1 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4 parameter. Use it for SFmode.
5 (rs6000_function_arg_advance_1): Adjust.
6 (rs6000_function_arg): Adjust.
7 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11 PR rtl-optimization/84169
12 PR rtl-optimization/84780
13 * combine.c (can_combine_p): Check for a 2-insn combination whether
14 the destination register is used between the two insns, too.
16 2018-03-12 Richard Biener <rguenther@suse.de>
18 PR tree-optimization/84803
19 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
20 for refs DR analysis didn't process.
22 2018-03-12 Richard Biener <rguenther@suse.de>
24 PR tree-optimization/84777
25 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
26 force-vectorize loops ignore whether we are optimizing for size.
28 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
30 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
31 (TARGET_MD_ASM_ADJUST): Define.
33 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
34 Kito Cheng <kito.cheng@gmail.com>
35 Chung-Ju Wu <jasonwucj@gmail.com>
37 * config/nds32/nds32.c (nds32_compute_stack_frame,
38 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
39 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
40 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
41 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
42 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
43 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
44 * config/nds32/nds32.md (prologue, epilogue): Use macro
45 NDS32_V3PUSH_AVAILABLE_P to do checking.
47 2018-03-11 Jakub Jelinek <jakub@redhat.com>
50 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
51 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
52 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
53 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
54 addition of most attributes on !orig_type_die or the attribute not
55 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
57 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
58 Chung-Ju Wu <jasonwucj@gmail.com>
60 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
62 * config/nds32/nds32.opt (mvh): New option.
64 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
65 Chung-Ju Wu <jasonwucj@gmail.com>
67 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
69 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
70 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
73 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
74 Chung-Ju Wu <jasonwucj@gmail.com>
76 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
78 * config/nds32/nds32-multiple.md (strlensi): New pattern.
79 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
81 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
82 Kito Cheng <kito.cheng@gmail.com>
83 Chung-Ju Wu <jasonwucj@gmail.com>
85 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
86 UNSPEC_FFMISM and UNSPEC_FLMISM.
87 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
88 for ffb, ffmism and flmism.
89 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
90 (unspec_ffmism): Ditto.
91 (unspec_flmism): Ditto.
92 (nds32_expand_builtin_impl): Check if string extension is available.
93 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
94 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
96 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
99 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
102 * lra-assigns.c (assign_by_spills): Return a flag of reload
103 assignment failure. Do not process the reload assignment
104 failures. Do not spill other reload pseudos if they has the same
106 (lra_assign): Add a return arg. Set up from the result of
107 assign_by_spills call.
108 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
109 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
110 usage_insns if it is not NULL.
111 (spill_hard_reg_in_range): New function.
112 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
113 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
115 (lra_assign): Change prototype.
116 * lra.c (lra): Add code to deal with fails by splitting hard reg
119 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
122 * config/i386/i386.opt: Replace Enforcment with Enforcement.
124 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
127 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
128 (dw_val_node): Add val_symbolic_view.
129 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
130 (symview_upper_bound): New.
131 (new_line_info_table): Initialize symviews_since_reset.
132 (dwarf2out_source_line): Count symviews_since_reset and set
134 (dw_val_equal_p): Handle symview.
135 (add_AT_symview): New.
136 (print_dw_val): Handle symview.
137 (attr_checksum, attr_checksum_ordered): Likewise.
138 (same_dw_val_p, size_of_die): Likewise.
139 (value_format, output_die): Likewise.
140 (add_high_low_attributes): Use add_AT_symview for entry_view.
141 (dwarf2out_finish): Reset symview_upper_bound, clear
144 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
147 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
148 Add strict argument and use it.
149 (rs6000_split_multireg_move): Update for new strict argument.
150 (mem_operand_gpr): Disallow all non-offsettable addresses.
151 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
153 2018-03-09 Jakub Jelinek <jakub@redhat.com>
156 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
157 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
158 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
161 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
162 decl, use remap_type if we want to use the type.
164 2018-03-09 Martin Sebor <msebor@redhat.com>
166 PR tree-optimization/84526
167 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
169 (builtin_access::generic_overlap): Be prepared to handle non-array
172 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
174 PR rtl-optimization/84682
175 * lra-constraints.c (process_address_1): Check is_address flag
176 for address constraints.
177 (process_alt_operands): Likewise.
178 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
179 preprocess_constraints.
180 * recog.h (preprocess_constraints): Add oploc parameter.
182 * recog.c (preprocess_constraints): Test address_operand for
183 CT_ADDRESS constraints.
185 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
188 * lra-assigns.c (assign_by_spills): Return a flag of reload
189 assignment failure. Do not process the reload assignment
190 failures. Do not spill other reload pseudos if they has the same
192 (lra_assign): Add a return arg. Set up from the result of
193 assign_by_spills call.
194 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
195 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
196 usage_insns if it is not NULL.
197 (spill_hard_reg_in_range): New function.
198 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
199 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
201 (lra_assign): Change prototype.
202 * lra.c (lra): Add code to deal with fails by splitting hard reg
205 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
208 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
209 Accept complain bool parameter. Only emit errors if it is true.
210 (arm_parse_cpu_option_name): Likewise.
211 (arm_target_thumb_only): Adjust callers of the above.
212 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
213 prototype to take a default true bool parameter.
214 (arm_parse_arch_option_name): Likewise.
216 2018-03-09 David Malcolm <dmalcolm@redhat.com>
217 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
221 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
222 * configure: Regenerate.
223 * configure.ac ("linker --version-script option"): New.
224 ("linker soname option"): New.
226 2018-03-09 Richard Biener <rguenther@suse.de>
228 PR tree-optimization/84775
229 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
230 immediate uses of predicate stmts and mark them modified.
233 PR tree-optimization/84178
234 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
236 (version_loop_for_if_conversion): Delay update_ssa call.
237 (tree_if_conversion): Delay update_ssa until after predicate
240 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
243 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
244 when the function accesses prior frames.
246 2018-03-08 Jakub Jelinek <jakub@redhat.com>
249 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
250 gen_llsym, otherwise call maybe_gen_llsym.
253 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
254 has ',' character inside of it.
256 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
259 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
260 as clobbering CC_REGNUM.
262 2018-03-08 Richard Biener <rguenther@suse.de>
265 * tree-scalar-evolution.c: Include tree-into-ssa.h.
266 (follow_copies_to_constant): Do not follow SSA names registered
269 2018-03-08 Richard Biener <rguenther@suse.de>
271 PR tree-optimization/84178
272 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
274 (version_loop_for_if_conversion): Delay update_ssa call.
275 (tree_if_conversion): Delay update_ssa until after predicate
278 2018-03-08 David Malcolm <dmalcolm@redhat.com>
280 PR tree-optimization/84178
281 * tree-if-conv.c (release_bb_predicate): Remove the
282 the assertion that the stmts have NULL use_ops.
283 Discard the statements, asserting that they haven't
284 yet been added to a BB.
286 2018-03-08 Richard Biener <rguenther@suse.de>
288 PR tree-optimization/84746
289 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
290 (phi_translate): Pass in destination ANTIC_OUT set.
291 (phi_translate_1): Likewise. For a simplified result lookup
292 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
293 (phi_translate_set): Adjust.
294 (do_pre_regular_insertion): Likewise.
295 (do_pre_partial_partial_insertion): Likewise.
297 2018-03-08 Martin Liska <mliska@suse.cz>
299 PR gcov-profile/84735
300 * doc/gcov.texi: Document usage of profile files.
301 * gcov-io.h: Document changes in the format.
303 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
307 * dwarf2out.c (struct dw_line_info_table): Update comments for
309 (FORCE_RESET_NEXT_VIEW): New.
310 (FORCE_RESETTING_VIEW_P): New.
311 (RESETTING_VIEW_P): Check for -1 too.
312 (ZERO_VIEW_P): Likewise.
313 (new_line_info_table): Force-reset next view.
314 (dwarf2out_begin_function): Likewise.
315 (dwarf2out_source_line): Simplify zero_view_p initialization.
316 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
317 view directly. Omit view when omitting .loc at line 0.
319 2018-03-08 Jakub Jelinek <jakub@redhat.com>
321 PR tree-optimization/84740
322 * tree-switch-conversion.c (process_switch): Call build_constructors
323 only if info.phi_count is non-zero.
325 PR tree-optimization/84739
326 * tree-tailcall.c (find_tail_calls): Check call arguments against
327 DECL_ARGUMENTS (current_function_decl) rather than
328 DECL_ARGUMENTS (func) when checking for tail recursion.
330 2018-03-07 Jakub Jelinek <jakub@redhat.com>
332 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
333 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
334 Volker Reichelt's entry and add entries for people that perform
335 GCC fuzzy testing and report numerous bugs.
337 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
340 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
341 readonly data in sdata, if that is disabled.
342 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
343 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
344 -mreadonly-in-sdata option.
346 2018-03-07 Martin Sebor <msebor@redhat.com>
348 PR tree-optimization/84468
349 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
350 basic block when looking for nul assignment.
352 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
355 * except.h (output_function_exception_table): Adjust prototype.
356 * except.c (output_function_exception_table): Remove FNNAME parameter
357 and add SECTION parameter. Ouput one part of the table at a time.
358 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
359 the first part of the exception table and emit unwind directives.
360 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
361 (i386_pe_seh_cold_init): Likewise.
362 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
363 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
364 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
365 (ix86_output_call_insn): Emit a nop in one more case for SEH.
366 * config/i386/winnt.c: Include except.h.
367 (struct seh_frame_state): Add reg_offset, after_prologue and
368 in_cold_section fields.
369 (i386_pe_seh_end_prologue): Set seh->after_prologue.
370 (i386_pe_seh_cold_init): New function.
371 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
372 to seh->in_cold_section.
373 (seh_emit_push): Record the offset of the push.
374 (seh_emit_save): Record the offet of the save.
375 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
376 Test seh->after_prologue to disregard the epilogue.
377 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
378 (i386_pe_end_cold_function): New function.
380 2018-03-07 Jakub Jelinek <jakub@redhat.com>
383 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
384 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
387 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
389 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
390 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
393 * multiple_target.c: Include tree-inline.h and intl.h.
394 (expand_target_clones): Diagnose and fail if node->definition and
395 !tree_versionable_function_p (node->decl).
397 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
399 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
401 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
402 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
403 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
405 2018-03-06 Jakub Jelinek <jakub@redhat.com>
408 * combine.c (try_combine): Use reg_or_subregno instead of handling
409 just paradoxical SUBREGs and REGs.
411 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
413 * config/arc/arc.c (arc_finalize_pic): Remove function.
414 (arc_must_save_register): We use single base PIC register, remove
415 checks to save/restore the PIC register.
416 (arc_expand_prologue): Likewise.
417 * config/arc/arc-protos.h (arc_set_default_type_attributes):
419 (arc_verify_short): Likewise.
420 (arc_attr_type): Likewise.
421 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
422 (walk_stores): Likewise.
423 (arc_address_cost): Make it static.
424 (arc_verify_short): Likewise.
425 (branch_dest): Likewise.
426 (arc_attr_type): Likewise.
427 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
428 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
429 (arc_final_prescan_insn): Remove inserting the nops due to
430 hardware hazards. It is done in reorg step.
431 (insn_length_variant_t): Remove.
432 (insn_length_parameters_t): Likewise.
433 (arc_insn_length_parameters): Likewise.
434 (arc_get_insn_variants): Likewise.
435 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
437 2018-03-06 Jakub Jelinek <jakub@redhat.com>
440 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
443 PR tree-optimization/84687
444 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
446 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
448 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
450 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
451 Rename to ppc_speculation_barrier.
452 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
453 __builtin_ppc_speculation_barrier.
455 2018-03-05 Jakub Jelinek <jakub@redhat.com>
458 * combine.c (combine_simplify_rtx): Don't try to simplify if
459 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
462 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
464 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
465 to 32 bytes when compiling for POWER9.
467 2018-03-05 Jakub Jelinek <jakub@redhat.com>
470 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
471 regparm >= 3 with no arg reg available also for calls with
472 flag_force_indirect_call. Pass decl to ix86_function_regparm.
475 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
477 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
479 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
482 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
484 2018-03-05 Richard Biener <rguenther@suse.de>
486 PR tree-optimization/84486
487 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
488 When inserting a __builtin_assume_aligned call set the LHS
489 SSA name alignment info accordingly.
491 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
493 PR tree-optimization/84114
494 * config/aarch64/aarch64.c (aarch64_reassociation_width)
495 Avoid reassociation of FLOAT_MODE addition.
497 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
499 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
500 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
501 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
502 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
503 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
504 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
505 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
507 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
508 __builtin_ia32_wbinvd): New builtins.
509 (SPECIAL_ARGS2): New.
510 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
511 (SPECIAL_ARGS2): New.
512 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
513 (ix86_valid_target_attribute_inner_p): Ditto.
514 (ix86_init_mmx_sse_builtins): Add special_args2.
515 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
516 TARGET_WBNOINVD_P): New.
517 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
518 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
519 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
520 * config/i386/immintrin.h (_wbinvd): New intrinsic.
521 * config/i386/pconfigintrin.h: New file.
522 * config/i386/wbnoinvdintrin.h: Ditto.
523 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
524 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
526 2018-03-05 Richard Biener <rguenther@suse.de>
528 PR tree-optimization/84670
529 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
531 (BB_VISITED_WITH_VISITED_SUCCS): New define.
532 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
533 (compute_antic_aux): Only assert the number of values in ANTIC_IN
534 doesn't grow if all successors (recursively) were visited at least
537 2018-03-05 Richard Biener <rguenther@suse.de>
539 PR tree-optimization/84650
540 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
541 if executed in the loop pipeline.
543 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
545 * doc/configfiles.texi (Configuration Files): Move info about
546 conditionalizing $target-protos.h to...
547 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
548 differs from $target-protos.h.
550 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
551 Chung-Ju Wu <jasonwucj@gmail.com>
553 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
554 * config/nds32/nds32-multiple.md (setmemsi): Define.
555 * config/nds32/nds32-memory-manipulation.c
556 (nds32_gen_dup_4_byte_to_word_value): New.
557 (emit_setmem_word_loop): New.
558 (emit_setmem_byte_loop): New.
559 (nds32_expand_setmem_loop): New.
560 (nds32_expand_setmem_loop_v3m): New.
561 (nds32_expand_setmem_unroll): New.
562 (nds32_expand_setmem): New.
564 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
565 Chung-Ju Wu <jasonwucj@gmail.com>
567 * config/nds32/nds32-memory-manipulation.c
568 (nds32_emit_load_store): New.
569 (nds32_emit_post_inc_load_store): New.
570 (nds32_emit_mem_move): New.
571 (nds32_emit_mem_move_block): New.
572 (nds32_expand_movmemsi_loop_unknown_size): New.
573 (nds32_expand_movmemsi_loop_known_size): New.
574 (nds32_expand_movmemsi_loop): New.
575 (nds32_expand_movmemsi_unroll): New.
576 (nds32_expand_movmemqi): Rename ...
577 (nds32_expand_movmemsi): ... to this.
578 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
579 (movmemsi): ... to this.
580 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
581 (nds32_expand_movmemsi): ... to this.
583 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
584 Monk Chiang <sh.chiang04@gmail.com>
585 Chung-Ju Wu <jasonwucj@gmail.com>
587 * config/nds32/nds32-protos.h
588 (nds32_expand_load_multiple): New arguments.
589 (nds32_expand_store_multiple): Ditto.
590 (nds32_valid_multiple_load_store): Rename ...
591 (nds32_valid_multiple_load_store_p): ... to this.
592 * config/nds32/nds32-memory-manipulation.c
593 (nds32_expand_load_multiple): Refine implementation.
594 (nds32_expand_store_multiple): Ditto.
595 * config/nds32/nds32-multiple.md
596 (load_multiple): Update nds32_expand_load_multiple interface.
597 (store_multiple): Update nds32_expand_store_multiple interface.
598 * config/nds32/nds32-predicates.c
599 (nds32_valid_multiple_load_store): Rename ...
600 (nds32_valid_multiple_load_store_p): ... to this and refine
602 * config/nds32/predicates.md
603 (nds32_load_multiple_and_update_address_operation): New predicate.
604 (nds32_store_multiple_and_update_address_operation): New predicate.
606 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
607 Chung-Ju Wu <jasonwucj@gmail.com>
609 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
610 (combo): New attribute.
611 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
613 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
615 * config/nds32/nds32.opt: Change -mcmodel= default value.
617 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
618 Monk Chiang <sh.chiang04@gmail.com>
619 Chung-Ju Wu <jasonwucj@gmail.com>
621 * config/nds32/constants.md (unspec_element): New enum.
622 * config/nds32/constraints.md (Umw): New constraint.
623 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
624 * config/nds32/nds32-intrinsic.md: Likewise.
625 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
626 (nds32_valid_smw_lwm_base_p): New.
627 (nds32_output_smw_single_word): New.
628 (nds32_output_lmw_single_word): New.
629 (nds32_expand_unaligned_load): New.
630 (nds32_expand_unaligned_store): New.
631 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
632 (nds32_output_smw_single_word): Declare.
633 (nds32_output_lmw_single_word): Declare.
634 (nds32_expand_unaligned_load): Declare.
635 (nds32_expand_unaligned_store): Declare.
636 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
637 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
638 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
639 NDS32_BUILTIN_UASTORE_DW.
640 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
643 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
644 Kito Cheng <kito.cheng@gmail.com>
645 Chung-Ju Wu <jasonwucj@gmail.com>
647 * config/nds32/nds32-intrinsic.c
648 (nds32_expand_builtin_null_ftype_reg): Delete.
649 (nds32_expand_builtin_reg_ftype_imm): Ditto.
650 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
651 (nds32_read_argument): New.
652 (nds32_legitimize_target): Ditto.
653 (nds32_legitimize_argument): Ditto.
654 (nds32_check_constant_argument): Ditto.
655 (nds32_expand_unop_builtin): Ditto.
656 (nds32_expand_unopimm_builtin): Ditto.
657 (nds32_expand_binop_builtin): Ditto.
658 (nds32_builtin_decl_impl): Ditto.
659 (builtin_description): Ditto.
660 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
661 (nds32_init_builtins_impl): Ditto.
662 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
663 (nds32_builtin_decl): New.
664 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
665 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
667 2018-03-02 Jeff Law <law@redhat.com>
669 * reorg.c (stop_search_p): Handle DEBUG_INSN.
670 (redundant_insn, fill_simple_delay_slots): Likewise.
671 (fill_slots_from_thread): Likewise.
672 * resource.c (mark_referenced_resources): Likewise.
673 (mark_set_resources, find_dead_or_set_registers): Likewise.
675 2018-03-02 Jakub Jelinek <jakub@redhat.com>
677 * substring-locations.h (format_warning_va): Formatting fix for
679 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
681 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
682 * substring-locations.c: Include intl.h.
683 (format_warning_va): Turned into small wrapper around
684 format_warning_n_va, renamed to ...
685 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
686 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
688 (format_warning_at_substring_n): New function.
689 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
690 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
691 format_warning_at_substring with just a shorter name instead of
692 const function pointer.
693 (fmtwarn_n): New function.
694 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
695 appropriate, get rid of all the fmtstr temporaries, move conditionals
696 with G_() wrapped string literals directly into fmtwarn arguments,
697 cast dir.len to (int), formatting fixes.
699 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
701 * doc/invoke.texi: Remove "Cilk Plus" references.
703 2018-03-02 Jakub Jelinek <jakub@redhat.com>
704 Richard Biener <rguenther@suse.de>
707 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
708 for error or warning attributes if CALL_FROM_THUNK_P is set.
711 2018-03-02 Jakub Jelinek <jakub@redhat.com>
714 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
715 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
718 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
719 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
721 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
722 instead of -1U in last predictors element's probability member.
724 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
727 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
728 arguments if they are comparable.
730 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
732 PR tree-optimization/84634
733 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
734 masks and masked_loop_p with a single loop_masks, making sure it's
735 null for bb vectorization.
737 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
739 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
740 (vect_analyze_data_ref_access): Use loop->safe_len rather than
741 loop->force_vectorize to check whether there is no alias.
743 2018-03-02 Jakub Jelinek <jakub@redhat.com>
746 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
748 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
750 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
751 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
752 instead of a loop around prev_real_insn.
753 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
757 * config/i386/i386.c (ix86_print_operand): Use conditional
758 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
761 2018-03-02 Richard Biener <rguenther@suse.de>
763 PR tree-optimization/84427
764 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
765 (bitmap_set_subtract_values): Rewrite to handle multiple
768 (prune_clobbered_mems): Likewise.
769 (phi_translate): Take edge instead of pred/phiblock.
770 (phi_translate_1): Likewise.
771 (phi_translate_set): Likewise. Insert all translated
772 exprs for a value into the set, keeping possibly multiple
773 expressions per value.
774 (compute_antic_aux): Adjust for phi_translate changes.
775 When intersecting union the expressions and prune those
776 not in the final value set, keeping possibly multiple
777 expressions per value. Do not use value-insertion
778 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
779 all expressions. Add verification that the value-sets
780 only shrink during iteration.
781 (compute_partial_antic_aux): Adjust for the phi_translate changes.
782 (do_pre_regular_insertion): Likewise.
783 (do_pre_partial_partial_insertion): Likewise.
785 2018-03-02 Richard Biener <rguenther@suse.de>
788 * config/darwin.c (saved_debug_info_level): New static global.
789 (darwin_asm_lto_start): Disable debug info generation for LTO out.
790 (darwin_asm_lto_end): Restore debug info generation settings.
792 2018-03-01 Martin Liska <mliska@suse.cz>
795 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
798 2018-03-01 Richard Biener <rguenther@suse.de>
801 * dwarf2out.c (gen_variable_die): Properly handle late VLA
802 type annotation with LTO when debug was disabled at compile-time.
804 2018-03-01 Matthew Fortune <mfortune@gmail.com>
806 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
808 (mips_final_postscan_insn): Likewise.
810 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
812 PR rtl-optimization/84528
813 * alias.c (init_alias_target): Add commentary.
814 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
815 a unique base value if the frame pointer is not eliminated
816 to the stack pointer.
818 2018-03-01 Tom de Vries <tom@codesourcery.com>
820 PR rtl-optimization/83327
821 * lra-int.h (hard_regs_spilled_into): Declare.
822 * lra.c (hard_regs_spilled_into): Define.
823 (init_reg_info): Init hard_regs_spilled_into.
824 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
825 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
826 (process_bb_lives): Handle hard_regs_spilled_into.
827 (lra_create_live_ranges_1): Before doing liveness propagation, clear
828 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
830 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
832 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
833 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
834 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
835 * config/rs6000/aix72.h: New file.
837 2018-02-28 Jakub Jelinek <jakub@redhat.com>
839 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
840 instead of warning_at with conditional singular and plural messages
844 * stor-layout.c (update_alignment_for_field): For
845 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
846 && !DECL_PACKED (field), do the alignment update, just use
847 only desired_align instead of MAX (type_align, desired_align)
849 (place_field): Don't do known_align < desired_align handling
850 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
851 is non-NULL, instead do it after rli->prev_field handling and
852 only if not within a bitfield word. For DECL_PACKED (field)
853 use type_align of BITS_PER_UNIT.
855 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
857 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
858 superfluous parentheses and trailing spaces.
860 2018-02-28 Richard Biener <rguenther@suse.de>
862 PR tree-optimization/84584
863 * graphite-scop-detection.c (scop_detection::add_scop): Discard
864 SCoPs with fake exit edge.
866 2018-02-28 Martin Liska <mliska@suse.cz>
869 * timevar.c (timer::print): Fix format to properly print 100%
872 2018-02-28 Richard Biener <rguenther@suse.de>
875 * genmatch.c (capture_info::walk_match): Do not mark
876 captured expressions without operands as expr_p given
877 they act more like predicates and should be subject to
878 "lost tail" side-effect preserving.
