise. Update call to
8502 altivec_expand_vec_perm_const_le.
8503 (rs6000_expand_vec_perm_const): Delete.
8504 (rs6000_vectorize_vec_perm_const_ok): Delete.
8505 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8506 reference to the SPE evmerge intructions.
8507 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8508 an element count and rtx array.
8509 (rs6000_expand_extract_even): Update call accordingly.
8510 (rs6000_expand_interleave): Likewise.
8511 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8512 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8514 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8516 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8518 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8519 vector mode and that that mode matches the mode of the data
8521 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8522 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8523 directly using expand_vec_perm_1 when forcing selectors into
8525 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8527 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8529 * optabs-query.h (can_vec_perm_p): Delete.
8530 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8531 * optabs-query.c (can_vec_perm_p): Split into...
8532 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8533 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8534 particular selector is valid.
8535 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8536 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8537 (vect_grouped_load_supported): Likewise.
8538 (vect_shift_permute_load_chain): Likewise.
8539 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8540 (vect_transform_slp_perm_load): Likewise.
8541 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8542 (vectorizable_bswap): Likewise.
8543 (vect_gen_perm_mask_checked): Likewise.
8544 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8545 implementations of variable permutation vectors into account
8546 when deciding which selector to use.
8547 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8548 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8549 with a false third argument.
8550 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8551 to test whether the constant selector is valid and can_vec_perm_var_p
8552 to test whether a variable selector is valid.
8554 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8556 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8557 * optabs-query.c (can_vec_perm_p): Likewise.
8558 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8559 instead of vec_perm_indices.
8560 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8561 (vect_gen_perm_mask_checked): Likewise,
8562 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8563 (vect_gen_perm_mask_checked): Likewise,
8565 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8567 * optabs-query.h (qimode_for_vec_perm): Declare.
8568 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8569 (qimode_for_vec_perm): ...this new function.
8570 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8572 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8574 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8575 does not have a conditional at the top.
8577 2018-01-02 Richard Biener <rguenther@suse.de>
8579 * ipa-inline.c (big_speedup_p): Fix expression.
8581 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8584 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8587 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8591 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8592 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8593 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8594 cond_taken_branch_cost 3->4.
8596 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8598 PR tree-optimization/83581
8599 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8600 TODO_cleanup_cfg if any changes have been made.
8603 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8604 convert_modes if target mode has the right side, but different mode
8608 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8609 last argument when extracting from CONCAT. If either from_real or
8610 from_imag is NULL, use expansion through memory. If result is not
8611 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8612 the parts directly to inner mode, if even that fails, use expansion
8616 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8617 check for bswap in mode rather than HImode and use that in expand_unop
8620 Copyright (C) 2018 Free Software Foundation, Inc.
8622 Copying and distribution of this file, with or without modification,
8623 are permitted in any medium without royalty provided the copyright
8624 notice and this notice are preserved.