880 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
882 PR rtl-optimization/81611
883 * auto-inc-dec.c (attempt_change): Move dead note from
884 mem_insn if it's the next use of regno
885 (find_address): Take address use of reg holding
886 non-incremented value. Add parm to limit search to the named
888 (merge_in_block): Attempt to use a mem insn that is the next
889 use of the original regno.
891 2018-02-27 Martin Sebor <msebor@redhat.com>
894 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
895 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
897 2018-02-27 Martin Sebor <msebor@redhat.com>
900 * diagnostic-core.h (warning_n, error_n, inform_n): Change
901 n argument to unsigned HOST_WIDE_INT.
902 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
903 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
904 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
905 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
907 2018-02-27 Richard Biener <rguenther@suse.de>
909 PR tree-optimization/84512
910 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
911 Do not use the estimate returned from record_stmt_cost for
912 the scalar iteration cost but sum properly using add_stmt_cost.
914 2018-02-27 Richard Biener <rguenther@suse.de>
916 PR tree-optimization/84466
917 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
918 Adjust last change to less strictly validate use operands.
920 2018-02-27 Martin Liska <mliska@suse.cz>
922 PR gcov-profile/84548
923 * gcov.c (process_file): Allow partial overlap and consider it
924 also as group functions.
925 (output_lines): Properly calculate range of lines for a group.
927 2018-02-27 Martin Liska <mliska@suse.cz>
929 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
930 'ggc' suffixes. Change first column width.
931 (timer::print): Fix formatting of the column.
933 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
935 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
936 preserve inline entry blocks for the sake of debug inline
937 entry point markers alone.
938 (remove_unused_locals): Suggest in comments a better place to
939 force the preservation of inline entry blocks that are
940 otherwise unused, but do not preserve them.
942 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
944 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
946 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
949 * config/i386/constraints.md (Bs): Replace
950 ix86_indirect_branch_register with
951 TARGET_INDIRECT_BRANCH_REGISTER.
953 * config/i386/i386.md (indirect_jump): Likewise.
954 (tablejump): Likewise.
955 (*sibcall_memory): Likewise.
956 (*sibcall_value_memory): Likewise.
957 Peepholes of indirect call and jump via memory: Likewise.
958 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
959 (*sibcall_value_GOT_32): Likewise.
960 * config/i386/predicates.md (indirect_branch_operand): Likewise.
961 (GOT_memory_operand): Likewise.
962 (call_insn_operand): Likewise.
963 (sibcall_insn_operand): Likewise.
964 (GOT32_symbol_operand): Likewise.
965 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
967 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
969 PR rtl-optimization/83496
970 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
971 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
972 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
973 redundant insn, if any.
974 (relax_delay_slots): Likewise.
975 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
977 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
979 PR tree-optimization/83965
980 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
981 that grouped statements are part of a reduction chain. Return
982 true if the statement is not marked as a reduction itself but
984 (vect_recog_dot_prod_pattern): Don't check whether the statement
985 is part of a group here.
986 (vect_recog_sad_pattern): Likewise.
987 (vect_recog_widen_sum_pattern): Likewise.
989 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
992 * final.c (rest_of_clean_state): Also look for calls inside sequences.
994 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
997 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
999 (ix86_output_indirect_function_return): New prototype.
1000 (ix86_split_simple_return_pop_internal): Likewise.
1001 * config/i386/i386.c (indirect_return_via_cx): New.
1002 (indirect_return_via_cx_bnd): Likewise.
1003 (indirect_thunk_name): Handle return va CX_REG.
1004 (output_indirect_thunk_function): Create alias for
1005 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1006 (ix86_output_indirect_jmp): Remove the bool argument.
1007 (ix86_output_indirect_function_return): New function.
1008 (ix86_split_simple_return_pop_internal): Likewise.
1009 * config/i386/i386.md (*indirect_jump): Don't pass false
1010 to ix86_output_indirect_jmp.
1011 (*tablejump_1): Likewise.
1012 (simple_return_pop_internal): Change it to define_insn_and_split.
1013 Call ix86_split_simple_return_pop_internal to split it for
1015 (simple_return_indirect_internal): Call
1016 ix86_output_indirect_function_return instead of
1017 ix86_output_indirect_jmp.
1019 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1022 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1023 memset and value initialization afterwards.
1025 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1027 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1029 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1032 * common/config/aarch64/aarch64-common.c
1033 (aarch_option_optimization_table[]): Switch
1034 off fomit-frame-pointer
1036 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1037 Chung-Ju Wu <jasonwucj@gmail.com>
1039 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1041 (store_multiple): Ditto.
1043 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1045 * config.gcc: Add --with-cpu support for nds32 target.
1046 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1047 * config/nds32/nds32.opt: Add -mcpu= option.
1049 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1051 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1052 isel=yes): Warn for these deprecated options.
1054 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1056 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1057 ISA_2_5_MASKS_EMBEDDED.
1059 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1061 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1062 p->max as pointers rather than using iterative_hash_expr.
1064 2018-02-23 Carl Love <cel@us.ibm.com>
1066 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1067 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1069 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1070 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1071 P8V_BUILTIN_VEC_VUNSIGNED2.
1073 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1076 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1077 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1079 * lra-constraints.c (curr_insn_transform): Set up
1080 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1082 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1083 * lra-eliminations.c (spill_pseudos): Ditto.
1084 (process_insn_for_elimination): Ditto.
1085 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1086 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1087 LRA_NON_CLOBBERED_ALT.
1089 2018-02-22 Martin Sebor <msebor@redhat.com>
1091 PR tree-optimization/84480
1092 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1093 to maybe_diag_stxncpy_trunc. Call it.
1094 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1095 from gimple_fold_builtin_strcpy. Print inlining stack.
1096 (handle_builtin_stxncpy): Print inlining stack.
1097 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1099 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1102 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1103 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1104 and -fcheck-pointer-bounds are used together.
1105 (indirect_thunk_prefix): New enum.
1106 (indirect_thunk_need_prefix): New function.
1107 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1108 "_nt" instead of "_bnd" for NOTRACK prefix.
1109 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1110 (output_indirect_thunk_function): Likewise.
1112 (ix86_code_end): Update output_indirect_thunk_function calls.
1113 (ix86_output_indirect_branch_via_reg): Replace
1114 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1115 (ix86_output_indirect_branch_via_push): Likewise.
1116 (ix86_output_function_return): Likewise.
1117 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1118 incompatible with -fcf-protection=branch and
1119 -fcheck-pointer-bounds.
1121 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1124 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1125 Change gcc_assert call to output_operand_lossage.
1127 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1129 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1131 2018-02-22 DJ Delorie <dj@redhat.com>
1132 Sebastian Perta <sebastian.perta@renesas.com>
1133 Oleg Endo <olegendo@gcc.gnu.org>
1135 * config/rx/rx.c (rx_rtx_costs): New function.
1136 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1138 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1140 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1142 2018-02-22 Martin Liska <mliska@suse.cz>
1145 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1146 Add "native" as a possible value.
1148 2018-02-22 Martin Liska <mliska@suse.cz>
1151 * config/i386/i386.c (ix86_option_override_internal):
1152 Add "native" as a possible value for -march and -mtune.
1154 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1157 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1158 to all type variants.
1160 PR tree-optimization/84503
1161 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1162 width as info->bitpos + info->bitsize - start.
1163 (merged_store_group::merge_overlapping): Simplify width computation.
1164 (check_no_overlap): New function.
1165 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1166 start + width and last_order of the group, fail if check_no_overlap
1168 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1169 to group if check_no_overlap fails.
1171 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1173 * config/rs6000/altivec.md: Delete contraint arguments to
1174 define_expand, define_split, and define_peephole2, and in
1175 define_insn_and_split if always unused.
1176 * config/rs6000/darwin.md: Ditto.
1177 * config/rs6000/dfp.md: Ditto.
1178 * config/rs6000/rs6000.md: Ditto.
1179 * config/rs6000/sync.md: Ditto.
1180 * config/rs6000/vector.md: Ditto.
1181 * config/rs6000/vsx.md: Ditto.
1183 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1185 * config/rs6000/altivec.md: Write output control strings as braced
1186 blocks instead of double-quoted strings.
1187 * config/rs6000/darwin.md: Ditto.
1188 * config/rs6000/rs6000.md: Ditto.
1189 * config/rs6000/vector.md: Ditto.
1190 * config/rs6000/vsx.md: Ditto.
1192 2018-02-21 Jason Merrill <jason@redhat.com>
1194 PR c++/84314 - ICE with templates and fastcall attribute.
1195 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1197 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1199 * ipa-cp.c (determine_versionability): Fix comment typos.
1201 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1204 * ipa-cp.c (determine_versionability): Do not version functions caling
1207 2018-02-21 Martin Liska <mliska@suse.cz>
1210 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1211 Add "native" as a possible value.
1212 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1213 the macro when native cpu detection is available.
1215 2018-02-21 Martin Liska <mliska@suse.cz>
1218 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1219 Add "native" as a possible value.
1220 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1221 when native cpu detection is available.
1223 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1224 Martin Sebor <msebor@redhat.com>
1226 PR tree-optimization/84478
1227 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1229 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1230 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1231 support which is conservatively correct, for 2 only stay conservative
1232 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1233 argument to the 2 argument get_range_strlen, adjust 6 arg
1234 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1236 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1237 (gimple_fold_builtin_strlen): Pass true as last argument to
1240 2018-02-20 Martin Sebor <msebor@redhat.com>
1243 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1244 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1245 (builtin_memref::builtin_memref): Factor out parts into
1246 set_base_and_offset and call it.
1248 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1251 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1252 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1253 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1254 search at the associated MODE_INT.
1256 2018-02-20 Jeff Law <law@redhat.com>
1259 PR tree-optimization/81592
1261 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1262 for range data rather than using global data.
1263 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1264 range data rather than using global data.
1265 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1266 pass it to children as needed.
1267 (struct directive::fmtresult): Similarly.
1268 (struct directive::set_width): Similarly.
1269 (struct directive::set_precision): Similarly.
1270 (format_integer, format_directive, parse_directive): Similarly.
1271 (format_none): Accept unnamed vr_values parameter.
1272 (format_percent, format_floating, format_character): Similarly.
1273 (format_string, format_plain): Similarly.
1274 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1275 the EVRP range analyzer for range data rather than using global data.
1276 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1277 gimple-ssa-evrp-analyze.h
1278 (class sprintf_dom_walker): Add after_dom_children member function.
1279 Add evrp_range_analyzer member.
1280 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1281 range analyzer as needed.
1282 (sprintf_dom_walker::after_dom_children): New member function.
1283 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1285 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1286 (evrp_range_analyzer::pop_to_marker): Likewise.
1288 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1290 PR tree-optimization/84419
1291 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1292 with the required type if its current type is compatible but
1295 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1298 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1299 after vectorization.
1301 2018-02-20 Martin Liska <mliska@suse.cz>
1304 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1305 possible values if we don't have a hint.
1307 2018-02-20 Martin Liska <mliska@suse.cz>
1311 * final.c (shorten_branches): Build align_tab array with one
1313 * opts.c (finish_options): Add alignment option limit check.
1314 (MAX_CODE_ALIGN): Likewise.
1315 (MAX_CODE_ALIGN_VALUE): Likewise.
1316 * doc/invoke.texi: Document maximum allowed option value for
1317 all -falign-* options.
1319 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1322 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1323 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1324 * var-tracking.c (emit_note_insn_var_location): Remove all references
1325 to NOTE_INSN_CALL_ARG_LOCATION.
1326 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1327 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1328 Use copy_rtx_if_shared.
1329 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1330 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1331 (dwarf2out_var_location): Remove handling of
1332 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1334 * final.c (final_scan_insn): Remove all references to
1335 NOTE_INSN_CALL_ARG_LOCATION.
1336 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1337 before dumping final insns.
1338 * except.c (emit_note_eh_region_end): Remove all references to
1339 NOTE_INSN_CALL_ARG_LOCATION.
1340 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1341 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1342 * config/arc/arc.c (hwloop_optimize): Likewise.
1343 * config/arm/arm.c (create_fix_barrier): Likewise.
1344 * config/s390/s390.c (s390_chunkify_start): Likewise.
1345 * config/sh/sh.c (find_barrier): Likewise.
1346 * config/i386/i386.c (rest_of_insert_endbranch,
1347 ix86_seh_fixup_eh_fallthru): Likewise.
1348 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1349 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1350 * config/frv/frv.c (frv_function_prologue): Likewise.
1351 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1353 (note_outside_basic_block_p): Remove all references to
1354 NOTE_INSN_CALL_ARG_LOCATION.
1355 * gengtype.c (adjust_field_rtx_def): Likewise.
1356 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1358 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1359 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1362 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1365 PR tree-optimization/84452
1366 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1367 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1370 2018-02-19 Martin Liska <mliska@suse.cz>
1373 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1375 2018-02-19 Martin Liska <mliska@suse.cz>
1376 Richard Sandiford <richard.sandiford@linaro.org>
1378 PR tree-optimization/82491
1379 * gimple-fold.c (get_base_constructor): Make earlier bail out
1382 2018-02-19 Carl Love <cel@us.ibm.com>
1384 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1385 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1387 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1388 P8V_BUILTIN_VEC_NEG.
1390 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1392 * config/rl78/rl78.md (movdf): New define expand.
1394 2018-02-19 Martin Liska <mliska@suse.cz>
1397 * doc/invoke.texi: Fix typo.
1398 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1400 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1402 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1403 handle rs6000_single_float and rs6000_double_float specially for
1406 2018-02-16 Jeff Law <law@redhat.com>
1408 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1411 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1414 * ipa-inline.c (inline_small_functions): Fix a typo.
1416 2018-02-16 Nathan Sidwell <nathan@acm.org>
1418 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1420 2018-02-16 Carl Love <cel@us.ibm.com>
1422 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1423 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1424 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1425 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1426 expansion to P8V_BUILTIN_VEC_FLOAT2.
1428 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1430 PR rtl-optimization/70023
1431 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1432 src_regno into account.
1434 2018-02-16 Carl Love <cel@us.ibm.com>
1436 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1437 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1438 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1439 * config/rs6000/rs6000.c: Remove case statements for
1440 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1441 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1442 and P9V_BUILTIN_VEC_VINSERT4B.
1443 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1444 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1445 * config/rs6000/vsx.md:
1446 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1449 2018-02-16 Carl Love <cel@us.ibm.com>
1451 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1453 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1455 * config/rs6000/rs6000-c.c: Add the definitions for
1456 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1457 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1458 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1459 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1460 definition for insert4b and define insn *insert3b_internal.
1461 * doc/extend.texi: Add documentation for vec_extract4b.
1463 2018-02-16 Nathan Sidwell <nathan@acm.org>
1465 * doc/extend.texi (Backwards Compatibility): Mention friend
1466 injection. Note for-scope is deprecated.
1467 * doc/invoke.texi (-ffriend-injection): Deprecate.
1469 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1471 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1472 that moved to I2, also allow destinations that are a paradoxical
1473 subreg (instead of a normal reg).
1475 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1478 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1481 2018-02-16 Richard Biener <rguenther@suse.de>
1483 PR tree-optimization/84037
1484 PR tree-optimization/84016
1486 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1487 Adjust vec_construct for the fact we need additional higher latency
1488 128bit inserts for AVX256 and AVX512 vector builds.
1489 (ix86_add_stmt_cost): Scale vector construction cost for
1492 2018-02-16 Richard Biener <rguenther@suse.de>
1494 PR tree-optimization/84417
1495 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1496 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1497 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1499 2018-02-16 Martin Liska <mliska@suse.cz>
1502 * internal-fn.def (ASAN_CHECK): Set proper flags.
1503 (ASAN_MARK): Likewise.
1505 2018-02-16 Julia Koval <julia.koval@intel.com>
1507 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1508 from PTA_CANNONLAKE.
1510 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1513 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1514 Use ++iter rather than iter++ for std::list iterators.
1515 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1516 defer deleting them until all nodes in the forest are processed. Do
1517 free even leaf nodes. Change to_process into auto_vec.
1520 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1521 * vec.h (vec_default_construct): Use memset instead of placement new
1522 if BROKEN_VALUE_INITIALIZATION is defined.
1523 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1524 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1527 PR rtl-optimization/83723
1528 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1529 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1530 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1532 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1534 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1536 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1538 PR rtl-optimization/81443
1539 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1540 from inner REGs to paradoxical SUBREGs.
1542 2018-02-16 Richard Biener <rguenther@suse.de>
1544 PR tree-optimization/84399
1545 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1546 For operands we can analyze at their definition make sure we can
1547 analyze them at each use as well.
1549 2018-02-16 Richard Biener <rguenther@suse.de>
1551 PR tree-optimization/84190
1552 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1553 volatile accesses if the decl isn't volatile.
1555 2018-02-15 Jason Merrill <jason@redhat.com>
1557 PR c++/84314 - ICE with templates and fastcall attribute.
1558 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1559 TYPE_CANONICAL on an existing type.
1561 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1563 PR tree-optimization/84383
1564 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1565 dstoff nor call operand_equal_p if dstbase is NULL.
1567 PR tree-optimization/84334
1568 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1569 also a CONSTANT_CLASS_P, punt.
1571 2018-02-14 Jim Wilson <jimw@sifive.com>
1573 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1574 first SMALL_OPERAND check. New local min_second_step. Move assert
1575 to where locals are set. Add TARGET_RVC support.
1576 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1578 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1580 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1582 2018-02-14 Martin Sebor <msebor@redhat.com>
1584 PR tree-optimization/83698
1585 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1586 arrays constrain the offset range to their bounds.
1587 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1588 (builtin_access::overlap): Avoid setting the size of overlap if it's
1590 (maybe_diag_overlap): Also consider arrays when deciding what values
1591 of offsets to include in diagnostics.
1593 2018-02-14 Martin Sebor <msebor@redhat.com>
1596 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1597 that correspond to the kind of a declaration.
1599 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1602 * config/pa/pa.md: Load address of PIC label using the linkage table
1603 if the label is nonlocal.
1605 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1607 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1608 warning message if user requests -maltivec=be.
1609 * doc/invoke.texi: Document deprecation of -maltivec=be.
1611 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1614 * config/rs6000/rs6000-c.c: Update definitions for
1615 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1616 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1618 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1621 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1622 add _get_ssp intrinsics. Remove argument from
1623 __builtin_ia32_rdssp[d|q].
1624 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1625 * config/i386/i386-builtin.def: Remove argument from
1626 __builtin_ia32_rdssp[d|q].
1627 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1628 ix86_expand_special_args_builtin for _rdssp[d|q].
1629 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1630 Clear register before usage.
1631 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1632 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1634 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1636 PR tree-optimization/84357
1637 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1638 operand 1 of an ARRAY_REF too.
1640 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1643 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1644 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1646 (set_of_reg): New struct.
1647 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1648 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1649 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1651 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1652 Split into bitclr, bitset, bitinvert patterns if appropriate.
1653 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1654 use rx_fuse_in_memory_bitop.
1655 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1656 to named insn, correct maximum insn length.
1658 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1661 * machmode.def: Define a complex mode for PARTIAL_INT.
1662 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1664 * doc/rtl.texi: Document CSPImode.
1665 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1667 (msp430_hard_regno_nregs_with_padding): Likewise.
1669 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1672 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1674 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1676 PR rtl-optimization/84169
1677 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1678 we generated a parallel as new i3 and we split that to new i2 and i3
1679 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1680 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1681 those to i2, not i1. Partially rewrite this scan code.
1683 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1686 * stor-layout.c (place_field): For variable length fields, adjust
1687 offset_align afterwards not just based on the field's alignment,
1688 but also on the size.
1691 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1692 of exps and logs in the use_exp2 case.
1694 2018-02-13 Jeff Law <law@redhat.com>
1696 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1699 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1702 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1706 * common.opt (gas-loc-support, gas-locview-support): New.
1707 (ginline-points, ginternal-reset-location-views): New.
1708 * doc/invoke.texi: Document them. Use @itemx where intended.
1709 (gvariable-location-views): Adjust.
1710 * target.def (reset_location_view): New.
1711 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1712 (TARGET_RESET_LOCATION_VIEW): New.
1713 * doc/tm.texi: Rebuilt.
1714 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1715 (dwarf2out_default_as_locview_support): New.
1716 (output_asm_line_debug_info): Use option variables.
1717 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1718 (output_loc_list): Likewise.
1719 (add_high_low_attributes): Check option variables.
1720 Don't output entry view attribute in strict mode.
1721 (gen_inlined_subroutine_die): Check option variables.
1722 (dwarf2out_inline_entry): Likewise.
1723 (init_sections_and_labels): Likewise.
1724 (dwarf2out_early_finish): Likewise.
1725 (maybe_reset_location_view): New, from...
1726 (dwarf2out_var_location): ... here. Call it.
1727 * debug.h (dwarf2out_default_as_loc_support): Declare.
1728 (dwarf2out_default_as_locview_support): Declare.
1729 * hooks.c (hook_int_rtx_insn_0): New.
1730 * hooks.h (hook_int_rtx_insn_0): Declare.
1731 * toplev.c (process_options): Take -gas-loc-support and
1732 -gas-locview-support from dwarf2out. Enable
1733 -gvariable-location-views by default only with locview
1734 assembler support. Enable -ginternal-reset-location-views by
1735 default only if the target defines the corresponding hook.
1736 Enable -ginline-points by default if location views are
1737 enabled; force it disabled if statement frontiers are
1739 * tree-inline.c (expand_call_inline): Check option variables.
1740 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1742 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1744 PR tree-optimization/84321
1745 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1746 handling. Also check whether the anti-range contains any values
1747 that satisfy the mask; switch to a VR_RANGE if not.
1749 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1752 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1754 2018-02-13 Martin Jambor <mjambor@suse.cz>
1757 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1758 of call statements, also set location of a load to a temporary.
1760 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1762 * config/rl78/rl78.c (add_vector_labels): New function.
1763 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1764 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1765 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1766 which checks that no arguments are passed.
1767 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1768 * doc/extend.texi: Documentation for the new attribute.
1770 2018-02-13 Andreas Schwab <schwab@suse.de>
1772 * config/riscv/linux.h (CPP_SPEC): Define.
1774 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1777 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1778 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1779 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1780 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1781 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1782 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1783 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1784 temporarily for AES and PCLMUL builtins.
1786 PR tree-optimization/84339
1787 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1788 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1792 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1793 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1794 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1796 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1798 * omp-simd-clone.h: New file.
1799 * omp-simd-clone.c: Include omp-simd-clone.h.
1800 (expand_simd_clones): No longer static.
1801 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1802 cgraph.h and omp-simd-clone.h.
1803 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1804 (vect_recog_widen_shift_pattern): Formatting fix.
1805 (vect_pattern_recog_1): Don't check optab for calls.
1808 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1809 operands[2] into a REG before using gen_lowpart on it.
1811 2018-02-12 Jeff Law <law@redhat.com>
1814 * config/sh/sh.c (find_barrier): Consider a sibling call
1817 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1818 successfully back substituting a reg.
1820 2018-02-12 Richard Biener <rguenther@suse.de>
1822 PR tree-optimization/84037
1823 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1824 parameter, move visited init to caller.
1825 (vect_slp_analyze_operations): Separate cost from validity
1826 check, initialize visited once for all instances.
1827 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1829 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1831 (vect_model_promotion_demotion_cost): Likewise.
1832 (vectorizable_bswap): Guard cost modeling with !slp_node
1833 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1835 (vectorizable_call): Likewise.
1836 (vectorizable_conversion): Likewise.
1837 (vectorizable_assignment): Likewise.
1838 (vectorizable_shift): Likewise.
1839 (vectorizable_operation): Likewise.
1840 (vectorizable_store): Likewise.
1841 (vectorizable_load): Likewise.
1842 (vectorizable_condition): Likewise.
1843 (vectorizable_comparison): Likewise.
1845 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1848 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1849 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1850 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1852 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1855 * params.def (inline-min-speedup): Increase from 8 to 15.
1856 (max-inline-insns-auto): Decrease from 40 to 30.
1857 * ipa-split.c (consider_split): Add some buffer for function to
1858 be considered inlining candidate.
1859 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1862 2018-02-12 Richard Biener <rguenther@suse.de>
1864 PR tree-optimization/84037
1865 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1866 matched stmts if we cannot swap the non-matched ones.
1868 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1870 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1871 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1872 _mm_maskz_scalef_round_ss): New intrinsics.
1873 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1874 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1875 __builtin_ia32_scalefss_round): Remove.
1876 (__builtin_ia32_scalefsd_mask_round,
1877 __builtin_ia32_scalefss_mask_round): New intrinsics.
1878 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1879 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1880 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1881 "<round_constraint>")): Changed to ...
1882 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1883 "<round_scalar_constraint>")): ... this.
1884 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1885 %0, %1, %2<round_op3>}"): Changed to ...
1886 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1887 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1888 %2<round_scalar_mask_op3>}"): ... this.
1889 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1891 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1893 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1894 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1895 (_mm_maskz_sqrt_round_ss): New intrinsics.
1896 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1897 (__builtin_ia32_sqrtsd_mask_round)
1898 (__builtin_ia32_sqrtss_mask_round): New builtins.
1899 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1900 (__builtin_ia32_sqrtss_round): Remove.
1901 (__builtin_ia32_sqrtsd_mask_round)
1902 (__builtin_ia32_sqrtss_mask_round): New builtins.
1903 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1904 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1905 ((match_operand:VF_128 1 "vector_operand"
1906 "xBm,<round_constraint>")): Changed to ...
1907 ((match_operand:VF_128 1 "vector_operand"
1908 "xBm,<round_scalar_constraint>")): ... this.
1909 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1910 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1911 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1912 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1913 %<iptr>1<round_scalar_mask_op3>}): ... this.
1914 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1915 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1917 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1920 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1921 Cast vec_cmpeq result to correct type.
1922 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1923 Cast vec_cmpgt result to correct type.
1925 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1927 * final.c (final_scan_insn_1): Renamed from...
1928 (final_scan_insn): ... this. New wrapper, to recover
1929 seen from the outermost call in recursive ones.
1930 * config/sparc/sparc.c (output_return): Drop seen from call.
1931 (output_sibcall): Likewise.
1932 * config/visium/visium.c (output_branch): Likewise.
1934 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1936 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1939 2018-02-10 Alan Modra <amodra@gmail.com>
1942 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1943 Specify LR as an input.
1945 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1948 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1949 remove_member_access_dummy_vars): New functions.
1950 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1951 lower_omp_1, execute_lower_omp): Use them.
1953 PR rtl-optimization/84308
1954 * shrink-wrap.c (spread_components): Release todo vector.
1956 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1958 PR rtl-optimization/57193
1959 * ira-color.c (struct allocno_color_data): Add member
1960 conflict_allocno_hard_prefs.
1961 (update_conflict_allocno_hard_prefs): New.
1962 (bucket_allocno_compare_func): Add a preference based on
1963 conflict_allocno_hard_prefs.
1964 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1965 (color_allocnos): Remove a dead code. Initiate
1966 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1968 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1971 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1972 constraint from =wa to wa. Avoid a subreg on the output operand,
1973 instead use a pseudo and subreg it in a move.
1974 (p9_xxbrd_<mode>): Changed to ...
1975 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1976 (p9_xxbrd_v2df): New expander.
1977 (p9_xxbrw_<mode>): Changed to ...
1978 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1979 (p9_xxbrw_v4sf): New expander.
1981 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1983 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1985 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1988 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1989 multiply in 32-bit mode.
1990 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1991 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1994 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1996 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1997 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1998 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1999 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2001 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2004 * dwarf2out.c (is_trivial_indirect_ref): New function.
2005 (dwarf2out_late_global_decl): Do not generate a location
2006 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2007 and that are not defined in the current unit.
2009 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2011 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2012 instead of a libcall for UNORDERED.
2014 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2017 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2018 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2020 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2023 * config/s390/s390.c (s390_set_current_function): Invoke
2024 s390_indirect_branch_settings also if fndecl didn't change.
2026 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2028 * config/rs6000/rs6000.md (blockage): Set length to zero.
2030 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2032 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2034 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2037 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2038 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2042 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2043 PARALLEL incoming that failed vt_get_decl_and_offset check.
2046 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2047 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2049 (get_variable_section): For decls in named .bss* sections pass true as
2050 second argument to bss_initializer_p.
2052 2018-02-09 Marek Polacek <polacek@redhat.com>
2053 Jakub Jelinek <jakub@redhat.com>
2056 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2057 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2058 Sync some changes from cxx_fold_indirect_ref.
2060 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2062 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2064 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2065 (BLOCK_INLINE_ENTRY_LABEL): New.
2066 (dwarf2out_var_location): Disregard inline entry markers.
2067 (inline_entry_data): New struct.
2068 (inline_entry_data_hasher): New hashtable type.
2069 (inline_entry_data_hasher::hash): New.
2070 (inline_entry_data_hasher::equal): New.
2071 (inline_entry_data_table): New variable.
2072 (add_high_low_attributes): Add DW_AT_entry_pc and
2073 DW_AT_GNU_entry_view attributes if a pending entry is found
2074 in inline_entry_data_table. Add old entry_pc attribute only
2075 if debug nonbinding markers are disabled.
2076 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2077 markers are enabled.
2078 (block_within_block_p, dwarf2out_inline_entry): New.
2079 (dwarf2out_finish): Check that no entries remained in
2080 inline_entry_data_table.
2081 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2082 (final_scan_insn, notice_source_line): Likewise.
2083 (rest_of_clean_state): Skip inline entry markers.
2084 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2086 * gimple.c (gimple_build_debug_inline_entry): New.
2087 * gimple.h (enum gimple_debug_subcode): Add
2088 GIMPLE_DEBUG_INLINE_ENTRY.
2089 (gimple_build_debug_inline_entry): Declare.
2090 (gimple_debug_inline_entry_p): New.
2091 (gimple_debug_nonbind_marker_p): Adjust.
2092 * insn-notes.def (INLINE_ENTRY): New.
2093 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2094 inline entry marker notes.
2095 (print_insn): Likewise.
2096 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2097 (INSN_DEBUG_MARKER_KIND): Likewise.
2098 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2099 * tree-inline.c (expand_call_inline): Build and insert
2100 debug_inline_entry stmt.
2101 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2102 inline entry blocks early, if nonbind markers are enabled.
2103 (dump_scope_block): Dump fragment info.
2104 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2105 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2106 (gimple_build_debug_inline_entry): New.
2107 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2108 Enable/disable inline entry points too.
2109 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2110 (DEBUG_INSN): Describe inline entry markers.
2112 * common.opt (gvariable-location-views): New.
2113 (gvariable-location-views=incompat5): New.
2114 * config.in: Rebuilt.
2115 * configure: Rebuilt.
2116 * configure.ac: Test assembler for view support.
2117 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2118 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2119 * dwarf2out.c (var_loc_view): New typedef.
2120 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2121 (dwarf2out_locviews_in_attribute): New.
2122 (dwarf2out_locviews_in_loclist): New.
2123 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2124 (enum dw_line_info_opcode): Add LI_adv_address.
2125 (struct dw_line_info_table): Add view.
2126 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2127 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2128 (zero_view_p): New variable.
2129 (ZERO_VIEW_P): New macro.
2130 (output_asm_line_debug_info): New.
2131 (struct var_loc_node): Add view.
2132 (add_AT_view_list, AT_loc_list): New.
2133 (add_var_loc_to_decl): Add view param. Test it against last.
2134 (new_loc_list): Add view params. Record them.
2135 (AT_loc_list_ptr): Handle loc and view lists.
2136 (view_list_to_loc_list_val_node): New.
2137 (print_dw_val): Handle dw_val_class_view_list.
2138 (size_of_die): Likewise.
2139 (value_format): Likewise.
2140 (loc_list_has_views): New.
2141 (gen_llsym): Set vl_symbol too.
2142 (maybe_gen_llsym, skip_loc_list_entry): New.
2143 (dwarf2out_maybe_output_loclist_view_pair): New.
2144 (output_loc_list): Output view list or entries too.
2145 (output_view_list_offset): New.
2146 (output_die): Handle dw_val_class_view_list.
2147 (output_dwarf_version): New.
2148 (output_compilation_unit_header): Use it.
2149 (output_skeleton_debug_sections): Likewise.
2150 (output_rnglists, output_line_info): Likewise.
2151 (output_pubnames, output_aranges): Update version comments.
2152 (output_one_line_info_table): Output view numbers in asm comments.
2153 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2154 Call maybe_gen_llsym.
2155 (loc_list_from_tree_1): Adjust.
2156 (add_AT_location_description): Create view list attribute if
2157 needed, check it's absent otherwise.
2158 (convert_cfa_to_fb_loc_list): Adjust.
2159 (maybe_emit_file): Call output_asm_line_debug_info for test.
2160 (dwarf2out_var_location): Reset views as needed. Precompute
2161 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2162 attribute. Set view.
2163 (new_line_info_table): Reset next view.
2164 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2165 (dwarf2out_source_line): Likewise. Output view resets and labels to
2166 the assembler, or select appropriate line info opcodes.
2167 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2168 (optimize_string_length): Catch it. Adjust.
2169 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2170 dw_val_class_view_list, and remove it if no longer needed.
2171 (hash_loc_list): Hash view numbers.
2172 (loc_list_hasher::equal): Compare them.
2173 (optimize_location_lists): Check whether a view list symbol is
2174 needed, and whether the locview attribute is present, and
2175 whether they match. Remove the locview attribute if no longer
2177 (index_location_lists): Call skip_loc_list_entry for test.
2178 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2179 Use output_dwarf_version.
2180 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2181 (struct dw_val_node): Add val_view_list.
2182 * final.c (SEEN_NEXT_VIEW): New.
2183 (set_next_view_needed): New.
2184 (clear_next_view_needed): New.
2185 (maybe_output_next_view): New.
2186 (final_start_function): Rename to...
2187 (final_start_function_1): ... this. Take pointer to FIRST,
2188 add SEEN parameter. Emit param bindings in the initial view.
2189 (final_start_function): Reintroduce SEEN-less interface.
2190 (final): Rename to...
2191 (final_1): ... this. Take SEEN parameter. Output final pending
2192 next view at the end.
2193 (final): Reintroduce seen-less interface.
2194 (final_scan_insn): Output pending next view before switching
2195 sections or ending a block. Mark the next view as needed when
2196 outputting variable locations. Notify debug backend of section
2197 changes, and of location view changes.
2198 (rest_of_handle_final): Adjust.
2199 * toplev.c (process_options): Autodetect value for debug variable
2200 location views option. Warn on incompat5 without -gdwarf-5.
2201 * doc/invoke.texi (gvariable-location-views): New.
2202 (gvariable-location-views=incompat5): New.
2203 (gno-variable-location-views): New.
2205 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2207 PR tree-optimization/84136
2208 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2209 that the result of find_edge is non-NULL.
2211 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2214 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2215 storing integer register in SImode. Fix cost of 256 and 512
2216 byte aligned SSE register store.
2218 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2220 * config/i386/i386.c (ix86_multiplication_cost): Fix
2221 multiplication cost for TARGET_AVX512DQ.
2223 2018-02-08 Marek Polacek <polacek@redhat.com>
2225 PR tree-optimization/84238
2226 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2229 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2231 PR tree-optimization/84265
2232 * tree-vect-stmts.c (vectorizable_store): Don't treat
2233 VMAT_CONTIGUOUS accesses as grouped.
2234 (vectorizable_load): Likewise.
2236 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2238 PR tree-optimization/81635
2239 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2240 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2241 (test_round_for_mask): New functions.
2242 (wide_int_cc_tests): Call test_round_for_mask.
2243 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2244 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2245 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2246 range returned by get_range_info.
2248 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2251 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2252 * symtab.c: Include builtins.h
2253 (symtab_node::output_to_lto_symbol_table_p): Move here
2254 from lto-streamer-out.c:output_symbol_p.
2255 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2256 (output_symbol_p): Move all logic to symtab.c
2257 (produce_symtab): Update.
2259 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2261 * config/s390/s390-opts.h (enum indirect_branch): Define.
2262 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2263 (s390_indirect_branch_via_thunk)
2264 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2265 (enum s390_indirect_branch_type): Define.
2266 * config/s390/s390.c (struct s390_frame_layout, struct
2267 machine_function): Remove.
2268 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2269 (indirect_branch_table_label_no, indirect_branch_table_name):
2271 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2272 (enum s390_indirect_branch_option): Define.
2273 (s390_return_addr_from_memory): New function.
2274 (s390_handle_string_attribute): New function.
2275 (s390_attribute_table): Add new attribute handler.
2276 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2277 (s390_indirect_branch_via_thunk): New function.
2278 (s390_indirect_branch_via_inline_thunk): New function.
2279 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2280 sibling call optimization for non z10 compiles.
2281 (s390_emit_call): Force indirect branch target to be a single
2282 register. Add r1 clobber for non-z10 compiles.
2283 (s390_emit_epilogue): Emit return jump via return_use expander.
2284 (s390_reorg): Handle JUMP_INSNs as execute targets.
2285 (s390_option_override_internal): Perform validity checks for the
2286 new command line options.
2287 (s390_indirect_branch_attrvalue): New function.
2288 (s390_indirect_branch_settings): New function.
2289 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2290 (s390_output_indirect_thunk_function): New function.
2291 (s390_code_end): Implement target hook.
2292 (s390_case_values_threshold): Implement target hook.
2293 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2295 * config/s390/s390.h (struct s390_frame_layout)
2296 (struct machine_function): Move here from s390.c.
2297 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2298 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2299 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2300 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2301 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2302 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2303 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2304 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2305 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2306 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2307 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2308 (mnemonic attribute): Add values which aren't recognized
2310 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2311 pattern for branch conversion. Fix mnemonic attribute.
2312 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2313 indirect branch via thunk if requested.
2314 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2315 ("*indirect_jump"): Disable for branch conversion using out of
2317 ("indirect_jump_via_thunk<mode>_z10")
2318 ("indirect_jump_via_thunk<mode>")
2319 ("indirect_jump_via_inlinethunk<mode>_z10")
2320 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2321 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2322 ("casesi_jump_via_inlinethunk<mode>_z10")
2323 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2324 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2325 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2326 ("*indirect2_jump"): Disable for branch conversion.
2327 ("casesi_jump"): Turn into expander and expand patterns for branch
2329 ("return_use"): New expander.
2330 ("*return"): Emit return via thunk and rename it to ...
2331 ("*return<mode>"): ... this one.
2332 * config/s390/s390.opt: Add new options and and enum for the
2335 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2337 * lra-constraints.c (match_reload): Unconditionally use
2338 gen_lowpart_SUBREG, rather than selecting between that
2339 and equivalent gen_rtx_SUBREG code.
2341 2018-02-08 Richard Biener <rguenther@suse.de>
2343 PR tree-optimization/84233
2344 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2345 changed flag instead of boguously re-using phi_inserted.
2347 2018-02-08 Martin Jambor <mjambor@suse.cz>
2349 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2350 static local variables.
2352 2018-02-08 Richard Biener <rguenther@suse.de>
2354 PR tree-optimization/84278
2355 * tree-vect-stmts.c (vectorizable_store): When looking for
2356 smaller vector types to perform grouped strided loads/stores
2357 make sure the mode is supported by the target.
2358 (vectorizable_load): Likewise.
2360 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2362 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2363 Increase LDP/STP opportunities by adding adjacent callee-saves.
2365 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2367 PR rtl-optimization/84068
2368 PR rtl-optimization/83459
2369 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2371 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2373 PR tree-optimization/84224
2374 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2375 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2378 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2381 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2382 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2383 count, remove one USE.
2385 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2387 * doc/install.texi (Configuration): Document the
2388 --with-long-double-format={ibm,ieee} PowerPC configuration
2392 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2393 Convert from define_expand to be define_insn_and_split. Rework
2394 float/double/_Float128 conversions to QI/HI/SImode to work with
2395 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2396 conversions to QI/HImode types did a store and then a load to
2397 truncate the value. For conversions to VSX registers, don't split
2398 the insn, instead emit the code directly. Use the code iterator
2399 any_fix to combine signed and unsigned conversions.
2400 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2401 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2402 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2403 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2404 (fix_<mode>di2_hw): Likewise.
2405 (fixuns_<mode>di2_hw): Likewise.
2406 (fix_<mode>si2_hw): Likewise.
2407 (fixuns_<mode>si2_hw): Likewise.
2408 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2409 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2410 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2411 fix<uns>_trunc<SFDF:mode>si2_p8.
2412 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2414 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2415 (fix<uns>_<mode>_mem): Likewise.
2416 (fctiw<u>z_<mode>_mem): Likewise.
2417 (fix<uns>_<mode>_mem): Likewise.
2418 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2419 the register allocator from doing a direct move to the GPRs to do
2420 a store, and instead use the ISA 3.0 store byte/half-word from
2421 vector register instruction. For IEEE 128-bit floating point,
2422 also optimize stores of 32-bit ints.
2423 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2425 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2427 * genextract.c (push_pathstr_operand): New function to support
2429 (walk_rtx): Call push_pathstr_operand.
2430 (print_path): Support [a-zA-Z].
2432 2018-02-07 Richard Biener <rguenther@suse.de>
2434 PR tree-optimization/84037
2435 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2436 (cse_and_gimplify_to_preheader): Declare.
2437 (vect_get_place_in_interleaving_chain): Likewise.
2438 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2440 (_loop_vec_info::~_loop_vec_info): Delete it.
2441 (cse_and_gimplify_to_preheader): New function.
2442 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2443 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2444 (vectorizable_load): Likewise. For grouped stores always base
2445 the IV on the first element.
2446 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2447 condition before gimplifying.
2449 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2451 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2452 *DIV_EXPR and *MOD_EXPR.
2454 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2457 * config/i386/i386.c (ix86_option_override_internal): Mask out
2458 the CF_SET bit when checking -fcf-protection.
2460 2018-02-07 Tom de Vries <tom@codesourcery.com>
2463 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2466 2018-02-07 Richard Biener <rguenther@suse.de>
2468 PR tree-optimization/84204
2469 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2472 PR tree-optimization/84205
2473 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2474 special-case isl_ast_op_zdiv_r.
2476 PR tree-optimization/84223
2477 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2478 Only add conditions from within the region.
2479 (gather_bbs::after_dom_children): Adjust.
2481 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2484 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2485 * config/avr/avr.md: Only post-reload split REG-REG moves if
2486 either register is GENERAL_REG_P.
2488 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2490 PR tree-optimization/84235
2491 * tree-ssa-scopedtables.c
2492 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2493 if the subtraction is performed in floating point type where NaNs are
2494 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2495 build 1. Formatting fix.
2497 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2500 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2501 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2502 and skip it regardless of bb boundaries. Use CALL_P macro,
2503 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2504 unnecessarily, formatting fix.
2506 2018-02-06 Michael Collison <michael.collison@arm.com>
2508 * config/arm/thumb2.md:
2509 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2510 (*thumb_mov_notscc): Ditto.
2512 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2515 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2516 unsigned_fix, not "s".
2518 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2520 * configure.ac (gcc_fn_eh_frame_ro): New function.
2521 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2522 correct .eh_frame permissions.
2523 * configure: Regenerate.
2525 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2527 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2530 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2532 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2533 Display warning message for -mno-speculate-indirect-jumps.
2535 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2537 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2539 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2541 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2543 PR tree-optimization/84225
2544 * tree-eh.c (find_trapping_overflow): Only call
2545 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2547 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2550 * config/i386/i386.c: Reimplement the check of possible options
2551 -mibt/-mshstk conbination. Change error messages.
2552 * doc/invoke.texi: Fix a typo: remove extra '='.
2554 2018-02-06 Marek Polacek <polacek@redhat.com>
2556 PR tree-optimization/84228
2557 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2559 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2562 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2563 emitted arch directives.
2564 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2565 __ARM_FEATURE_COPROC before changing architectures.
2567 2018-02-06 Richard Biener <rguenther@suse.de>
2569 * config/i386/i386.c (print_reg): Fix typo.
2570 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2572 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2574 * configure: Regenerate.
2576 2018-02-05 Martin Sebor <msebor@redhat.com>
2578 PR tree-optimization/83369
2579 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2582 2018-02-05 Martin Liska <mliska@suse.cz>
2584 * doc/invoke.texi: Cherry-pick upstream r323995.
2586 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2588 * ira.c (ira_init_register_move_cost): Adjust comment.
2590 2018-02-05 Martin Liska <mliska@suse.cz>
2592 PR gcov-profile/84137
2593 * doc/gcov.texi: Fix typo in documentation.
2595 2018-02-05 Martin Liska <mliska@suse.cz>
2597 PR gcov-profile/83879
2598 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2599 using dlopen functionality.
2601 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2603 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2604 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2605 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2606 _mm_maskz_range_round_ss): New intrinsics.
2607 (__builtin_ia32_rangesd128_round)
2608 (__builtin_ia32_rangess128_round): Remove.
2609 (__builtin_ia32_rangesd128_mask_round,
2610 __builtin_ia32_rangess128_mask_round): New builtins.
2611 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2612 __builtin_ia32_rangess128_round): Remove.
2613 (__builtin_ia32_rangesd128_mask_round,
2614 __builtin_ia32_rangess128_mask_round): New builtins.
2615 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2616 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2617 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2618 "<round_saeonly_constraint>")): Changed to ...
2619 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2620 "<round_saeonly_scalar_constraint>")): ... this.
2621 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2622 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2623 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2624 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2625 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2627 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2629 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2631 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2632 Remove all values except native, 8540 and 8548.
2634 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2636 * config/i386/i386.c (ix86_output_function_return): Pass
2637 INVALID_REGNUM, instead of -1, as invalid register number to
2638 indirect_thunk_name and output_indirect_thunk.
2640 2018-02-02 Julia Koval <julia.koval@intel.com>
2642 * config.gcc: Add -march=icelake.
2643 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2644 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2645 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2646 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2647 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2648 (processor_target_table): Add icelake.
2649 (ix86_option_override_internal): Handle new PTAs.
2650 (get_builtin_code_for_version): Handle icelake.
2651 (M_INTEL_COREI7_ICELAKE): New.
2652 (fold_builtin_cpu): Handle icelake.
2653 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2654 * doc/invoke.texi: Add -march=icelake.
2656 2018-02-02 Julia Koval <julia.koval@intel.com>
2658 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2659 to wide_int_bitmask.
2660 * wide-int-bitmask.h: New.
2662 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2665 * config/i386/i386.md: Replace Pmode with word_mode in
2666 builtin_setjmp_setup and builtin_longjmp to support x32.
2668 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2672 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2674 (rs6000_supported_cpu_names): New static variable.
2675 (linux_cpu_translation_table): Likewise.
2676 (elf_platform) <cpu>: Define new static variable and use it.
2677 Translate kernel AT_PLATFORM name to canonical name if needed.
2678 Error if platform name is unknown.
2680 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2683 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2685 2018-02-01 Jeff Law <law@redhat.com>
2688 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2689 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2690 the scratch if RELEASE_VIA_POP is false.
2691 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2692 If we have to save a temporary register, decrement SIZE appropriately.
2693 Pass new arguments to release_scratch_register_on_entry.
2694 (ix86_adjust_stack_and_probe): Likewise.
2695 (ix86_emit_probe_stack_range): Pass new arguments to
2696 release_scratch_register_on_entry.
2698 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2700 PR rtl-optimization/84157
2701 * combine.c (change_zero_ext): Use REG_P predicate in
2702 front of HARD_REGISTER_P predicate.
2704 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2706 * config/avr/avr.c (avr_option_override): Move disabling of
2707 -fdelete-null-pointer-checks to...
2708 * common/config/avr/avr-common.c (avr_option_optimization_table):
2711 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2713 PR tree-optimization/81635
2714 * tree-data-ref.c (split_constant_offset_1): For types that
2715 wrap on overflow, try to use range info to prove that wrapping
2718 2018-02-01 Renlin Li <renlin.li@arm.com>
2721 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2723 (aarch64_register_move_cost): Likewise.
2724 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2726 (REG_CLASS_NAMES): Likewise.
2727 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2728 TAILCALL_ADDR_REGS. Remove IP registers.
2729 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2731 2018-02-01 Richard Biener <rguenther@suse.de>
2733 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2734 for specifying RPO order and allow NULL for that.
2735 * domwalk.c (dom_walker::dom_walker): Likewise.
2736 (dom_walker::walk): Handle NULL RPO order.
2737 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2739 (rewrite_update_dom_walker): Likewise.
2740 (mark_def_dom_walker): Likewise.
2742 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2744 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2745 (aarch64_maybe_expand_sve_subreg_move): Declare.
2746 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2747 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2749 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2750 that are semantically a reverse operation.
2751 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2752 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2753 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2755 (aarch64_can_change_mode_class): For big-endian, forbid changes
2756 between two SVE modes if they have different element sizes.
2758 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2760 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2761 the TImode handling for big-endian targets.
2763 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2765 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2766 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2768 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2769 Remove BSWAP handing for big-endian targets and use the form of
2770 LD1RQ appropariate for the mode.
2772 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2774 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2775 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2778 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2781 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2782 check for operands that need to go through aarch64_sve_reload_be.
2784 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2786 PR tree-optimization/81661
2787 PR tree-optimization/84117
2788 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2789 * tree-eh.c: Include gimplify.h.
2790 (find_trapping_overflow, replace_trapping_overflow,
2791 rewrite_to_non_trapping_overflow): New functions.
2792 * tree-vect-loop.c: Include tree-eh.h.
2793 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2794 * tree-data-ref.c: Include tree-eh.h.
2795 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2797 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2799 PR rtl-optimization/84123
2800 * combine.c (change_zero_ext): Check if hard register satisfies
2801 can_change_dest_mode before calling gen_lowpart_SUBREG.
2803 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2806 * ira.c (ira_init_register_move_cost): Remove assert.
2808 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2810 PR rtl-optimization/84071
2811 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2812 * doc/tm.texi: Regenerate.
2814 2018-01-31 Richard Biener <rguenther@suse.de>
2816 PR tree-optimization/84132
2817 * tree-data-ref.c (analyze_miv_subscript): Properly
2818 check whether evolution_function_is_affine_multivariate_p
2819 before calling gcd_of_steps_may_divide_p.
2821 2018-01-31 Julia Koval <julia.koval@intel.com>
2824 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2825 * config/i386/i386.md (rdpid_rex64) New.
2826 (rdpid): Make 32bit only.
2828 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2831 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2832 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2834 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2837 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2839 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2841 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2843 PR rtl-optimization/84071
2844 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2845 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2847 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2849 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2850 (arc_attribute_table): Add 'aux' attribute.
2851 (arc_in_small_data_p): Consider aux like variables.
2852 (arc_is_aux_reg_p): New function.
2853 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2854 (arc_get_aux_arg): New function.
2855 (prepare_move_operands): Handle aux-register access.
2856 (arc_handle_aux_attribute): New function.
2857 * doc/extend.texi (ARC Variable attributes): Add subsection.
2859 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2861 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2862 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2863 (arc_attribute_table): Add 'uncached' attribute.
2864 (arc_print_operand): Print '.di' flag for uncached memory
2866 (arc_in_small_data_p): Do not consider for small data the uncached
2868 (arc_is_uncached_mem_p): New function.
2869 * config/arc/predicates.md (compact_store_memory_operand): Check
2870 for uncached memory accesses.
2871 (nonvol_nonimm_operand): Likewise.
2872 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2874 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2877 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2878 falign-loops=): Add Optimization flag.
2880 2018-01-30 Jeff Law <law@redhat.com>
2883 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2884 INT_REGISTERS_SAVED. Check it prior to calling
2885 get_scratch_register_on_entry.
2886 (ix86_adjust_stack_and_probe): Similarly.
2887 (ix86_emit_probe_stack_range): Similarly.
2888 (ix86_expand_prologue): Corresponding changes.
2890 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2893 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2894 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2896 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2899 * lra-constraints.c (curr_insn_transform): Process AND in the
2902 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2904 PR rtl-optimization/83986
2905 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2906 dependence against last_pending_memory_flush in addition to
2909 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2911 PR tree-optimization/81611
2912 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2915 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2918 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2921 2018-01-30 Richard Biener <rguenther@suse.de>
2922 Jakub Jelinek <jakub@redhat.com>
2924 PR tree-optimization/84111
2925 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2926 inner loops added during recursion, as they don't have up-to-date
2929 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2932 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2933 (can_inline_edge_by_limits_p): ... here.
2934 (can_early_inline_edge_p, check_callers,
2935 update_caller_keys, update_callee_keys, recursive_inlining,
2936 add_new_edges_to_heap, speculation_useful_p,
2937 inline_small_functions,
2938 inline_small_functions, flatten_function,
2939 inline_to_all_callers_1): Update.
2941 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2943 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2946 2018-01-30 Richard Biener <rguenther@suse.de>
2948 PR tree-optimization/83008
2949 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2950 invariant and constant vector uses in stmts when they need
2953 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2956 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2957 * configure: Regenerate.
2959 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2961 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2963 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2964 Use gen_rtx_REG rather than gen_lowpart.
2966 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2968 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2969 rather than 0 when creating partial subregs.
2971 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2973 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2976 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2979 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2980 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2981 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2982 flags. This restores the settings used before the 2017-07-24.
2983 Turning off pre increment/decrement/modify allows IVOPTS to
2984 optimize DF/SF loops where the index is an int.
2986 2018-01-29 Richard Biener <rguenther@suse.de>
2987 Kelvin Nilsen <kelvin@gcc.gnu.org>
2990 * tree-vect-stmts.c (vectorizable_call): Don't call
2991 targetm.vectorize_builtin_md_vectorized_function if callee is
2994 2018-01-22 Carl Love <cel@us.ibm.com>
2996 * doc/extend.tex: Fix typo in second arg in
2997 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2999 2018-01-29 Richard Biener <rguenther@suse.de>
3001 PR tree-optimization/84086
3002 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3003 (flush_ssaname_freelist): When SSA names were released reset
3004 the SCEV hash table.
3006 2018-01-29 Richard Biener <rguenther@suse.de>
3008 PR tree-optimization/84057
3009 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3010 removed paths when removing edges.
3012 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3014 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3015 -mfunction-return=@var{choice}.
3017 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3020 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3022 (layout::print_source_line): Likewise.
3023 (test_get_line_width_without_trailing_whitespace): Add test cases.
3025 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3028 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3031 2018-01-26 Jim Wilson <jimw@sifive.com>
3033 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3035 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3038 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3040 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3041 and CMP + SUB-immediate -> SUBS.
3043 2018-01-26 Martin Sebor <msebor@redhat.com>
3045 PR tree-optimization/83896
3046 * tree-ssa-strlen.c (get_string_len): Rename...
3047 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3048 Avoid assuming length is constant.
3049 (handle_char_store): Use HOST_WIDE_INT for string length.
3051 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3054 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3055 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3057 2018-01-26 Richard Biener <rguenther@suse.de>
3059 PR rtl-optimization/84003
3060 * dse.c (record_store): Only record redundant stores when
3061 the earlier store aliases at least all accesses the later one does.
3063 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3065 PR rtl-optimization/83985
3066 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3067 REG_CFA_RESTORE insns.
3068 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3069 REG_CFA_RESTORE insns here.
3072 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3073 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3075 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3077 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3078 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3079 (arc_init): Likewise.
3080 (arc_override_options): Likewise.
3081 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3083 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3085 * config/arc/arc.h (TARGET_DBNZ): Define.
3086 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3087 properly set the tune attribute.
3088 (dbnz): Use TARGET_DBNZ guard.
3089 * config/arc/arc.opt (mtune): Add core3 option.
3091 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3093 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3094 recognize new pic like addresses.
3095 (arc_delegitimize_address): Clean up.
3097 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3099 * config/arc/arc-arches.def: Option mrf16 valid for all
3101 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3102 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3103 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3104 * config/arc/arc-tables.opt: Regenerate.
3105 * config/arc/arc.c (arc_conditional_register_usage): Handle
3106 reduced register file case.
3107 (arc_file_start): Set must have build attributes.
3108 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3110 * config/arc/arc.opt (mrf16): Add new option.
3111 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3112 * config/arc/genmultilib.awk: Handle new mrf16 option.
3113 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3114 * config/arc/t-multilib: Regenerate.
3115 * doc/invoke.texi (ARC Options): Document mrf16 option.
3117 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3119 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3120 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3121 (arc_attribute_table): Add 'secure_call' attribute.
3122 (arc_print_operand): Print secure call operand.
3123 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3125 (arc_is_secure_call_p): New function. * config/arc/arc.md
3126 (call_i): Add support for sjli instruction.
3127 (call_value_i): Likewise.
3128 * config/arc/constraints.md (Csc): New constraint.
3130 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3131 John Eric Martin <John.Martin@emmicro-us.com>
3133 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3134 * config/arc/arc.c (_arc_jli_section): New struct.
3135 (arc_jli_section): New type.
3136 (rc_jli_sections): New static variable.
3137 (arc_handle_jli_attribute): New function.
3138 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3139 (arc_file_end): New function.
3140 (TARGET_ASM_FILE_END): Define.
3141 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3142 (arc_add_jli_section): New function.
3143 (jli_call_scan): Likewise.
3144 (arc_reorg): Call jli_call_scan.
3145 (arc_output_addsi): Remove 'S' from printing asm operand.
3146 (arc_is_jli_call_p): New function.
3147 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3149 (movhi_insn): Likewise.
3150 (movsi_insn): Likewise.
3151 (movsi_set_cc_insn): Likewise.
3152 (loadqi_update): Likewise.
3153 (load_zeroextendqisi_update): Likewise.
3154 (load_signextendqisi_update): Likewise.
3155 (loadhi_update): Likewise.
3156 (load_zeroextendhisi_update): Likewise.
3157 (load_signextendhisi_update): Likewise.
3158 (loadsi_update): Likewise.
3159 (loadsf_update): Likewise.
3160 (movsicc_insn): Likewise.
3161 (bset_insn): Likewise.
3162 (bxor_insn): Likewise.
3163 (bclr_insn): Likewise.
3164 (bmsk_insn): Likewise.
3165 (bicsi3_insn): Likewise.
3166 (cmpsi_cc_c_insn): Likewise.
3167 (movsi_ne): Likewise.
3168 (movsi_cond_exec): Likewise.
3169 (clrsbsi2): Likewise.
3180 (call_i): Remove 'S' asm letter, add jli instruction.
3181 (call_value_i): Likewise.
3182 * config/arc/arc.op (mjli-always): New option.
3183 * config/arc/constraints.md (Cji): New constraint.
3184 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3186 (subsf3_fpx): Likewise.
3187 (mulsf3_fpx): Likewise.
3188 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3190 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3192 * doc/invoke.texi (ARC): Document mjli-always option.
3194 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3196 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3197 avoid addition with 0 and use incw and decw where possible.
3199 2018-01-26 Richard Biener <rguenther@suse.de>
3201 PR tree-optimization/81082
3202 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3203 association if it requires casting to unsigned.
3204 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3205 from fold_plusminus_mult_expr to catch important cases late when
3206 range info is available.
3208 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3210 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3211 * configure.ac (hidden_linkonce): New test.
3212 * configure: Regenerate.
3213 * config.in: Regenerate.
3215 2018-01-26 Julia Koval <julia.koval@intel.com>
3217 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3218 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3219 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3220 _mm_mask_bitshuffle_epi64_mask): Fix type.
3221 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3222 USI_FTYPE_V4DI_V4DI_USI): Remove.
3223 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3224 __builtin_ia32_vpshufbitqmb256_mask,
3225 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3226 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3227 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3229 2018-01-26 Alan Modra <amodra@gmail.com>
3232 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3233 UNSPEC_VBPERMQ. Sort other unspecs.
3235 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3237 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3239 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3242 * predict.c (drop_profile): Do not push/pop cfun; update also
3244 (handle_missing_profiles): Fix logic looking for zero profiles.
3246 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3249 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3250 on functions with #pragma omp declare simd or functions with simd
3252 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3253 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3254 Remove trailing \n from warning_at calls.
3256 2018-01-25 Tom de Vries <tom@codesourcery.com>
3259 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3260 for neutered workers.
3262 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3265 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3266 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3268 2018-01-24 Jeff Law <law@redhat.com>
3271 * i386.c (get_probe_interval): Move to earlier point.
3272 (ix86_compute_frame_layout): If -fstack-clash-protection and
3273 the frame is larger than the probe interval, then use pushes
3274 to save registers rather than reg->mem moves.
3275 (ix86_expand_prologue): Remove conditional for int_registers_saved
3278 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3281 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3282 min/max for never referenced object.
3284 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3287 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3289 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3290 attributes from DECL_ATTRIBUTES (decl) without affecting
3291 DECL_ATTRIBUTES (current_function_decl).
3292 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3293 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3295 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3297 PR tree-optimization/83979
3298 * fold-const.c (fold_comparison): Use constant_boolean_node
3299 instead of boolean_{true,false}_node.
3301 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3303 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3306 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3308 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3309 Simplify the clause that sets the length attribute.
3310 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3311 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3312 clause that sets the length attribute.
3313 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3315 2018-01-24 Tom de Vries <tom@codesourcery.com>
3318 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3319 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3320 Add strict parameter.
3321 (prevent_branch_around_nothing): Insert dummy insn between branch to
3322 label and label with no ptx insn inbetween.
3323 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3325 2018-01-24 Tom de Vries <tom@codesourcery.com>
3328 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3329 for neutered threads in warp.
3330 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3332 2018-01-24 Richard Biener <rguenther@suse.de>
3334 PR tree-optimization/83176
3335 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3338 2018-01-24 Richard Biener <rguenther@suse.de>
3340 PR tree-optimization/82819
3341 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3342 code generating pluses that are no-ops in the target precision.
3344 2018-01-24 Richard Biener <rguenther@suse.de>
3347 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3349 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3351 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3352 to merge probabilities.
3353 * predict.c (probably_never_executed): Also mark as cold functions
3354 with global 0 profile and guessed local profile.
3355 * profile-count.c (profile_probability::combine_with_count): New
3357 * profile-count.h (profile_probability::operator*,
3358 profile_probability::operator*=, profile_probability::operator/,
3359 profile_probability::operator/=): Reduce precision to adjusted
3360 and set value to guessed on contradictory divisions.
3361 (profile_probability::combine_with_freq): Remove.
3362 (profile_probability::combine_wiht_count): Declare.
3363 (profile_count::force_nonzero):: Set to adjusted.
3364 (profile_count::probability_in):: Set quality to adjusted.
3365 * tree-ssa-tail-merge.c (replace_block_by): Use
3368 2018-01-23 Andrew Waterman <andrew@sifive.com>
3369 Jim Wilson <jimw@sifive.com>
3371 * config/riscv/riscv.c (riscv_stack_boundary): New.
3372 (riscv_option_override): Set riscv_stack_boundary. Handle
3373 riscv_preferred_stack_boundary_arg.
3374 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3375 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3376 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3377 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3378 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3379 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3381 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3384 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3385 of struct ix86_frame.
3386 (ix86_expand_epilogue): Likewise. Add a local variable for
3387 the reg_save_offset field in struct ix86_frame.
3389 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3391 PR tree-optimization/82604
3392 * tree-loop-distribution.c (enum partition_kind): New enum item
3393 PKIND_PARTIAL_MEMSET.
3394 (partition_builtin_p): Support above new enum item.
3395 (generate_code_for_partition): Ditto.
3396 (compute_access_range): Differentiate cases that equality can be
3397 proven at all loops, the innermost loops or no loops.
3398 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3399 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3400 (finalize_partitions, distribute_loop): Don't fuse partition of
3401 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3402 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3405 2018-01-23 Martin Liska <mliska@suse.cz>
3407 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3408 order to ignore the predictor.
3409 (PRED_POLYMORPHIC_CALL): Likewise.
3410 (PRED_RECURSIVE_CALL): Likewise.
3412 2018-01-23 Martin Liska <mliska@suse.cz>
3414 * tree-profile.c (tree_profiling): Print function header to
3415 aware reader which function we are working on.
3416 * value-prof.c (gimple_find_values_to_profile): Do not print
3417 not interesting value histograms.
3419 2018-01-23 Martin Liska <mliska@suse.cz>
3421 * profile-count.h (enum profile_quality): Add
3422 profile_uninitialized as the first value. Do not number values
3423 as they are zero based.
3424 (profile_count::verify): Update sanity check.
3425 (profile_probability::verify): Likewise.
3427 2018-01-23 Nathan Sidwell <nathan@acm.org>
3429 * doc/invoke.texi (ffor-scope): Deprecate.
3431 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3433 PR tree-optimization/83510
3434 * domwalk.c (set_all_edges_as_executable): New function.
3435 (dom_walker::dom_walker): Convert bool param
3436 "skip_unreachable_blocks" to enum reachability. Move setup of
3437 edge flags to set_all_edges_as_executable and only do it when
3438 reachability is REACHABLE_BLOCKS.
3439 * domwalk.h (enum dom_walker::reachability): New enum.
3440 (dom_walker::dom_walker): Convert bool param
3441 "skip_unreachable_blocks" to enum reachability.
3442 (set_all_edges_as_executable): New decl.
3443 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3444 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3446 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3447 but converting true to REACHABLE_BLOCKS.
3448 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3450 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3451 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3452 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3454 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3455 if check_all_array_refs will be called.
3457 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3459 * tree.c (selftest::test_location_wrappers): Add more test
3462 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3464 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3465 (selftest::test_bit_in_range): Likewise.
3467 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3470 * doc/sourcebuild.texi (vect_float): Say that the selector
3471 only describes the situation when -funsafe-math-optimizations is on.
3472 (vect_float_strict): Document.
3474 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3476 PR tree-optimization/83965
3477 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3478 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3479 instead of checking only for a reduction.
3480 (vect_recog_widen_sum_pattern): Likewise.
3482 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3484 * predict.c (probably_never_executed): Only use precise profile info.
3485 (compute_function_frequency): Skip after inlining hack since we now
3486 have quality checking.
3488 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3490 * profile-count.h (profile_probability::very_unlikely,
3491 profile_probability::unlikely, profile_probability::even): Set
3492 precision to guessed.
3494 2018-01-23 Richard Biener <rguenther@suse.de>
3496 PR tree-optimization/83963
3497 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3498 Properly terminate dominator walk when crossing the exit edge not
3499 when visiting its source block.
3501 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3504 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3505 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3507 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3509 PR tree-optimization/83957
3510 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3511 semicolon after for body surrounded by braces.
3513 PR tree-optimization/83081
3514 * profile-count.h (profile_probability::split): New method.
3515 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3516 Use profile_probability::split.
3517 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3518 when splitting a single conditional jump into 2.
3520 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3522 PR tree-optimization/69452
3523 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3526 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3528 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3529 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3530 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3532 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3534 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3535 * config/rl78/rl78.md (movdi): New define_expand.
3536 * config/rl78/rl78.c (rl78_split_movdi): New function.
3538 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3541 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3543 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3544 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3545 128-bit to produce an UNSPEC move to get the double word with the
3546 signbit and then a shift directly to do signbit.
3547 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3548 implementation with a new version that just does either a direct
3549 move or a regular move. Move memory interface to separate insns.
3550 Move insns so they are next to the expander.
3551 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3552 with signbit move. Split big and little endian case.
3553 (signbit<mode>2_dm_mem_le): Likewise.
3554 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3555 (signbit<mode>2_dm2): Likewise.
3557 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3559 * config/rl78/rl78.md (anddi3): New define_expand.
3561 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3563 * config/rl78/rl78.md (umindi3): New define_expand.
3565 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3567 * config/rl78/rl78.md (smindi3): New define_expand.
3569 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3571 * config/rl78/rl78.md (smaxdi3): New define_expand.
3573 2018-01-22 Carl Love <cel@us.ibm.com>
3575 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3576 LVX_V1TI): Add macro expansion.
3577 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3578 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3579 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3580 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3581 Change check to determine if the instruction is a byte reversing
3582 entry. Fix typo in comment.
3583 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3584 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3585 Add def_builtin calls for new builtins.
3586 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3587 Add define_insn expansion.
3589 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3591 * config/rl78/rl78.md (umaxdi3): New define_expand.
3593 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3595 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3596 for non-QImode registers.
3598 2018-01-22 Richard Biener <rguenther@suse.de>
3600 PR tree-optimization/83963
3601 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3602 including the loop exit block.
3603 (scop_detection::merge_sese): Likewise.
3604 (scop_detection::add_scop): Do it here instead.
3606 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3608 * doc/sourcebuild.texi (arm_softfloat): Document.
3610 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3613 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3614 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3615 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3617 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3618 David Edelsohn <dje.gcc@gmail.com>
3621 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3622 Change "crset eq" to "crset 2".
3623 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3624 (*call_indirect_aix<mode>_nospec): Likewise.
3625 (*call_value_indirect_aix<mode>_nospec): Likewise.
3626 (*call_indirect_elfv2<mode>_nospec): Likewise.
3627 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3628 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3629 change assembly output from . to $.
3630 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3631 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3632 (*tablejump<mode>_internal1_nospec): Likewise.
3634 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3637 * config/sh/sh_optimize_sett_clrt.cc:
3638 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3640 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3642 PR tree-optimization/83940
3643 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3644 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3645 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3646 use it to pass back the definition type.
3647 (vect_check_store_rhs): Likewise rhs_dt_out.
3648 (vect_build_gather_load_calls): Add a mask_dt argument and use
3649 it instead of a call to vect_is_simple_use.
3650 (vectorizable_store): Update calls to vect_check_load_store_mask
3651 and vect_check_store_rhs. Use the dt returned by the latter instead
3652 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3653 instead of calls to vect_is_simple_use. Pass the scalar rather
3654 than the vector operand to vect_is_simple_use when handling
3655 second and subsequent copies of an rhs value.
3656 (vectorizable_load): Update calls to vect_check_load_store_mask
3657 and vect_build_gather_load_calls. Use the cached mask_dt and
3658 gs_info.offset_dt instead of calls to vect_is_simple_use.
3660 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3663 * tree-emutls.c: Include gimplify.h.
3664 (lower_emutls_2): New function.
3665 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3666 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3667 it before further processing.
3670 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3671 UINTVAL (trueop1) instead of INTVAL (op1).
3673 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3677 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3678 INCOMING_FRAME_SP_OFFSET if not defined.
3679 (scan_trace): Add ENTRY argument. If true and
3680 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3681 emit a note to adjust the CFA offset.
3682 (create_cfi_notes): Adjust scan_trace callers.
3683 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3684 INCOMING_FRAME_SP_OFFSET in the CIE.
3685 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3686 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3688 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3689 * doc/tm.texi: Regenerated.
3691 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3693 PR rtl-optimization/83147
3694 * lra-constraints.c (remove_inheritance_pseudos): Use
3695 lra_substitute_pseudo_within_insn.
3697 2018-01-19 Tom de Vries <tom@codesourcery.com>
3698 Cesar Philippidis <cesar@codesourcery.com>
3701 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3703 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3706 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3707 spaces for function labels.
3709 2018-01-19 Martin Liska <mliska@suse.cz>
3711 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3712 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3713 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3714 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3715 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3716 (PRED_CONST_RETURN): Change from 69 to 65.
3717 (PRED_NULL_RETURN): Change from 91 to 71.
3718 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3719 (PRED_LOOP_GUARD): Change from 66 to 73.
3721 2018-01-19 Martin Liska <mliska@suse.cz>
3723 * predict.c (predict_insn_def): Add new assert.
3724 (struct branch_predictor): Change type to signed integer.
3725 (test_prediction_value_range): Amend test to cover
3727 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3728 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3729 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3730 (PRED_LOOP_IV_COMPARE): Likewise.
3731 * predict.h (PROB_UNINITIALIZED): Define new constant.
3733 2018-01-19 Martin Liska <mliska@suse.cz>
3735 * predict.c (dump_prediction): Add new format for
3736 analyze_brprob.py script which is enabled with -details
3738 * profile-count.h (precise_p): New function.
3740 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3742 PR tree-optimization/83922
3743 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3744 there are no statements that need masking.
3745 (vect_active_double_reduction_p): New function.
3746 (vect_analyze_loop_operations): Use it when handling phis that
3747 are not in the loop header.
3749 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3751 PR tree-optimization/83914
3752 * tree-vect-loop.c (vectorizable_induction): Don't convert
3753 init_expr or apply the peeling adjustment for inductions
3754 that are nested within the vectorized loop.
3756 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3758 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3761 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3765 * function.h (gimplify_parameters): Add gimple_seq * argument.
3766 * function.c: Include gimple.h and options.h.
3767 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3768 for the added local temporaries if needed.
3769 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3770 if there are any parameter cleanups, wrap whole body into a
3771 try/finally with the cleanups.
3773 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3776 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3777 Use GET_MODE_CLASS for scalar floating point.
3779 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3783 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3784 Fix call of call_cgraph_insertion_hooks.
3786 2018-01-18 Martin Sebor <msebor@redhat.com>
3788 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3790 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3793 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3796 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3799 * common.opt: (-ffile-prefix-map): New option.
3800 * opts.c (common_handle_option): Defer it.
3801 * opts-global.c (handle_common_deferred_options): Handle it.
3802 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3803 * file-prefix-map.h: New file.
3804 (remap_debug_filename, add_debug_prefix_map): ...here.
3805 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3806 * final.c (debug_prefix_map, add_debug_prefix_map
3807 remap_debug_filename): Move to...
3808 * file-prefix-map.c: New file.
3809 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3810 generalize, get rid of alloca(), use strrchr() instead of strchr().
3811 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3812 Implement in terms of add_prefix_map().
3813 (remap_macro_filename, remap_debug_filename): Implement in term of
3815 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3816 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3817 * dbxout.c: Include file-prefix-map.h.
3818 * varasm.c: Likewise.
3819 * vmsdbgout.c: Likewise.
3820 * xcoffout.c: Likewise.
3821 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3822 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3823 * doc/invoke.texi (-ffile-prefix-map): Document.
3824 (-fdebug-prefix-map): Update description.
3826 2018-01-18 Martin Liska <mliska@suse.cz>
3828 * config/i386/i386.c (indirect_thunk_name): Document that also
3830 (output_indirect_thunk): Document why both instructions
3831 (pause and lfence) are generated.
3833 2018-01-18 Richard Biener <rguenther@suse.de>
3835 PR tree-optimization/83887
3836 * graphite-scop-detection.c
3837 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3838 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3839 (scop_detection::merge_sese): Re-implement with a flood-fill
3840 algorithm that properly finds a SESE region if it exists.
3842 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3845 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3846 pointer_diff optimizations use view_convert instead of convert.
3848 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3850 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3851 Generate different code for -mno-speculate-indirect-jumps.
3852 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3853 (*call_indirect_aix<mode>): Disable for
3854 -mno-speculate-indirect-jumps.
3855 (*call_indirect_aix<mode>_nospec): New define_insn.
3856 (*call_value_indirect_aix<mode>): Disable for
3857 -mno-speculate-indirect-jumps.
3858 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3859 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3860 -mno-speculate-indirect-jumps.
3861 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3863 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3865 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3866 long double type, set the flags for noting the default long double
3867 type, even if we don't pass or return a long double type.
3869 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3872 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3875 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3877 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3878 support for merge[hl].
3879 (fold_mergehl_helper): New helper function.
3880 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3881 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3882 (altivec_vmrglw_direct): Add xxmrglw insn.
3884 2018-01-17 Andrew Waterman <andrew@sifive.com>
3886 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3887 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3889 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3892 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3893 call the lto_location_cache before reading the
3894 DECL_SOURCE_LOCATION of the types.
3896 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3897 Richard Sandiford <richard.sandiford@linaro.org>
3899 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3900 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3901 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3902 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3903 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3905 * config/aarch64/constraints.md (aarch64_movti_operand):
3907 * config/aarch64/predicates.md (Uti): Add new constraint.
3909 2018-01-17 Carl Love <cel@us.ibm.com>
3911 * config/rs6000/vsx.md (define_expand xl_len_r,
3912 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3913 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3915 (define_expand, define_insn): Move the shift left from the
3916 define_insn to the define_expand for lxvl and stxvl instructions.
3917 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3918 and XL_LEN_R definitions to PURE.
3920 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3922 * config/i386/i386.c (indirect_thunk_name): Declare regno
3923 as unsigned int. Compare regno with INVALID_REGNUM.
3924 (output_indirect_thunk): Ditto.
3925 (output_indirect_thunk_function): Ditto.
3926 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3927 in the call to output_indirect_thunk_function.
3929 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3932 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3933 rather than the size of inner_type to determine the stack slot size
3934 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3936 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3939 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3942 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3944 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3945 endian Linux systems to optionally enable multilibs for selecting
3946 the long double type if the user configured an explicit type.
3947 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3948 have no long double multilibs if not defined.
3949 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3950 warn if the user used -mabi={ieee,ibm}longdouble and we built
3951 multilibs for long double.
3952 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3953 appropriate multilib option.
3954 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3956 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3957 for building long double multilibs.
3958 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3960 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3962 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3965 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3967 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3970 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3973 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3976 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3978 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3979 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3981 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3983 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3984 different rtl trees depending on TARGET_64BIT.
3985 (rs6000_gen_lvx): Likewise.
3987 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3989 * config/visium/visium.md (nop): Tweak comment.
3990 (hazard_nop): Likewise.
3992 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3994 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3995 -mspeculate-indirect-jumps.
3996 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3997 for -mno-speculate-indirect-jumps.
3998 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3999 (*call_value_indirect_elfv2<mode>): Disable for
4000 -mno-speculate-indirect-jumps.
4001 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4002 (indirect_jump): Emit different RTL for
4003 -mno-speculate-indirect-jumps.
4004 (*indirect_jump<mode>): Disable for
4005 -mno-speculate-indirect-jumps.
4006 (*indirect_jump<mode>_nospec): New define_insn.
4007 (tablejump): Emit different RTL for
4008 -mno-speculate-indirect-jumps.
4009 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4010 (tablejumpsi_nospec): New define_expand.
4011 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4012 (tablejumpdi_nospec): New define_expand.
4013 (*tablejump<mode>_internal1): Disable for
4014 -mno-speculate-indirect-jumps.
4015 (*tablejump<mode>_internal1_nospec): New define_insn.
4016 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4019 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4021 * caller-save.c (insert_save): Drop unnecessary parameter. All
4024 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4025 Richard Biener <rguenth@suse.de>
4028 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4029 return early, inline manually is_gimple_sizepos. Make sure if we
4030 call gimplify_expr we don't end up with a gimple constant.
4031 * tree.c (variably_modified_type_p): Don't return true for
4032 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4033 * gimplify.h (is_gimple_sizepos): Remove.
4035 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4037 PR tree-optimization/83857
4038 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4039 vectorizable_live_operation for pure SLP statements.
4040 (vectorizable_live_operation): Handle PHIs.
4042 2018-01-16 Richard Biener <rguenther@suse.de>
4044 PR tree-optimization/83867
4045 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4046 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4048 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4051 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4052 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4053 If off is not INTEGER_CST, issue a may not be aligned warning
4054 rather than isn't aligned. Use isn%'t rather than isn't.
4055 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4057 <case MULT_EXPR>: Improve the case when bottom and one of the
4058 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4059 operand, in that case check if the other operand is multiple of
4060 bottom divided by the INTEGER_CST operand.
4062 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4065 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4066 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4067 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4068 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4069 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4070 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4071 (pa_function_arg_size): New function.
4073 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4075 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4076 in a separate statement.
4078 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4080 PR tree-optimization/83847
4081 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4082 group gathers and scatters.
4084 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4086 PR rtl-optimization/86620
4087 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4089 PR rtl-optimization/83213
4090 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4091 to last if both are JUMP_INSNs.
4093 PR tree-optimization/83843
4094 * gimple-ssa-store-merging.c
4095 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4096 store_immediate_info for bswap/nop orig_stores.
4098 2018-01-15 Andrew Waterman <andrew@sifive.com>
4100 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4102 <UDIV>: Increase cost if !TARGET_DIV.
4104 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4106 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4107 (define_attr "cr_logical_3op"): New.
4108 (cceq_ior_compare): Adjust.
4109 (cceq_ior_compare_complement): Adjust.
4110 (*cceq_rev_compare): Adjust.
4111 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4112 (is_cracked_insn): Adjust.
4113 (insn_must_be_first_in_group): Adjust.
4114 * config/rs6000/40x.md: Adjust.
4115 * config/rs6000/440.md: Adjust.
4116 * config/rs6000/476.md: Adjust.
4117 * config/rs6000/601.md: Adjust.
4118 * config/rs6000/603.md: Adjust.
4119 * config/rs6000/6xx.md: Adjust.
4120 * config/rs6000/7450.md: Adjust.
4121 * config/rs6000/7xx.md: Adjust.
4122 * config/rs6000/8540.md: Adjust.
4123 * config/rs6000/cell.md: Adjust.
4124 * config/rs6000/e300c2c3.md: Adjust.
4125 * config/rs6000/e500mc.md: Adjust.
4126 * config/rs6000/e500mc64.md: Adjust.
4127 * config/rs6000/e5500.md: Adjust.
4128 * config/rs6000/e6500.md: Adjust.
4129 * config/rs6000/mpc.md: Adjust.
4130 * config/rs6000/power4.md: Adjust.
4131 * config/rs6000/power5.md: Adjust.
4132 * config/rs6000/power6.md: Adjust.
4133 * config/rs6000/power7.md: Adjust.
4134 * config/rs6000/power8.md: Adjust.
4135 * config/rs6000/power9.md: Adjust.
4136 * config/rs6000/rs64.md: Adjust.
4137 * config/rs6000/titan.md: Adjust.
4139 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4141 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4142 ix86_indirect_branch_register logic.
4144 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4146 * config/i386/constraints.md (Bs): Update
4147 ix86_indirect_branch_register check. Don't check
4148 ix86_indirect_branch_register with GOT_memory_operand.
4150 * config/i386/predicates.md (GOT_memory_operand): Don't check
4151 ix86_indirect_branch_register here.
4152 (GOT32_symbol_operand): Likewise.
4154 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4156 * config/i386/predicates.md (constant_call_address_operand):
4157 Rewrite ix86_indirect_branch_register logic.
4158 (sibcall_insn_operand): Likewise.
4160 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4162 * config/i386/constraints.md (Bs): Replace
4163 ix86_indirect_branch_thunk_register with
4164 ix86_indirect_branch_register.
4166 * config/i386/i386.md (indirect_jump): Likewise.
4167 (tablejump): Likewise.
4168 (*sibcall_memory): Likewise.
4169 (*sibcall_value_memory): Likewise.
4170 Peepholes of indirect call and jump via memory: Likewise.
4171 * config/i386/i386.opt: Likewise.
4172 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4173 (GOT_memory_operand): Likewise.
4174 (call_insn_operand): Likewise.
4175 (sibcall_insn_operand): Likewise.
4176 (GOT32_symbol_operand): Likewise.
4178 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4181 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4182 type rather than type addr's type points to.
4183 (expand_omp_atomic_mutex): Likewise.
4184 (expand_omp_atomic): Likewise.
4186 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4189 * config/i386/i386.c (output_indirect_thunk_function): Use
4190 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4191 for __x86_return_thunk.
4193 2018-01-15 Richard Biener <rguenther@suse.de>
4196 * expmed.c (extract_bit_field_1): Fix typo.
4198 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4201 * config/arm/iterators.md (VF): New mode iterator.
4202 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4203 Remove integer-related logic from pattern.
4204 (neon_vabd<mode>_3): Likewise.
4206 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4209 * common.opt (fstrict-overflow): No longer an alias.
4210 (fwrapv-pointer): New option.
4211 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4212 also for pointer types based on flag_wrapv_pointer.
4213 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4214 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4215 opts->x_flag_wrapv got set.
4216 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4217 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4218 POINTER_TYPE_OVERFLOW_UNDEFINED.
4219 * match.pd: Likewise in address comparison pattern.
4220 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4222 2018-01-15 Richard Biener <rguenther@suse.de>
4225 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4226 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4227 Reset type names to their identifier if their TYPE_DECL doesn't
4228 have linkage (and thus is used for ODR and devirt).
4229 (save_debug_info_for_decl): Remove.
4230 (save_debug_info_for_type): Likewise.
4231 (add_tree_to_fld_list): Adjust.
4232 * tree-pretty-print.c (dump_generic_node): Make dumping of
4233 type names more robust.
4235 2018-01-15 Richard Biener <rguenther@suse.de>
4237 * BASE-VER: Bump to 8.0.1.
4239 2018-01-14 Martin Sebor <msebor@redhat.com>
4242 * builtins.c (check_access): Avoid warning when the no-warning bit
4245 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4247 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4248 * ira-color (allocno_hard_regs_compare): Likewise.
4250 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4253 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4254 Use .pushsection/.popsection.
4256 2018-01-14 Martin Sebor <msebor@redhat.com>
4259 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4261 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4263 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4264 entry from extra_headers.
4265 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4266 extra_headers, make the list bitwise identical to the i?86-*-* one.
4268 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4270 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4271 -mcmodel=large with -mindirect-branch=thunk,
4272 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4273 -mfunction-return=thunk-extern.
4274 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4275 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4276 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4278 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4280 * config/i386/i386.c (print_reg): Print the name of the full
4281 integer register without '%'.
4282 (ix86_print_operand): Handle 'V'.
4283 * doc/extend.texi: Document 'V' modifier.
4285 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4287 * config/i386/constraints.md (Bs): Disallow memory operand for
4288 -mindirect-branch-register.
4290 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4291 (GOT_memory_operand): Likewise.
4292 (call_insn_operand): Likewise.
4293 (sibcall_insn_operand): Likewise.
4294 (GOT32_symbol_operand): Likewise.
4295 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4296 for -mindirect-branch-register.
4297 (tablejump): Likewise.
4298 (*sibcall_memory): Likewise.
4299 (*sibcall_value_memory): Likewise.
4300 Disallow peepholes of indirect call and jump via memory for
4301 -mindirect-branch-register.
4302 (*call_pop): Replace m with Bw.
4303 (*call_value_pop): Likewise.
4304 (*sibcall_pop_memory): Replace m with Bs.
4305 * config/i386/i386.opt (mindirect-branch-register): New option.
4306 * doc/invoke.texi: Document -mindirect-branch-register option.
4308 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4310 * config/i386/i386-protos.h (ix86_output_function_return): New.
4311 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4312 set function_return_type.
4313 (indirect_thunk_name): Add ret_p to indicate thunk for function
4315 (output_indirect_thunk_function): Pass false to
4316 indirect_thunk_name.
4317 (ix86_output_indirect_branch_via_reg): Likewise.
4318 (ix86_output_indirect_branch_via_push): Likewise.
4319 (output_indirect_thunk_function): Create alias for function
4320 return thunk if regno < 0.
4321 (ix86_output_function_return): New function.
4322 (ix86_handle_fndecl_attribute): Handle function_return.
4323 (ix86_attribute_table): Add function_return.
4324 * config/i386/i386.h (machine_function): Add
4325 function_return_type.
4326 * config/i386/i386.md (simple_return_internal): Use
4327 ix86_output_function_return.
4328 (simple_return_internal_long): Likewise.
4329 * config/i386/i386.opt (mfunction-return=): New option.
4330 (indirect_branch): Mention -mfunction-return=.
4331 * doc/extend.texi: Document function_return function attribute.
4332 * doc/invoke.texi: Document -mfunction-return= option.
4334 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4336 * config/i386/i386-opts.h (indirect_branch): New.
4337 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4338 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4339 with local indirect jump when converting indirect call and jump.
4340 (ix86_set_indirect_branch_type): New.
4341 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4342 (indirectlabelno): New.
4343 (indirect_thunk_needed): Likewise.
4344 (indirect_thunk_bnd_needed): Likewise.
4345 (indirect_thunks_used): Likewise.
4346 (indirect_thunks_bnd_used): Likewise.
4347 (INDIRECT_LABEL): Likewise.
4348 (indirect_thunk_name): Likewise.
4349 (output_indirect_thunk): Likewise.
4350 (output_indirect_thunk_function): Likewise.
4351 (ix86_output_indirect_branch_via_reg): Likewise.
4352 (ix86_output_indirect_branch_via_push): Likewise.
4353 (ix86_output_indirect_branch): Likewise.
4354 (ix86_output_indirect_jmp): Likewise.
4355 (ix86_code_end): Call output_indirect_thunk_function if needed.
4356 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4358 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4359 (ix86_attribute_table): Add indirect_branch.
4360 * config/i386/i386.h (machine_function): Add indirect_branch_type
4361 and has_local_indirect_jump.
4362 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4364 (tablejump): Likewise.
4365 (*indirect_jump): Use ix86_output_indirect_jmp.
4366 (*tablejump_1): Likewise.
4367 (simple_return_indirect_internal): Likewise.
4368 * config/i386/i386.opt (mindirect-branch=): New option.
4369 (indirect_branch): New.
4372 (thunk-inline): Likewise.
4373 (thunk-extern): Likewise.
4374 * doc/extend.texi: Document indirect_branch function attribute.
4375 * doc/invoke.texi: Document -mindirect-branch= option.
4377 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4380 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4382 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4384 * ipa-inline.c (want_inline_small_function_p): Return false if
4385 inlining has already failed with CIF_FINAL_ERROR.
4386 (update_caller_keys): Call want_inline_small_function_p before
4388 (update_callee_keys): Likewise.
4390 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4392 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4394 (rs6000_quadword_masked_address_p): Likewise.
4395 (quad_aligned_load_p): Likewise.
4396 (quad_aligned_store_p): Likewise.
4397 (const_load_sequence_p): Add comment to describe the outer-most loop.
4398 (mimic_memory_attributes_and_flags): New function.
4399 (rs6000_gen_stvx): Likewise.
4400 (replace_swapped_aligned_store): Likewise.
4401 (rs6000_gen_lvx): Likewise.
4402 (replace_swapped_aligned_load): Likewise.
4403 (replace_swapped_load_constant): Capitalize argument name in
4404 comment describing this function.
4405 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4406 and stores that access quad-word aligned addresses and replace
4407 with stvx or lvx instructions when appropriate.
4408 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4409 New function prototype.
4410 (rs6000_quadword_masked_address_p): Likewise.
4411 (rs6000_gen_lvx): Likewise.
4412 (rs6000_gen_stvx): Likewise.
4413 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4414 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4415 when memory address is aligned.
4416 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4417 this split to select lvx instruction when memory address is aligned.
4418 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4419 instruction when memory address is aligned.
4420 (*vsx_le_perm_load_v16qi): Likewise.
4421 (four unnamed splitters): Modify to select the stvx instruction
4422 when memory is aligned.
4424 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4426 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4427 which appears in the queue multiple times.
4429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4430 Alan Hayward <alan.hayward@arm.com>
4431 David Sherwood <david.sherwood@arm.com>
4433 * tree-vectorizer.h (vec_lower_bound): New structure.
4434 (_loop_vec_info): Add check_nonzero and lower_bounds.
4435 (LOOP_VINFO_CHECK_NONZERO): New macro.
4436 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4437 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4438 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4439 fields. Make seg_len the distance travelled, not including the
4441 (dr_direction_indicator): Declare.
4442 (dr_zero_step_indicator): Likewise.
4443 (dr_known_forward_stride_p): Likewise.
4444 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4446 (runtime_alias_check_p): Allow runtime alias checks with
4448 (operator ==): Compare access_size and align.
4449 (prune_runtime_alias_test_list): Rework for new distinction between
4450 the access_size and seg_len.
4451 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4453 (get_segment_min_max): New function.
4454 (create_intersect_range_checks): Use it.
4455 (dr_step_indicator): New function.
4456 (dr_direction_indicator): Likewise.
4457 (dr_zero_step_indicator): Likewise.
4458 (dr_known_forward_stride_p): Likewise.
4459 * tree-loop-distribution.c (data_ref_segment_size): Return
4460 DR_STEP * (niters - 1).
4461 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4463 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4464 (vect_preserves_scalar_order_p): New function, split out from...
4465 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4466 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4467 (vect_vfa_access_size): New function.
4468 (vect_vfa_align): Likewise.
4469 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4470 (dump_lower_bound): New function.
4471 (vect_check_lower_bound): Likewise.
4472 (vect_small_gap_p): Likewise.
4473 (vectorizable_with_step_bound_p): Likewise.
4474 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4475 depencies if the vectorization factor is 1. Convert the checks
4476 for nonzero steps into checks on the bounds of DR_STEP. Try using
4477 a bunds check for variable steps if the minimum required step is
4478 relatively small. Update calls to the dr_with_seg_len
4479 constructor and to vect_compile_time_alias.
4480 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4482 (vect_loop_versioning): Call it.
4483 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4485 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4488 Alan Hayward <alan.hayward@arm.com>
4489 David Sherwood <david.sherwood@arm.com>
4491 * doc/sourcebuild.texi (vect_scatter_store): Document.
4492 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4494 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4496 * genopinit.c (main): Add supports_vec_scatter_store and
4497 supports_vec_scatter_store_cached to target_optabs.
4498 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4499 IFN_MASK_SCATTER_STORE.
4500 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4502 * internal-fn.h (internal_store_fn_p): Declare.
4503 (internal_fn_stored_value_index): Likewise.
4504 * internal-fn.c (scatter_store_direct): New macro.
4505 (expand_scatter_store_optab_fn): New function.
4506 (direct_scatter_store_optab_supported_p): New macro.
4507 (internal_store_fn_p): New function.
4508 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4509 IFN_MASK_SCATTER_STORE.
4510 (internal_fn_mask_index): Likewise.
4511 (internal_fn_stored_value_index): New function.
4512 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4514 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4515 * optabs-query.c (supports_vec_scatter_store_p): New function.
4516 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4517 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4518 true for scatter stores.
4519 (vect_gather_scatter_fn_p): Handle scatter stores too.
4520 (vect_check_gather_scatter): Consider using scatter stores if
4521 supports_vec_scatter_store_p.
4522 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4524 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4525 internal_fn_stored_value_index.
4526 (check_load_store_masking): Handle scatter stores too.
4527 (vect_get_store_rhs): Make public.
4528 (vectorizable_call): Use internal_store_fn_p.
4529 (vectorizable_store): Handle scatter store internal functions.
4530 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4531 when deciding whether the end of the group has been reached.
4532 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4533 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4534 (mask_scatter_store<mode>): New insns.
4536 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4537 Alan Hayward <alan.hayward@arm.com>
4538 David Sherwood <david.sherwood@arm.com>
4540 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4541 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4542 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4544 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4545 Use vect_truncate_gather_scatter_offset if we can't treat the
4546 operation as a normal gather load or scatter store.
4547 (get_group_load_store_type): Take the gather_scatter_info
4548 as argument. Try using a gather load or scatter store for
4549 single-element groups.
4550 (get_load_store_type): Update calls to get_group_load_store_type
4551 and vect_use_strided_gather_scatters_p.
4553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4554 Alan Hayward <alan.hayward@arm.com>
4555 David Sherwood <david.sherwood@arm.com>
4557 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4558 optional tree argument.
4559 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4561 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4562 but continue to use the current value as a fallback.
4563 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4564 to compare the updates.
4565 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4566 (get_load_store_type): Use it when handling a strided access.
4567 (vect_get_strided_load_store_ops): New function.
4568 (vect_get_data_ptr_increment): Likewise.
4569 (vectorizable_load): Handle strided gather loads. Always pass
4570 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4573 Alan Hayward <alan.hayward@arm.com>
4574 David Sherwood <david.sherwood@arm.com>
4576 * doc/md.texi (gather_load@var{m}): Document.
4577 (mask_gather_load@var{m}): Likewise.
4578 * genopinit.c (main): Add supports_vec_gather_load and
4579 supports_vec_gather_load_cached to target_optabs.
4580 * optabs-tree.c (init_tree_optimization_optabs): Use
4581 ggc_cleared_alloc to allocate target_optabs.
4582 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4583 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4585 * internal-fn.h (internal_load_fn_p): Declare.
4586 (internal_gather_scatter_fn_p): Likewise.
4587 (internal_fn_mask_index): Likewise.
4588 (internal_gather_scatter_fn_supported_p): Likewise.
4589 * internal-fn.c (gather_load_direct): New macro.
4590 (expand_gather_load_optab_fn): New function.
4591 (direct_gather_load_optab_supported_p): New macro.
4592 (direct_internal_fn_optab): New function.
4593 (internal_load_fn_p): Likewise.
4594 (internal_gather_scatter_fn_p): Likewise.
4595 (internal_fn_mask_index): Likewise.
4596 (internal_gather_scatter_fn_supported_p): Likewise.
4597 * optabs-query.c (supports_at_least_one_mode_p): New function.
4598 (supports_vec_gather_load_p): Likewise.
4599 * optabs-query.h (supports_vec_gather_load_p): Declare.
4600 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4601 and memory_type field.
4602 (NUM_PATTERNS): Bump to 15.
4603 * tree-vect-data-refs.c: Include internal-fn.h.
4604 (vect_gather_scatter_fn_p): New function.
4605 (vect_describe_gather_scatter_call): Likewise.
4606 (vect_check_gather_scatter): Try using internal functions for
4607 gather loads. Recognize existing calls to a gather load function.
4608 (vect_analyze_data_refs): Consider using gather loads if
4609 supports_vec_gather_load_p.
4610 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4611 (vect_get_gather_scatter_offset_type): Likewise.
4612 (vect_convert_mask_for_vectype): Likewise.
4613 (vect_add_conversion_to_patterm): Likewise.
4614 (vect_try_gather_scatter_pattern): Likewise.
4615 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4616 (vect_vect_recog_func_ptrs): Add it.
4617 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4618 internal_fn_mask_index and internal_gather_scatter_fn_p.
4619 (check_load_store_masking): Take the gather_scatter_info as an
4620 argument and handle gather loads.
4621 (vect_get_gather_scatter_ops): New function.
4622 (vectorizable_call): Check internal_load_fn_p.
4623 (vectorizable_load): Likewise. Handle gather load internal
4625 (vectorizable_store): Update call to check_load_store_masking.
4626 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4627 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4628 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4629 (aarch64_gather_scale_operand_d): New predicates.
4630 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4631 (mask_gather_load<mode>): New insns.
4633 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4634 Alan Hayward <alan.hayward@arm.com>
4635 David Sherwood <david.sherwood@arm.com>
4637 * optabs.def (fold_left_plus_optab): New optab.
4638 * doc/md.texi (fold_left_plus_@var{m}): Document.
4639 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4640 * internal-fn.c (fold_left_direct): Define.
4641 (expand_fold_left_optab_fn): Likewise.
4642 (direct_fold_left_optab_supported_p): Likewise.
4643 * fold-const-call.c (fold_const_fold_left): New function.
4644 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4645 * tree-parloops.c (valid_reduction_p): New function.
4646 (gather_scalar_reductions): Use it.
4647 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4648 (vect_finish_replace_stmt): Declare.
4649 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4650 (needs_fold_left_reduction_p): New function, split out from...
4651 (vect_is_simple_reduction): ...here. Accept reductions that
4652 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4653 (vect_force_simple_reduction): Also store the reduction type in
4654 the assignment's STMT_VINFO_REDUC_TYPE.
4655 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4656 (merge_with_identity): New function.
4657 (vect_expand_fold_left): Likewise.
4658 (vectorize_fold_left_reduction): Likewise.
4659 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4660 scalar phi in place for it. Check for target support and reject
4661 cases that would reassociate the operation. Defer the transform
4662 phase to vectorize_fold_left_reduction.
4663 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4664 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4665 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4667 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4669 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4670 call to ifc_temp_var.
4672 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4673 Alan Hayward <alan.hayward@arm.com>
4674 David Sherwood <david.sherwood@arm.com>
4676 * target.def (legitimize_address_displacement): Take the original
4677 offset as a poly_int.
4678 * targhooks.h (default_legitimize_address_displacement): Update
4680 * targhooks.c (default_legitimize_address_displacement): Likewise.
4681 * doc/tm.texi: Regenerate.
4682 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4683 as an argument, moving assert of ad->disp == ad->disp_term to...
4684 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4685 Try calling targetm.legitimize_address_displacement before expanding
4686 the address rather than afterwards, and adjust for the new interface.
4687 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4688 Match the new hook interface. Handle SVE addresses.
4689 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4692 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4694 * Makefile.in (OBJS): Add early-remat.o.
4695 * target.def (select_early_remat_modes): New hook.
4696 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4697 * doc/tm.texi: Regenerate.
4698 * targhooks.h (default_select_early_remat_modes): Declare.
4699 * targhooks.c (default_select_early_remat_modes): New function.
4700 * timevar.def (TV_EARLY_REMAT): New timevar.
4701 * passes.def (pass_early_remat): New pass.
4702 * tree-pass.h (make_pass_early_remat): Declare.
4703 * early-remat.c: New file.
4704 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4706 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4709 Alan Hayward <alan.hayward@arm.com>
4710 David Sherwood <david.sherwood@arm.com>
4712 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4713 vfm1 with a bound_epilog parameter.
4714 (vect_do_peeling): Update calls accordingly, and move the prologue
4715 call earlier in the function. Treat the base bound_epilog as 0 for
4716 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4717 this base when peeling for gaps.
4718 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4719 with fully-masked loops.
4720 (vect_estimate_min_profitable_iters): Handle the single peeled
4721 iteration in that case.
4723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4724 Alan Hayward <alan.hayward@arm.com>
4725 David Sherwood <david.sherwood@arm.com>
4727 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4728 single-element interleaving even if the size is not a power of 2.
4729 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4730 accesses for single-element interleaving if the group size is
4733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4734 Alan Hayward <alan.hayward@arm.com>
4735 David Sherwood <david.sherwood@arm.com>
4737 * doc/md.texi (fold_extract_last_@var{m}): Document.
4738 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4739 * optabs.def (fold_extract_last_optab): New optab.
4740 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4741 * internal-fn.c (fold_extract_direct): New macro.
4742 (expand_fold_extract_optab_fn): Likewise.
4743 (direct_fold_extract_optab_supported_p): Likewise.
4744 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4745 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4746 EXTRACT_LAST_REDUCTION.
4747 (get_initial_def_for_reduction): Do not create an initial vector
4748 for EXTRACT_LAST_REDUCTION reductions.
4749 (vectorizable_reduction): Leave the scalar phi in place for
4750 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4751 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4752 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4753 transform phase to vectorizable_condition.
4754 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4756 (vect_finish_stmt_generation): ...here.
4757 (vect_finish_replace_stmt): New function.
4758 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4759 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4761 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4764 Alan Hayward <alan.hayward@arm.com>
4765 David Sherwood <david.sherwood@arm.com>
4767 * doc/md.texi (extract_last_@var{m}): Document.
4768 * optabs.def (extract_last_optab): New optab.
4769 * internal-fn.def (EXTRACT_LAST): New internal function.
4770 * internal-fn.c (cond_unary_direct): New macro.
4771 (expand_cond_unary_optab_fn): Likewise.
4772 (direct_cond_unary_optab_supported_p): Likewise.
4773 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4774 loops using EXTRACT_LAST.
4775 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4776 (extract_last_<mode>): ...this optab.
4777 (vec_extract<mode><Vel>): Update accordingly.
4779 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4780 Alan Hayward <alan.hayward@arm.com>
4781 David Sherwood <david.sherwood@arm.com>
4783 * target.def (empty_mask_is_expensive): New hook.
4784 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4785 * doc/tm.texi: Regenerate.
4786 * targhooks.h (default_empty_mask_is_expensive): Declare.
4787 * targhooks.c (default_empty_mask_is_expensive): New function.
4788 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4789 if the target says that empty masks are expensive.
4790 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4792 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4795 Alan Hayward <alan.hayward@arm.com>
4796 David Sherwood <david.sherwood@arm.com>
4798 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4799 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4800 (vect_use_loop_mask_for_alignment_p): New function.
4801 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4802 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4803 niters_skip argument. Make sure that the first niters_skip elements
4804 of the first iteration are inactive.
4805 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4806 Update call to vect_set_loop_masks_directly.
4807 (get_misalign_in_elems): New function, split out from...
4808 (vect_gen_prolog_loop_niters): ...here.
4809 (vect_update_init_of_dr): Take a code argument that specifies whether
4810 the adjustment should be added or subtracted.
4811 (vect_update_init_of_drs): Likewise.
4812 (vect_prepare_for_masked_peels): New function.
4813 (vect_do_peeling): Skip prologue peeling if we're using a mask
4814 instead. Update call to vect_update_inits_of_drs.
4815 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4817 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4818 alignment. Do not include the number of peeled iterations in
4819 the minimum threshold in that case.
4820 (vectorizable_induction): Adjust the start value down by
4821 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4822 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4823 Take the number of skipped iterations into account when calculating
4825 * tree-vect-stmts.c (vect_gen_while_not): New function.
4827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4828 Alan Hayward <alan.hayward@arm.com>
4829 David Sherwood <david.sherwood@arm.com>
4831 * doc/sourcebuild.texi (vect_fully_masked): Document.
4832 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4834 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4836 (vect_analyze_loop_2): ...here. Don't check the vectorization
4837 factor against the number of loop iterations if the loop is
4840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4841 Alan Hayward <alan.hayward@arm.com>
4842 David Sherwood <david.sherwood@arm.com>
4844 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4845 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4846 (dump_groups): Update accordingly.
4847 (iv_use::mem_type): New member variable.
4848 (address_p): New function.
4849 (record_use): Add a mem_type argument and initialize the new
4851 (record_group_use): Add a mem_type argument. Use address_p.
4852 Remove obsolete null checks of base_object. Update call to record_use.
4853 (find_interesting_uses_op): Update call to record_group_use.
4854 (find_interesting_uses_cond): Likewise.
4855 (find_interesting_uses_address): Likewise.
4856 (get_mem_type_for_internal_fn): New function.
4857 (find_address_like_use): Likewise.
4858 (find_interesting_uses_stmt): Try find_address_like_use before
4859 calling find_interesting_uses_op.
4860 (addr_offset_valid_p): Use the iv mem_type field as the type
4861 of the addressed memory.
4862 (add_autoinc_candidates): Likewise.
4863 (get_address_cost): Likewise.
4864 (split_small_address_groups_p): Use address_p.
4865 (split_address_groups): Likewise.
4866 (add_iv_candidate_for_use): Likewise.
4867 (autoinc_possible_for_pair): Likewise.
4868 (rewrite_groups): Likewise.
4869 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4870 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4871 (get_alias_ptr_type_for_ptr_address): New function.
4872 (rewrite_use_address): Rewrite address uses in calls that were
4873 identified by find_address_like_use.
4875 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4876 Alan Hayward <alan.hayward@arm.com>
4877 David Sherwood <david.sherwood@arm.com>
4879 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4881 * gimple-expr.h (is_gimple_addressable: Likewise.
4882 * gimple-expr.c (is_gimple_address): Likewise.
4883 * internal-fn.c (expand_call_mem_ref): New function.
4884 (expand_mask_load_optab_fn): Use it.
4885 (expand_mask_store_optab_fn): Likewise.
4887 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4888 Alan Hayward <alan.hayward@arm.com>
4889 David Sherwood <david.sherwood@arm.com>
4891 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4892 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4893 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4894 (cond_umax@var{mode}): Document.
4895 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4896 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4897 (cond_umin_optab, cond_umax_optab): New optabs.
4898 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4899 (COND_IOR, COND_XOR): New internal functions.
4900 * internal-fn.h (get_conditional_internal_fn): Declare.
4901 * internal-fn.c (cond_binary_direct): New macro.
4902 (expand_cond_binary_optab_fn): Likewise.
4903 (direct_cond_binary_optab_supported_p): Likewise.
4904 (get_conditional_internal_fn): New function.
4905 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4906 Cope with reduction statements that are vectorized as calls rather
4908 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4909 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4910 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4911 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4912 (UNSPEC_COND_EOR): New unspecs.
4913 (optab): Add mappings for them.
4914 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4915 (sve_int_op, sve_fp_op): New int attributes.
4917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4918 Alan Hayward <alan.hayward@arm.com>
4919 David Sherwood <david.sherwood@arm.com>
4921 * optabs.def (while_ult_optab): New optab.
4922 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4923 * internal-fn.def (WHILE_ULT): New internal function.
4924 * internal-fn.h (direct_internal_fn_supported_p): New override
4925 that takes two types as argument.
4926 * internal-fn.c (while_direct): New macro.
4927 (expand_while_optab_fn): New function.
4928 (convert_optab_supported_p): Likewise.
4929 (direct_while_optab_supported_p): New macro.
4930 * wide-int.h (wi::udiv_ceil): New function.
4931 * tree-vectorizer.h (rgroup_masks): New structure.
4932 (vec_loop_masks): New typedef.
4933 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4935 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4936 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4937 (vect_max_vf): New function.
4938 (slpeel_make_loop_iterate_ntimes): Delete.
4939 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4940 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4941 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4942 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4943 internal-fn.h, stor-layout.h and optabs-query.h.
4944 (vect_set_loop_mask): New function.
4945 (add_preheader_seq): Likewise.
4946 (add_header_seq): Likewise.
4947 (interleave_supported_p): Likewise.
4948 (vect_maybe_permute_loop_masks): Likewise.
4949 (vect_set_loop_masks_directly): Likewise.
4950 (vect_set_loop_condition_masked): Likewise.
4951 (vect_set_loop_condition_unmasked): New function, split out from
4952 slpeel_make_loop_iterate_ntimes.
4953 (slpeel_make_loop_iterate_ntimes): Rename to..
4954 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4955 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4956 (vect_do_peeling): Update call accordingly.
4957 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4959 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4960 mask_compare_type, can_fully_mask_p and fully_masked_p.
4961 (release_vec_loop_masks): New function.
4962 (_loop_vec_info): Use it to free the loop masks.
4963 (can_produce_all_loop_masks_p): New function.
4964 (vect_get_max_nscalars_per_iter): Likewise.
4965 (vect_verify_full_masking): Likewise.
4966 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4967 retries, and free the mask rgroups before retrying. Check loop-wide
4968 reasons for disallowing fully-masked loops. Make the final decision
4969 about whether use a fully-masked loop or not.
4970 (vect_estimate_min_profitable_iters): Do not assume that peeling
4971 for the number of iterations will be needed for fully-masked loops.
4972 (vectorizable_reduction): Disable fully-masked loops.
4973 (vectorizable_live_operation): Likewise.
4974 (vect_halve_mask_nunits): New function.
4975 (vect_double_mask_nunits): Likewise.
4976 (vect_record_loop_mask): Likewise.
4977 (vect_get_loop_mask): Likewise.
4978 (vect_transform_loop): Handle the case in which the final loop
4979 iteration might handle a partial vector. Call vect_set_loop_condition
4980 instead of slpeel_make_loop_iterate_ntimes.
4981 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4982 (check_load_store_masking): New function.
4983 (prepare_load_store_mask): Likewise.
4984 (vectorizable_store): Handle fully-masked loops.
4985 (vectorizable_load): Likewise.
4986 (supportable_widening_operation): Use vect_halve_mask_nunits for
4988 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4989 (vect_gen_while): New function.
4990 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4991 (aarch64_uqdec<mode>): New insn.
4993 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4994 Alan Hayward <alan.hayward@arm.com>
4995 David Sherwood <david.sherwood@arm.com>
4997 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4998 (reduc_xor_scal_optab): New optabs.
4999 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5000 (reduc_xor_scal_@var{m}): Document.
5001 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5002 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5004 * fold-const-call.c (fold_const_call): Handle them.
5005 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5006 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5007 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5008 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5009 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5010 (UNSPEC_XORV): New unspecs.
5011 (optab): Add entries for them.
5012 (BITWISEV): New int iterator.
5013 (bit_reduc_op): New int attributes.
5015 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5016 Alan Hayward <alan.hayward@arm.com>
5017 David Sherwood <david.sherwood@arm.com>
5019 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5020 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5021 * optabs.def (vec_shl_insert_optab): New optab.
5022 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5023 (duplicate_and_interleave): Likewise.
5024 * tree-vect-loop.c: Include internal-fn.h.
5025 (neutral_op_for_slp_reduction): New function, split out from
5026 get_initial_defs_for_reduction.
5027 (get_initial_def_for_reduction): Handle option 2 for variable-length
5028 vectors by loading the neutral value into a vector and then shifting
5029 the initial value into element 0.
5030 (get_initial_defs_for_reduction): Replace the code argument with
5031 the neutral value calculated by neutral_op_for_slp_reduction.
5032 Use gimple_build_vector for constant-length vectors.
5033 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5034 but the first group_size elements have a neutral value.
5035 Use duplicate_and_interleave otherwise.
5036 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5037 Update call to get_initial_defs_for_reduction. Handle SLP
5038 reductions for variable-length vectors by creating one vector
5039 result for each scalar result, with the elements associated
5040 with other scalar results stubbed out with the neutral value.
5041 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5042 Require IFN_VEC_SHL_INSERT for double reductions on
5043 variable-length vectors, or SLP reductions that have
5044 a neutral value. Require can_duplicate_and_interleave_p
5045 support for variable-length unchained SLP reductions if there
5046 is no neutral value, such as for MIN/MAX reductions. Also require
5047 the number of vector elements to be a multiple of the number of
5048 SLP statements when doing variable-length unchained SLP reductions.
5049 Update call to vect_create_epilog_for_reduction.
5050 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5051 and remove initial values.
5052 (duplicate_and_interleave): Make public.
5053 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5054 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5057 Alan Hayward <alan.hayward@arm.com>
5058 David Sherwood <david.sherwood@arm.com>
5060 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5061 (can_duplicate_and_interleave_p): New function.
5062 (vect_get_and_check_slp_defs): Take the vector of statements
5063 rather than just the current one. Remove excess parentheses.
5064 Restriction rejectinon of vect_constant_def and vect_external_def
5065 for variable-length vectors to boolean types, or types for which
5066 can_duplicate_and_interleave_p is false.
5067 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5068 (duplicate_and_interleave): New function.
5069 (vect_get_constant_vectors): Use gimple_build_vector for
5070 constant-length vectors and suitable variable-length constant
5071 vectors. Use duplicate_and_interleave for other variable-length
5072 vectors. Don't defer the update when inserting new statements.
5074 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5075 Alan Hayward <alan.hayward@arm.com>
5076 David Sherwood <david.sherwood@arm.com>
5078 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5079 min_profitable_iters doesn't go negative.
5081 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5082 Alan Hayward <alan.hayward@arm.com>
5083 David Sherwood <david.sherwood@arm.com>
5085 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5086 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5087 * optabs.def (vec_mask_load_lanes_optab): New optab.
5088 (vec_mask_store_lanes_optab): Likewise.
5089 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5090 (MASK_STORE_LANES): Likewise.
5091 * internal-fn.c (mask_load_lanes_direct): New macro.
5092 (mask_store_lanes_direct): Likewise.
5093 (expand_mask_load_optab_fn): Handle masked operations.
5094 (expand_mask_load_lanes_optab_fn): New macro.
5095 (expand_mask_store_optab_fn): Handle masked operations.
5096 (expand_mask_store_lanes_optab_fn): New macro.
5097 (direct_mask_load_lanes_optab_supported_p): Likewise.
5098 (direct_mask_store_lanes_optab_supported_p): Likewise.
5099 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5101 (vect_load_lanes_supported): Likewise.
5102 * tree-vect-data-refs.c (strip_conversion): New function.
5103 (can_group_stmts_p): Likewise.
5104 (vect_analyze_data_ref_accesses): Use it instead of checking
5105 for a pair of assignments.
5106 (vect_store_lanes_supported): Take a masked_p parameter.
5107 (vect_load_lanes_supported): Likewise.
5108 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5109 vect_store_lanes_supported and vect_load_lanes_supported.
5110 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5111 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5112 parameter. Don't allow gaps for masked accesses.
5113 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5114 and vect_load_lanes_supported.
5115 (get_load_store_type): Take a masked_p parameter and update
5116 call to get_group_load_store_type.
5117 (vectorizable_store): Update call to get_load_store_type.
5118 Handle IFN_MASK_STORE_LANES.
5119 (vectorizable_load): Update call to get_load_store_type.
5120 Handle IFN_MASK_LOAD_LANES.
5122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5123 Alan Hayward <alan.hayward@arm.com>
5124 David Sherwood <david.sherwood@arm.com>
5126 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5128 * config/aarch64/aarch64-protos.h
5129 (aarch64_sve_struct_memory_operand_p): Declare.
5130 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5131 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5132 (VPRED, vpred): Handle SVE structure modes.
5133 * config/aarch64/constraints.md (Utx): New constraint.
5134 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5135 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5136 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5137 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5138 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5139 structure modes. Split into pieces after RA.
5140 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5141 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5143 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5144 SVE structure modes.
5145 (aarch64_classify_address): Likewise.
5146 (sizetochar): Move earlier in file.
5147 (aarch64_print_operand): Handle SVE register lists.
5148 (aarch64_array_mode): New function.
5149 (aarch64_sve_struct_memory_operand_p): Likewise.
5150 (TARGET_ARRAY_MODE): Redefine.
5152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5153 Alan Hayward <alan.hayward@arm.com>
5154 David Sherwood <david.sherwood@arm.com>
5156 * target.def (array_mode): New target hook.
5157 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5158 * doc/tm.texi: Regenerate.
5159 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5160 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5161 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5163 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5166 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5167 Alan Hayward <alan.hayward@arm.com>
5168 David Sherwood <david.sherwood@arm.com>
5170 * fold-const.c (fold_binary_loc): Check the argument types
5171 rather than the result type when testing for a vector operation.
5173 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5175 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5176 * doc/tm.texi: Regenerate.
5178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5179 Alan Hayward <alan.hayward@arm.com>
5180 David Sherwood <david.sherwood@arm.com>
5182 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5183 (sve): Document new AArch64 extension.
5184 * doc/md.texi (w): Extend the description of the AArch64
5185 constraint to include SVE vectors.
5186 (Upl, Upa): Document new AArch64 predicate constraints.
5187 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5189 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5190 (msve-vector-bits=): New option.
5191 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5192 SVE when these are disabled.
5193 (sve): New extension.
5194 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5195 modes. Adjust their number of units based on aarch64_sve_vg.
5196 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5197 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5198 aarch64_addr_query_type.
5199 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5200 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5201 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5202 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5203 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5204 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5205 (aarch64_simd_imm_zero_p): Delete.
5206 (aarch64_check_zero_based_sve_index_immediate): Declare.
5207 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5208 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5209 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5210 (aarch64_sve_float_mul_immediate_p): Likewise.
5211 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5213 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5214 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5215 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5216 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5217 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5218 (aarch64_regmode_natural_size): Likewise.
5219 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5220 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5222 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5223 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5224 for VG and the SVE predicate registers.
5225 (V_ALIASES): Add a "z"-prefixed alias.
5226 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5227 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5228 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5229 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5230 (REG_CLASS_NAMES): Add entries for them.
5231 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5232 and the predicate registers.
5233 (aarch64_sve_vg): Declare.
5234 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5235 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5236 (REGMODE_NATURAL_SIZE): Define.
5237 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5239 * config/aarch64/aarch64.c: Include cfgrtl.h.
5240 (simd_immediate_info): Add a constructor for series vectors,
5241 and an associated step field.
5242 (aarch64_sve_vg): New variable.
5243 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5244 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5245 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5246 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5247 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5248 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5249 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5250 (aarch64_get_mask_mode): New functions.
5251 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5252 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5253 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5254 predicate modes and predicate registers. Explicitly restrict
5255 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5256 to store a vector mode if it is recognized by
5257 aarch64_classify_vector_mode.
5258 (aarch64_regmode_natural_size): New function.
5259 (aarch64_hard_regno_caller_save_mode): Return the original mode
5261 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5262 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5263 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5264 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5266 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5267 does not overlap dest if the function is frame-related. Handle
5269 (aarch64_split_add_offset): New function.
5270 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5271 them aarch64_add_offset.
5272 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5273 and update call to aarch64_sub_sp.
5274 (aarch64_add_cfa_expression): New function.
5275 (aarch64_expand_prologue): Pass extra temporary registers to the
5276 functions above. Handle the case in which we need to emit new
5277 DW_CFA_expressions for registers that were originally saved
5278 relative to the stack pointer, but now have to be expressed
5279 relative to the frame pointer.
5280 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5282 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5283 IP0 and IP1 values for SVE frames.
5284 (aarch64_expand_vec_series): New function.
5285 (aarch64_expand_sve_widened_duplicate): Likewise.
5286 (aarch64_expand_sve_const_vector): Likewise.
5287 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5288 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5289 into the register, rather than emitting a SET directly.
5290 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5291 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5292 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5293 (offset_9bit_signed_scaled_p): New functions.
5294 (aarch64_replicate_bitmask_imm): New function.
5295 (aarch64_bitmask_imm): Use it.
5296 (aarch64_cannot_force_const_mem): Reject expressions involving
5297 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5298 (aarch64_classify_index): Handle SVE indices, by requiring
5299 a plain register index with a scale that matches the element size.
5300 (aarch64_classify_address): Handle SVE addresses. Assert that
5301 the mode of the address is VOIDmode or an integer mode.
5302 Update call to aarch64_classify_symbol.
5303 (aarch64_classify_symbolic_expression): Update call to
5304 aarch64_classify_symbol.
5305 (aarch64_const_vec_all_in_range_p): New function.
5306 (aarch64_print_vector_float_operand): Likewise.
5307 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5308 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5309 and the FP immediates 1.0 and 0.5.
5310 (aarch64_print_address_internal): Handle SVE addresses.
5311 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5312 (aarch64_regno_regclass): Handle predicate registers.
5313 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5315 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5316 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5317 (aarch64_convert_sve_vector_bits): New function.
5318 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5319 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5321 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5322 Handle SVE vector and predicate modes. Accept VL-based constants
5323 that need only one temporary register, and VL offsets that require
5324 no temporary registers.
5325 (aarch64_conditional_register_usage): Mark the predicate registers
5326 as fixed if SVE isn't available.
5327 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5328 Return true for SVE vector and predicate modes.
5329 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5330 rather than an unsigned int. Handle SVE modes.
5331 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5333 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5335 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5336 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5337 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5338 (aarch64_sve_float_mul_immediate_p): New functions.
5339 (aarch64_sve_valid_immediate): New function.
5340 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5341 Explicitly reject structure modes. Check for INDEX constants.
5342 Handle PTRUE and PFALSE constants.
5343 (aarch64_check_zero_based_sve_index_immediate): New function.
5344 (aarch64_simd_imm_zero_p): Delete.
5345 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5346 vector modes. Accept constants in the range of CNT[BHWD].
5347 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5348 ask for an Advanced SIMD mode.
5349 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5350 (aarch64_simd_vector_alignment): Handle SVE predicates.
5351 (aarch64_vectorize_preferred_vector_alignment): New function.
5352 (aarch64_simd_vector_alignment_reachable): Use it instead of
5354 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5355 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5357 (MAX_VECT_LEN): Delete.
5358 (expand_vec_perm_d): Add a vec_flags field.
5359 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5360 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5361 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5363 (aarch64_evpc_rev): Rename to...
5364 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5365 (aarch64_evpc_rev_global): New function.
5366 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5367 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5369 (aarch64_evpc_sve_tbl): New function.
5370 (aarch64_expand_vec_perm_const_1): Update after rename of
5371 aarch64_evpc_rev. Handle SVE permutes too, trying
5372 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5373 than aarch64_evpc_tbl.
5374 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5375 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5376 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5377 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5378 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5379 (aarch64_expand_sve_vcond): New functions.
5380 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5381 of aarch64_vector_mode_p.
5382 (aarch64_dwarf_poly_indeterminate_value): New function.
5383 (aarch64_compute_pressure_classes): Likewise.
5384 (aarch64_can_change_mode_class): Likewise.
5385 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5386 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5387 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5388 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5389 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5390 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5391 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5392 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5394 (Dn, Dl, Dr): Accept const as well as const_vector.
5395 (Dz): Likewise. Compare against CONST0_RTX.
5396 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5397 of "vector" where appropriate.
5398 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5399 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5400 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5401 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5402 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5403 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5404 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5405 (v_int_equiv): Extend to SVE modes.
5406 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5408 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5409 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5410 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5411 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5412 (SVE_COND_FP_CMP): New int iterators.
5413 (perm_hilo): Handle the new unpack unspecs.
5414 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5416 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5417 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5418 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5419 (aarch64_equality_operator, aarch64_constant_vector_operand)
5420 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5421 (aarch64_sve_nonimmediate_operand): Likewise.
5422 (aarch64_sve_general_operand): Likewise.
5423 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5424 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5425 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5426 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5427 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5428 (aarch64_sve_float_arith_immediate): Likewise.
5429 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5430 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5431 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5432 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5433 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5434 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5435 (aarch64_sve_float_arith_operand): Likewise.
5436 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5437 (aarch64_sve_float_mul_operand): Likewise.
5438 (aarch64_sve_vec_perm_operand): Likewise.
5439 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5440 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5441 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5442 as well as const_vector.
5443 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5444 in file. Use CONST0_RTX and CONSTM1_RTX.
5445 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5446 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5447 Use aarch64_simd_imm_zero.
5448 * config/aarch64/aarch64-sve.md: New file.
5449 * config/aarch64/aarch64.md: Include it.
5450 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5451 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5452 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5453 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5454 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5455 (sve): New attribute.
5456 (enabled): Disable instructions with the sve attribute unless
5458 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5459 aarch64_expand_mov_immediate.
5460 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5461 CNT[BHSD] immediates.
5462 (movti): Split CONST_POLY_INT moves into two halves.
5463 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5464 Split additions that need a temporary here if the destination
5465 is the stack pointer.
5466 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5467 (*add<mode>3_poly_1): New instruction.
5468 (set_clobber_cc): New expander.
5470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5472 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5473 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5474 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5475 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5476 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5477 Change innermode from fixed_mode_size to machine_mode.
5478 (simplify_subreg): Update call accordingly. Handle a constant-sized
5479 subreg of a variable-length CONST_VECTOR.
5481 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5482 Alan Hayward <alan.hayward@arm.com>
5483 David Sherwood <david.sherwood@arm.com>
5485 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5486 (add_offset_to_base): New function, split out from...
5487 (create_mem_ref): ...here. When handling a scale other than 1,
5488 check first whether the address is valid without the offset.
5489 Add it into the base if so, leaving the index and scale as-is.
5491 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5494 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5495 fold_for_warn before checking if arg2 is INTEGER_CST.
5497 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5499 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5500 (store_multiple_operation): Delete.
5501 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5502 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5503 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5504 guarded by TARGET_STRING.
5505 (rs6000_output_load_multiple): Delete.
5506 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5507 OPTION_MASK_STRING / TARGET_STRING handling.
5508 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5509 (const rs6000_opt_masks) <"string">: Change mask to 0.
5510 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5511 (MASK_STRING): Delete.
5512 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5514 (load_multiple): Delete.
5521 (store_multiple): Delete.
5528 (movmemsi_8reg): Delete.
5529 (corresponding unnamed define_insn): Delete.
5530 (movmemsi_6reg): Delete.
5531 (corresponding unnamed define_insn): Delete.
5532 (movmemsi_4reg): Delete.
5533 (corresponding unnamed define_insn): Delete.
5534 (movmemsi_2reg): Delete.
5535 (corresponding unnamed define_insn): Delete.
5536 (movmemsi_1reg): Delete.
5537 (corresponding unnamed define_insn): Delete.
5538 * config/rs6000/rs6000.opt (mno-string): New.
5539 (mstring): Replace by deprecation warning stub.
5540 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5542 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5544 * regrename.c (regrename_do_replace): If replacing the same
5545 reg multiple times, try to reuse last created gen_raw_REG.
5548 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5549 main to workaround a bug in GDB.
5551 2018-01-12 Tom de Vries <tom@codesourcery.com>
5554 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5556 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5558 PR rtl-optimization/80481
5559 * ira-color.c (get_cap_member): New function.
5560 (allocnos_conflict_by_live_ranges_p): Use it.
5561 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5562 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5564 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5567 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5568 (*saddl_se_1): Ditto.
5570 (*ssubl_se_1): Ditto.
5572 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5574 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5575 rather than wi::to_widest for DR_INITs.
5576 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5577 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5578 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5580 (vect_analyze_group_access_1): Note that here.
5582 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5584 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5585 polynomial type sizes.
5587 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5589 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5590 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5591 (gimple_add_tmp_var): Likewise.
5593 2018-01-12 Martin Liska <mliska@suse.cz>
5595 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5596 (gimple_alloc_sizes): Likewise.
5597 (dump_gimple_statistics): Use PRIu64 in printf format.
5598 * gimple.h: Change uint64_t to int.
5600 2018-01-12 Martin Liska <mliska@suse.cz>
5602 * tree-core.h: Use uint64_t instead of int.
5603 * tree.c (tree_node_counts): Likewise.
5604 (tree_node_sizes): Likewise.
5605 (dump_tree_statistics): Use PRIu64 in printf format.
5607 2018-01-12 Martin Liska <mliska@suse.cz>
5609 * Makefile.in: As qsort_chk is implemented in vec.c, add
5610 vec.o to linkage of gencfn-macros.
5611 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5612 passing the info to record_node_allocation_statistics.
5613 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5615 * ggc-common.c (struct ggc_usage): Add operator== and use
5616 it in operator< and compare function.
5617 * mem-stats.h (struct mem_usage): Likewise.
5618 * vec.c (struct vec_usage): Remove operator< and compare
5619 function. Can be simply inherited.
5621 2018-01-12 Martin Jambor <mjambor@suse.cz>
5624 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5625 * tree-ssa-math-opts.c: Include domwalk.h.
5626 (convert_mult_to_fma_1): New function.
5627 (fma_transformation_info): New type.
5628 (fma_deferring_state): Likewise.
5629 (cancel_fma_deferring): New function.
5630 (result_of_phi): Likewise.
5631 (last_fma_candidate_feeds_initial_phi): Likewise.
5632 (convert_mult_to_fma): Added deferring logic, split actual
5633 transformation to convert_mult_to_fma_1.
5634 (math_opts_dom_walker): New type.
5635 (math_opts_dom_walker::after_dom_children): New method, body moved
5636 here from pass_optimize_widening_mul::execute, added deferring logic
5638 (pass_optimize_widening_mul::execute): Moved most of code to
5639 math_opts_dom_walker::after_dom_children.
5640 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5641 * config/i386/i386.c (ix86_option_override_internal): Added
5642 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5644 2018-01-12 Richard Biener <rguenther@suse.de>
5647 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5648 inline instance vars.
5650 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5653 * config/rx/rx.c (rx_is_restricted_memory_address):
5656 2018-01-12 Richard Biener <rguenther@suse.de>
5658 PR tree-optimization/80846
5659 * target.def (split_reduction): New target hook.
5660 * targhooks.c (default_split_reduction): New function.
5661 * targhooks.h (default_split_reduction): Declare.
5662 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5663 target requests first reduce vectors by combining low and high
5665 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5666 (get_vectype_for_scalar_type_and_size): Export.
5667 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5668 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5669 * doc/tm.texi: Regenerate.
5670 * config/i386/i386.c (ix86_split_reduction): Implement
5671 TARGET_VECTORIZE_SPLIT_REDUCTION.
5673 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5676 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5677 in PIC mode except for TARGET_VXWORKS_RTP.
5678 * config/sparc/sparc.c: Include cfgrtl.h.
5679 (TARGET_INIT_PIC_REG): Define.
5680 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5681 (sparc_pic_register_p): New predicate.
5682 (sparc_legitimate_address_p): Use it.
5683 (sparc_legitimize_pic_address): Likewise.
5684 (sparc_delegitimize_address): Likewise.
5685 (sparc_mode_dependent_address_p): Likewise.
5686 (gen_load_pcrel_sym): Remove 4th parameter.
5687 (load_got_register): Adjust call to above. Remove obsolete stuff.
5688 (sparc_expand_prologue): Do not call load_got_register here.
5689 (sparc_flat_expand_prologue): Likewise.
5690 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5691 (sparc_use_pseudo_pic_reg): New function.
5692 (sparc_init_pic_reg): Likewise.
5693 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5694 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5696 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5698 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5699 Add item for branch_cost.
5701 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5703 PR rtl-optimization/83565
5704 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5705 not extend the result to a larger mode for rotate operations.
5706 (num_sign_bit_copies1): Likewise.
5708 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5711 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5713 Use values-Xc.o for -pedantic.
5714 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5716 2018-01-12 Martin Liska <mliska@suse.cz>
5719 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5721 (possible_polymorphic_call_targets): Use it.
5722 (ipa_devirt): Likewise.
5724 2018-01-12 Martin Liska <mliska@suse.cz>
5726 * profile-count.h (enum profile_quality): Use 0 as invalid
5727 enum value of profile_quality.
5729 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5731 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5732 -mext-string options.
5734 2018-01-12 Richard Biener <rguenther@suse.de>
5736 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5737 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5738 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5740 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5742 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5744 * configure.ac (--with-long-double-format): Add support for the
5745 configuration option to change the default long double format on
5747 * config.gcc (powerpc*-linux*-*): Likewise.
5748 * configure: Regenerate.
5749 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5750 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5751 used without modification.
5753 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5755 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5756 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5757 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5758 MISC_BUILTIN_SPEC_BARRIER.
5759 (rs6000_init_builtins): Likewise.
5760 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5762 (speculation_barrier): New define_insn.
5763 * doc/extend.texi: Document __builtin_speculation_barrier.
5765 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5768 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5769 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5770 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5772 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5773 integral modes instead of "ss" and "sd".
5774 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5775 vectors with 32-bit and 64-bit elements.
5776 (vecdupssescalarmodesuffix): New mode attribute.
5777 (vec_dup<mode>): Use it.
5779 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5782 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5783 frame if argument is passed on stack.
5785 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5788 * ree.c (combine_reaching_defs): Optimize also
5789 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5790 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5792 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5795 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5797 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5800 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5801 after they are computed.
5803 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5805 PR tree-optimization/83695
5806 * gimple-loop-linterchange.cc
5807 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5808 reset cached scev information after interchange.
5809 (pass_linterchange::execute): Remove call to scev_reset_htab.
5811 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5813 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5814 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5815 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5816 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5817 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5818 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5819 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5820 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5821 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5822 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5823 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5824 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5825 (V_lane_reg): Likewise.
5826 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5828 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5829 (vfmal_lane_low<mode>_intrinsic,
5830 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5831 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5832 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5833 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5834 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5835 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5837 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5839 * config/arm/arm-cpus.in (fp16fml): New feature.
5840 (ALL_SIMD): Add fp16fml.
5841 (armv8.2-a): Add fp16fml as an option.
5842 (armv8.3-a): Likewise.
5843 (armv8.4-a): Add fp16fml as part of fp16.
5844 * config/arm/arm.h (TARGET_FP16FML): Define.
5845 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5847 * config/arm/arm-modes.def (V2HF): Define.
5848 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5849 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5850 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5851 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5852 vfmsl_low, vfmsl_high): New set of builtins.
5853 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5854 (vfml_op): New code attribute.
5855 (VFMLHALVES): New int iterator.
5856 (VFML, VFMLSEL): New mode attributes.
5857 (V_reg): Define mapping for V2HF.
5858 (V_hi, V_lo): New mode attributes.
5859 (VF_constraint): Likewise.
5860 (vfml_half, vfml_half_selector): New int attributes.
5861 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5863 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5864 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5866 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5867 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5868 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5869 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5871 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5872 Document new effective target and option set.
5874 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5876 * config/arm/arm-cpus.in (armv8_4): New feature.
5877 (ARMv8_4a): New fgroup.
5878 (armv8.4-a): New arch.
5879 * config/arm/arm-tables.opt: Regenerate.
5880 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5881 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5882 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5883 Add matching rules for -march=armv8.4-a and extensions.
5884 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5886 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5889 * config/rx/rx.md (BW): New mode attribute.
5890 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5892 2018-01-11 Richard Biener <rguenther@suse.de>
5894 PR tree-optimization/83435
5895 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5896 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5897 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5899 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5900 Alan Hayward <alan.hayward@arm.com>
5901 David Sherwood <david.sherwood@arm.com>
5903 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5905 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5906 (aarch64_print_address_internal): Use it to check for a zero offset.
5908 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5909 Alan Hayward <alan.hayward@arm.com>
5910 David Sherwood <david.sherwood@arm.com>
5912 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5913 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5914 Return a poly_int64 rather than a HOST_WIDE_INT.
5915 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5916 rather than a HOST_WIDE_INT.
5917 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5918 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5919 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5920 final_offset from HOST_WIDE_INT to poly_int64.
5921 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5922 to_constant when getting the number of units in an Advanced SIMD
5924 (aarch64_builtin_vectorized_function): Check for a constant number
5926 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5928 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5929 attribute instead of GET_MODE_NUNITS.
5930 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5931 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5932 GET_MODE_SIZE for fixed-size registers.
5933 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5934 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5935 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5936 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5937 (aarch64_print_operand, aarch64_print_address_internal)
5938 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5939 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5940 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5941 Handle polynomial GET_MODE_SIZE.
5942 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5943 wider than SImode without modification.
5944 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5945 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5946 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5947 passing and returning SVE modes.
5948 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5949 rather than GEN_INT.
5950 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5951 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5952 (aarch64_allocate_and_probe_stack_space): Likewise.
5953 (aarch64_layout_frame): Cope with polynomial offsets.
5954 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5955 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5957 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5958 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5959 poly_int64 rather than a HOST_WIDE_INT.
5960 (aarch64_get_separate_components, aarch64_process_components)
5961 (aarch64_expand_prologue, aarch64_expand_epilogue)
5962 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5963 (aarch64_anchor_offset): New function, split out from...
5964 (aarch64_legitimize_address): ...here.
5965 (aarch64_builtin_vectorization_cost): Handle polynomial
5966 TYPE_VECTOR_SUBPARTS.
5967 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5969 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5970 number of elements from the PARALLEL rather than the mode.
5971 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5972 rather than GET_MODE_BITSIZE.
5973 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5974 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5975 (aarch64_expand_vec_perm_const_1): Handle polynomial
5976 d->perm.length () and d->perm elements.
5977 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5978 Apply to_constant to d->perm elements.
5979 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5980 polynomial CONST_VECTOR_NUNITS.
5981 (aarch64_move_pointer): Take amount as a poly_int64 rather
5983 (aarch64_progress_pointer): Avoid temporary variable.
5984 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5985 the mode attribute instead of GET_MODE.
5987 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5988 Alan Hayward <alan.hayward@arm.com>
5989 David Sherwood <david.sherwood@arm.com>
5991 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5992 x exists before using it.
5993 (aarch64_add_constant_internal): Rename to...
5994 (aarch64_add_offset_1): ...this. Replace regnum with separate
5995 src and dest rtxes. Handle the case in which they're different,
5996 including when the offset is zero. Replace scratchreg with an rtx.
5997 Use 2 additions if there is no spare register into which we can
5998 move a 16-bit constant.
5999 (aarch64_add_constant): Delete.
6000 (aarch64_add_offset): Replace reg with separate src and dest
6001 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6002 Use aarch64_add_offset_1.
6003 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6004 an rtx rather than an int. Take the delta as a poly_int64
6005 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6006 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6007 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6008 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6009 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6011 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6012 aarch64_add_constant.
6014 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6016 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6017 Use scalar_float_mode.
6019 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6021 * config/aarch64/aarch64-simd.md
6022 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6023 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6024 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6025 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6026 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6027 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6028 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6029 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6030 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6031 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6033 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6036 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6037 targ_options->x_arm_arch_string is non NULL.
6039 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6041 * config/aarch64/aarch64.h
6042 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6044 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6047 * expmed.c (emit_store_flag_force): Swap if const op0
6048 and change VOIDmode to mode of op0.
6050 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6052 PR rtl-optimization/83761
6053 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6054 than bytes to mode_for_size.
6056 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6059 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6060 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6063 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6066 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6067 when in layout mode.
6068 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6069 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6072 2018-01-10 Michael Collison <michael.collison@arm.com>
6074 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6075 * config/aarch64/aarch64-option-extension.def: Add
6076 AARCH64_OPT_EXTENSION of 'fp16fml'.
6077 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6078 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6079 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6080 * config/aarch64/constraints.md (Ui7): New constraint.
6081 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6082 (VFMLA_SEL_W): Ditto.
6085 (VFMLA16_LOW): New int iterator.
6086 (VFMLA16_HIGH): Ditto.
6087 (UNSPEC_FMLAL): New unspec.
6088 (UNSPEC_FMLSL): Ditto.
6089 (UNSPEC_FMLAL2): Ditto.
6090 (UNSPEC_FMLSL2): Ditto.
6091 (f16mac): New code attribute.
6092 * config/aarch64/aarch64-simd-builtins.def
6093 (aarch64_fmlal_lowv2sf): Ditto.
6094 (aarch64_fmlsl_lowv2sf): Ditto.
6095 (aarch64_fmlalq_lowv4sf): Ditto.
6096 (aarch64_fmlslq_lowv4sf): Ditto.
6097 (aarch64_fmlal_highv2sf): Ditto.
6098 (aarch64_fmlsl_highv2sf): Ditto.
6099 (aarch64_fmlalq_highv4sf): Ditto.
6100 (aarch64_fmlslq_highv4sf): Ditto.
6101 (aarch64_fmlal_lane_lowv2sf): Ditto.
6102 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6103 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6104 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6105 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6106 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6107 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6108 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6109 (aarch64_fmlal_lane_highv2sf): Ditto.
6110 (aarch64_fmlsl_lane_highv2sf): Ditto.
6111 (aarch64_fmlal_laneq_highv2sf): Ditto.
6112 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6113 (aarch64_fmlalq_lane_highv4sf): Ditto.
6114 (aarch64_fmlsl_lane_highv4sf): Ditto.
6115 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6116 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6117 * config/aarch64/aarch64-simd.md:
6118 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6119 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6120 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6121 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6122 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6123 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6124 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6125 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6126 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6127 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6128 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6129 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6130 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6131 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6132 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6133 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6134 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6135 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6136 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6137 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6138 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6139 (vfmlsl_low_u32): Ditto.
6140 (vfmlalq_low_u32): Ditto.
6141 (vfmlslq_low_u32): Ditto.
6142 (vfmlal_high_u32): Ditto.
6143 (vfmlsl_high_u32): Ditto.
6144 (vfmlalq_high_u32): Ditto.
6145 (vfmlslq_high_u32): Ditto.
6146 (vfmlal_lane_low_u32): Ditto.
6147 (vfmlsl_lane_low_u32): Ditto.
6148 (vfmlal_laneq_low_u32): Ditto.
6149 (vfmlsl_laneq_low_u32): Ditto.
6150 (vfmlalq_lane_low_u32): Ditto.
6151 (vfmlslq_lane_low_u32): Ditto.
6152 (vfmlalq_laneq_low_u32): Ditto.
6153 (vfmlslq_laneq_low_u32): Ditto.
6154 (vfmlal_lane_high_u32): Ditto.
6155 (vfmlsl_lane_high_u32): Ditto.
6156 (vfmlal_laneq_high_u32): Ditto.
6157 (vfmlsl_laneq_high_u32): Ditto.
6158 (vfmlalq_lane_high_u32): Ditto.
6159 (vfmlslq_lane_high_u32): Ditto.
6160 (vfmlalq_laneq_high_u32): Ditto.
6161 (vfmlslq_laneq_high_u32): Ditto.
6162 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6163 (AARCH64_FL_FOR_ARCH8_4): New.
6164 (AARCH64_ISA_F16FML): New ISA flag.
6165 (TARGET_F16FML): New feature flag for fp16fml.
6166 (doc/invoke.texi): Document new fp16fml option.
6168 2018-01-10 Michael Collison <michael.collison@arm.com>
6170 * config/aarch64/aarch64-builtins.c:
6171 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6172 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6173 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6174 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6175 (AARCH64_ISA_SHA3): New ISA flag.
6176 (TARGET_SHA3): New feature flag for sha3.
6177 * config/aarch64/iterators.md (sha512_op): New int attribute.
6178 (CRYPTO_SHA512): New int iterator.
6179 (UNSPEC_SHA512H): New unspec.
6180 (UNSPEC_SHA512H2): Ditto.
6181 (UNSPEC_SHA512SU0): Ditto.
6182 (UNSPEC_SHA512SU1): Ditto.
6183 * config/aarch64/aarch64-simd-builtins.def
6184 (aarch64_crypto_sha512hqv2di): New builtin.
6185 (aarch64_crypto_sha512h2qv2di): Ditto.
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/aarch64-simd.md:
6193 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6194 (aarch64_crypto_sha512su0qv2di): Ditto.
6195 (aarch64_crypto_sha512su1qv2di): Ditto.
6196 (aarch64_eor3qv8hi): Ditto.
6197 (aarch64_rax1qv2di): Ditto.
6198 (aarch64_xarqv2di): Ditto.
6199 (aarch64_bcaxqv8hi): Ditto.
6200 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6201 (vsha512h2q_u64): Ditto.
6202 (vsha512su0q_u64): Ditto.
6203 (vsha512su1q_u64): Ditto.
6204 (veor3q_u16): Ditto.
6205 (vrax1q_u64): Ditto.
6207 (vbcaxq_u16): Ditto.
6208 * config/arm/types.md (crypto_sha512): New type attribute.
6209 (crypto_sha3): Ditto.
6210 (doc/invoke.texi): Document new sha3 option.
6212 2018-01-10 Michael Collison <michael.collison@arm.com>
6214 * config/aarch64/aarch64-builtins.c:
6215 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6216 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6217 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6218 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6219 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6220 (AARCH64_ISA_SM4): New ISA flag.
6221 (TARGET_SM4): New feature flag for sm4.
6222 * config/aarch64/aarch64-simd-builtins.def
6223 (aarch64_sm3ss1qv4si): Ditto.
6224 (aarch64_sm3tt1aq4si): Ditto.
6225 (aarch64_sm3tt1bq4si): Ditto.
6226 (aarch64_sm3tt2aq4si): Ditto.
6227 (aarch64_sm3tt2bq4si): Ditto.
6228 (aarch64_sm3partw1qv4si): Ditto.
6229 (aarch64_sm3partw2qv4si): Ditto.
6230 (aarch64_sm4eqv4si): Ditto.
6231 (aarch64_sm4ekeyqv4si): Ditto.
6232 * config/aarch64/aarch64-simd.md:
6233 (aarch64_sm3ss1qv4si): Ditto.
6234 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6235 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6236 (aarch64_sm4eqv4si): Ditto.
6237 (aarch64_sm4ekeyqv4si): Ditto.
6238 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6239 (sm3part_op): Ditto.
6240 (CRYPTO_SM3TT): Ditto.
6241 (CRYPTO_SM3PART): Ditto.
6242 (UNSPEC_SM3SS1): New unspec.
6243 (UNSPEC_SM3TT1A): Ditto.
6244 (UNSPEC_SM3TT1B): Ditto.
6245 (UNSPEC_SM3TT2A): Ditto.
6246 (UNSPEC_SM3TT2B): Ditto.
6247 (UNSPEC_SM3PARTW1): Ditto.
6248 (UNSPEC_SM3PARTW2): Ditto.
6249 (UNSPEC_SM4E): Ditto.
6250 (UNSPEC_SM4EKEY): Ditto.
6251 * config/aarch64/constraints.md (Ui2): New constraint.
6252 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6253 * config/arm/types.md (crypto_sm3): New type attribute.
6254 (crypto_sm4): Ditto.
6255 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6256 (vsm3tt1aq_u32): Ditto.
6257 (vsm3tt1bq_u32): Ditto.
6258 (vsm3tt2aq_u32): Ditto.
6259 (vsm3tt2bq_u32): Ditto.
6260 (vsm3partw1q_u32): Ditto.
6261 (vsm3partw2q_u32): Ditto.
6262 (vsm4eq_u32): Ditto.
6263 (vsm4ekeyq_u32): Ditto.
6264 (doc/invoke.texi): Document new sm4 option.
6266 2018-01-10 Michael Collison <michael.collison@arm.com>
6268 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6269 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6270 (AARCH64_FL_FOR_ARCH8_4): New.
6271 (AARCH64_FL_V8_4): New flag.
6272 (doc/invoke.texi): Document new armv8.4-a option.
6274 2018-01-10 Michael Collison <michael.collison@arm.com>
6276 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6277 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6278 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6279 * config/aarch64/aarch64-option-extension.def: Add
6280 AARCH64_OPT_EXTENSION of 'sha2'.
6281 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6282 (crypto): Disable sha2 and aes if crypto disabled.
6283 (crypto): Enable aes and sha2 if enabled.
6284 (simd): Disable sha2 and aes if simd disabled.
6285 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6287 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6288 (TARGET_SHA2): New feature flag for sha2.
6289 (TARGET_AES): New feature flag for aes.
6290 * config/aarch64/aarch64-simd.md:
6291 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6292 conditional on TARGET_AES.
6293 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6294 (aarch64_crypto_sha1hsi): Make pattern conditional
6296 (aarch64_crypto_sha1hv4si): Ditto.
6297 (aarch64_be_crypto_sha1hv4si): Ditto.
6298 (aarch64_crypto_sha1su1v4si): Ditto.
6299 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6300 (aarch64_crypto_sha1su0v4si): Ditto.
6301 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6302 (aarch64_crypto_sha256su0v4si): Ditto.
6303 (aarch64_crypto_sha256su1v4si): Ditto.
6304 (doc/invoke.texi): Document new aes and sha2 options.
6306 2018-01-10 Martin Sebor <msebor@redhat.com>
6308 PR tree-optimization/83781
6309 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6312 2018-01-11 Martin Sebor <msebor@gmail.com>
6313 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6315 PR tree-optimization/83501
6316 PR tree-optimization/81703
6318 * tree-ssa-strlen.c (get_string_cst): Rename...
6319 (get_string_len): ...to this. Handle global constants.
6320 (handle_char_store): Adjust.
6322 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6323 Jim Wilson <jimw@sifive.com>
6325 * config/riscv/riscv-protos.h (riscv_output_return): New.
6326 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6327 (riscv_attribute_table, riscv_output_return),
6328 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6329 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6330 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6332 (riscv_expand_prologue): Add early return for naked function.
6333 (riscv_expand_epilogue): Likewise.
6334 (riscv_function_ok_for_sibcall): Return false for naked function.
6335 (riscv_set_current_function): New.
6336 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6337 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6338 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6339 * doc/extend.texi (RISC-V Function Attributes): New.
6341 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6343 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6344 check for 128-bit long double before checking TCmode.
6345 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6346 128-bit long doubles before checking TFmode or TCmode.
6347 (FLOAT128_IBM_P): Likewise.
6349 2018-01-10 Martin Sebor <msebor@redhat.com>
6351 PR tree-optimization/83671
6352 * builtins.c (c_strlen): Unconditionally return zero for the empty
6354 Use -Warray-bounds for warnings.
6355 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6356 for non-constant array indices with COMPONENT_REF, arrays of
6357 arrays, and pointers to arrays.
6358 (gimple_fold_builtin_strlen): Determine and set length range for
6359 non-constant character arrays.
6361 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6364 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6367 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6369 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6371 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6374 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6375 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6376 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6377 indexed_or_indirect_operand predicate.
6378 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6379 (*vsx_le_perm_load_v8hi): Likewise.
6380 (*vsx_le_perm_load_v16qi): Likewise.
6381 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6382 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6383 (*vsx_le_perm_store_v8hi): Likewise.
6384 (*vsx_le_perm_store_v16qi): Likewise.
6385 (eight unnamed splitters): Likewise.
6387 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6389 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6390 * config/rs6000/emmintrin.h: Likewise.
6391 * config/rs6000/mmintrin.h: Likewise.
6392 * config/rs6000/xmmintrin.h: Likewise.
6394 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6397 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6399 * tree.c (tree_nop_conversion): Return true for location wrapper
6401 (maybe_wrap_with_location): New function.
6402 (selftest::check_strip_nops): New function.
6403 (selftest::test_location_wrappers): New function.
6404 (selftest::tree_c_tests): Call it.
6405 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6406 (maybe_wrap_with_location): New decl.
6407 (EXPR_LOCATION_WRAPPER_P): New macro.
6408 (location_wrapper_p): New inline function.
6409 (tree_strip_any_location_wrapper): New inline function.
6411 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6414 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6415 stack_realign_offset for the largest alignment of stack slot
6417 (ix86_find_max_used_stack_alignment): New function.
6418 (ix86_finalize_stack_frame_flags): Use it. Set
6419 max_used_stack_alignment if we don't realign stack.
6420 * config/i386/i386.h (machine_function): Add
6421 max_used_stack_alignment.
6423 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6425 * config/arm/arm.opt (-mbranch-cost): New option.
6426 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6429 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6432 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6433 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6435 2018-01-10 Richard Biener <rguenther@suse.de>
6438 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6439 early out so it also covers the case where we have a non-NULL
6442 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6444 PR tree-optimization/83753
6445 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6446 for non-strided grouped accesses if the number of elements is 1.
6448 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6451 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6452 * i386.h (TARGET_USE_GATHER): Define.
6453 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6455 2018-01-10 Martin Liska <mliska@suse.cz>
6458 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6459 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6461 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6462 CLEANUP_NO_PARTITIONING is not set.
6464 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6466 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6467 for vectors, as a partial revert of r254296.
6468 * rtl.h (const_vec_p): Delete.
6469 (const_vec_duplicate_p): Don't test for vector CONSTs.
6470 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6471 * expmed.c (make_tree): Likewise.
6474 * common.md (E, F): Use CONSTANT_P instead of checking for
6476 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6477 checking for CONST_VECTOR.
6479 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6482 * predict.c (force_edge_cold): Handle in more sane way edges
6485 2018-01-09 Carl Love <cel@us.ibm.com>
6487 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6489 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6490 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6491 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6492 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6493 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6494 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6495 * config/rs6000/rs6000-protos.h: Add extern defition for
6496 rs6000_generate_float2_double_code.
6497 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6499 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6500 (float2_v2df): Add define_expand.
6502 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6505 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6506 op_mode in the force_to_mode call.
6508 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6510 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6511 instead of checking each element individually.
6512 (aarch64_evpc_uzp): Likewise.
6513 (aarch64_evpc_zip): Likewise.
6514 (aarch64_evpc_ext): Likewise.
6515 (aarch64_evpc_rev): Likewise.
6516 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6517 instead of checking each element individually. Return true without
6519 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6520 whether all selected elements come from the same input, instead of
6521 checking each element individually. Remove calls to gen_rtx_REG,
6522 start_sequence and end_sequence and instead assert that no rtl is
6525 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6527 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6528 order of HIGH and CONST checks.
6530 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6532 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6533 if the destination isn't an SSA_NAME.
6535 2018-01-09 Richard Biener <rguenther@suse.de>
6537 PR tree-optimization/83668
6538 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6540 (canonicalize_loop_form): ... here, renamed from ...
6541 (canonicalize_loop_closed_ssa_form): ... this and amended to
6542 swap successor edges for loop exit blocks to make us use
6543 the RPO order we need for initial schedule generation.
6545 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6547 PR tree-optimization/64811
6548 * match.pd: When optimizing comparisons with Inf, avoid
6549 introducing or losing exceptions from comparisons with NaN.
6551 2018-01-09 Martin Liska <mliska@suse.cz>
6554 * asan.c (shadow_mem_size): Add gcc_assert.
6556 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6558 Don't save registers in main().
6561 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6562 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6563 * config/avr/avr.c (avr_set_current_function): Don't error if
6564 naked, OS_task or OS_main are specified at the same time.
6565 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6567 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6569 * common/config/avr/avr-common.c (avr_option_optimization_table):
6570 Switch on -mmain-is-OS_task for optimizing compilations.
6572 2018-01-09 Richard Biener <rguenther@suse.de>
6574 PR tree-optimization/83572
6575 * graphite.c: Include cfganal.h.
6576 (graphite_transform_loops): Connect infinite loops to exit
6577 and remove fake edges at the end.
6579 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6581 * ipa-inline.c (edge_badness): Revert accidental checkin.
6583 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6586 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6587 symbols; not inline clones.
6589 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6592 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6593 hard registers. Formatting fixes.
6595 PR preprocessor/83722
6596 * gcc.c (try_generate_repro): Pass
6597 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6598 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6601 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6602 Kito Cheng <kito.cheng@gmail.com>
6604 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6605 (riscv_leaf_function_p): Delete.
6606 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6608 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6610 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6612 (do_ifelse): New function.
6613 (do_isel): New function.
6614 (do_sub3): New function.
6615 (do_add3): New function.
6616 (do_load_mask_compare): New function.
6617 (do_overlap_load_compare): New function.
6618 (expand_compare_loop): New function.
6619 (expand_block_compare): Call expand_compare_loop() when appropriate.
6620 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6622 (-mblock-compare-inline-loop-limit): New option.
6624 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6627 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6628 Reverse order of second and third operands in first alternative.
6629 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6630 of first and second elements in UNSPEC_VPERMR vector.
6631 (altivec_expand_vec_perm_le): Likewise.
6633 2018-01-08 Jeff Law <law@redhat.com>
6635 PR rtl-optimizatin/81308
6636 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6637 (process_switch): If group_case_labels makes a change, then set
6639 (pass_convert_switch::execute): If a switch is converted, then
6640 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6642 PR rtl-optimization/81308
6643 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6646 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6648 PR target/83663 - Revert r255946
6649 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6650 generation for cases where splatting a value is not useful.
6651 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6652 across a vec_duplicate and a paradoxical subreg forming a vector
6653 mode to a vec_concat.
6655 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6657 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6658 -march=armv8.3-a variants.
6659 * config/arm/t-multilib: Likewise.
6660 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6662 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6664 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6666 (cceq_ior_compare_complement): Give it a name so I can use it, and
6667 change boolean_or_operator predicate to boolean_operator so it can
6668 be used to generate a crand.
6669 (eqne): New code iterator.
6670 (bd/bd_neg): New code_attrs.
6671 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6672 a single define_insn.
6673 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6674 decrement (bdnzt/bdnzf/bdzt/bdzf).
6675 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6676 with the new names of the branch decrement patterns, and added the
6677 names of the branch decrement conditional patterns.
6679 2018-01-08 Richard Biener <rguenther@suse.de>
6681 PR tree-optimization/83563
6682 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6685 2018-01-08 Richard Biener <rguenther@suse.de>
6688 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6690 2018-01-08 Richard Biener <rguenther@suse.de>
6692 PR tree-optimization/83685
6693 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6694 references to abnormals.
6696 2018-01-08 Richard Biener <rguenther@suse.de>
6699 * dwarf2out.c (output_indirect_strings): Handle empty
6700 skeleton_debug_str_hash.
6701 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6703 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6705 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6706 (emit_store_direct): Likewise.
6707 (arc_trampoline_adjust_address): Likewise.
6708 (arc_asm_trampoline_template): New function.
6709 (arc_initialize_trampoline): Use asm_trampoline_template.
6710 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6711 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6712 * config/arc/arc.md (flush_icache): Delete pattern.
6714 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6716 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6717 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6720 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6723 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6724 by not USED_FOR_TARGET.
6725 (make_pass_resolve_sw_modes): Likewise.
6727 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6729 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6732 2018-01-08 Richard Biener <rguenther@suse.de>
6735 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6737 2018-01-08 Richard Biener <rguenther@suse.de>
6740 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6742 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6745 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6746 basic blocks with a small number of successors.
6747 (convert_control_dep_chain_into_preds): Improve handling of
6749 (dump_predicates): Split apart into...
6750 (dump_pred_chain): ...here...
6751 (dump_pred_info): ...and here.
6752 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6753 (can_chain_union_be_invalidated_p): Improve check for invalidation
6755 (uninit_uses_cannot_happen): Avoid unnecessary if
6756 convert_control_dep_chain_into_preds yielded nothing.
6758 2018-01-06 Martin Sebor <msebor@redhat.com>
6760 PR tree-optimization/83640
6761 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6762 subtracting negative offset from size.
6763 (builtin_access::overlap): Adjust offset bounds of the access to fall
6764 within the size of the object if possible.
6766 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6768 PR rtl-optimization/83699
6769 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6770 extract_bit_field_as_subreg to cases in which the extracted
6771 value is also a vector.
6773 * lra-constraints.c (process_alt_operands): Test for the equivalence
6774 substitutions when detecting a possible reload cycle.
6776 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6779 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6780 by default if flag_selective_schedling{,2}. Formatting fixes.
6782 PR rtl-optimization/83682
6783 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6784 if it has non-VECTOR_MODE element mode.
6785 (vec_duplicate_p): Likewise.
6788 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6789 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6791 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6794 * config/i386/i386-builtin.def
6795 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6796 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6797 Require also OPTION_MASK_ISA_AVX512F in addition to
6798 OPTION_MASK_ISA_GFNI.
6799 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6800 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6801 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6802 to OPTION_MASK_ISA_GFNI.
6803 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6804 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6805 OPTION_MASK_ISA_AVX512BW.
6806 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6807 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6808 addition to OPTION_MASK_ISA_GFNI.
6809 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6810 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6811 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6812 to OPTION_MASK_ISA_GFNI.
6813 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6814 a requirement for all ISAs rather than any of them with a few
6816 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6818 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6819 bitmasks to be enabled with 3 exceptions, instead of requiring any
6820 enabled ISA with lots of exceptions.
6821 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6822 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6823 Change avx512bw in isa attribute to avx512f.
6824 * config/i386/sgxintrin.h: Add license boilerplate.
6825 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6826 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6827 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6828 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6830 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6831 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6832 temporarily sse2 rather than sse if not enabled already.
6835 * config/i386/sse.md (VI248_VLBW): Rename to ...
6836 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6837 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6838 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6839 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6840 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6841 mode iterator instead of VI248_VLBW.
6843 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6845 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6846 (record_modified): Skip clobbers; add debug output.
6847 (param_change_prob): Use sreal frequencies.
6849 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6851 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6852 punt for user-aligned variables.
6854 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6856 * tree-chrec.c (chrec_contains_symbols): Return true for
6859 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6862 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6863 of (x|y) == x for BICS pattern.
6865 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6867 PR tree-optimization/83605
6868 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6869 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6872 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6874 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6875 * config/epiphany/rtems.h: New file.
6877 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6878 Uros Bizjak <ubizjak@gmail.com>
6881 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6882 QIreg_operand instead of register_operand predicate.
6883 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6884 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6885 comments instead of -fmitigate[-_]rop.
6887 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6890 * cgraphunit.c (symbol_table::compile): Switch to text_section
6891 before calling assembly_start debug hook.
6892 * run-rtl-passes.c (run_rtl_passes): Likewise.
6895 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6897 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6898 range_int_cst_p rather than !symbolic_range_p before calling
6899 extract_range_from_multiplicative_op_1.
6901 2018-01-04 Jeff Law <law@redhat.com>
6903 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6904 redundant test in assertion.
6906 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6908 * doc/rtl.texi: Document machine_mode wrapper classes.
6910 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6912 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6915 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6917 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6918 the VEC_PERM_EXPR fold to fail.
6920 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6923 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6924 to switched_sections.
6926 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6929 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6932 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6935 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6936 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6938 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6941 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6942 is BLKmode and bitpos not zero or mode change is needed.
6944 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6947 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6950 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6953 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6954 instead of MULT rtx. Update all corresponding splitters.
6956 (*ssub<modesuffix>): Ditto.
6958 (*cmp_sadd_di): Update split patterns.
6959 (*cmp_sadd_si): Ditto.
6960 (*cmp_sadd_sidi): Ditto.
6961 (*cmp_ssub_di): Ditto.
6962 (*cmp_ssub_si): Ditto.
6963 (*cmp_ssub_sidi): Ditto.
6964 * config/alpha/predicates.md (const23_operand): New predicate.
6965 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6966 Look for ASHIFT, not MULT inner operand.
6967 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6969 2018-01-04 Martin Liska <mliska@suse.cz>
6971 PR gcov-profile/83669
6972 * gcov.c (output_intermediate_file): Add version to intermediate
6974 * doc/gcov.texi: Document new field 'version' in intermediate
6975 file format. Fix location of '-k' option of gcov command.
6977 2018-01-04 Martin Liska <mliska@suse.cz>
6980 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6982 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6984 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6986 2018-01-03 Martin Sebor <msebor@redhat.com>
6988 PR tree-optimization/83655
6989 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6990 checking calls with invalid arguments.
6992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6994 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6995 (vectorizable_mask_load_store): Delete.
6996 (vectorizable_call): Return false for masked loads and stores.
6997 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6998 instead of gimple_assign_rhs1.
6999 (vectorizable_load): Handle IFN_MASK_LOAD.
7000 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7004 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7006 (vectorizable_mask_load_store): ...here.
7007 (vectorizable_load): ...and here.
7009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7011 * tree-vect-stmts.c (vect_build_all_ones_mask)
7012 (vect_build_zero_merge_argument): New functions, split out from...
7013 (vectorizable_load): ...here.
7015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7017 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7019 (vectorizable_mask_load_store): ...here.
7020 (vectorizable_store): ...and here.
7022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7024 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7026 (vectorizable_mask_load_store): ...here.
7028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7030 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7031 (vect_model_store_cost): Take a vec_load_store_type instead of a
7033 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7034 (vect_model_store_cost): Take a vec_load_store_type instead of a
7036 (vectorizable_mask_load_store): Update accordingly.
7037 (vectorizable_store): Likewise.
7038 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7042 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7043 IFN_MASK_LOAD calls here rather than...
7044 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
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 * expmed.c (extract_bit_field_1): For vector extracts,
7051 fall back to extract_bit_field_as_subreg if vec_extract
7054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7055 Alan Hayward <alan.hayward@arm.com>
7056 David Sherwood <david.sherwood@arm.com>
7058 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7059 they are variable or constant sized.
7060 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7061 slots for constant-sized data.
7063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7064 Alan Hayward <alan.hayward@arm.com>
7065 David Sherwood <david.sherwood@arm.com>
7067 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7068 handling COND_EXPRs with boolean comparisons, try to find a better
7069 basis for the mask type than the boolean itself.
7071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7073 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7074 is calculated and how it can be overridden.
7075 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7076 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7078 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7082 Alan Hayward <alan.hayward@arm.com>
7083 David Sherwood <david.sherwood@arm.com>
7085 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7086 Remove the mode argument.
7087 (aarch64_simd_valid_immediate): Remove the mode and inverse
7089 * config/aarch64/iterators.md (bitsize): New iterator.
7090 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7091 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7092 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7093 aarch64_simd_valid_immediate.
7094 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7095 (aarch64_reg_or_bic_imm): Likewise.
7096 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7097 with an insn_type enum and msl with a modifier_type enum.
7098 Replace element_width with a scalar_mode. Change the shift
7099 to unsigned int. Add constructors for scalar_float_mode and
7100 scalar_int_mode elements.
7101 (aarch64_vect_float_const_representable_p): Delete.
7102 (aarch64_can_const_movi_rtx_p)
7103 (aarch64_simd_scalar_immediate_valid_for_move)
7104 (aarch64_simd_make_constant): Update call to
7105 aarch64_simd_valid_immediate.
7106 (aarch64_advsimd_valid_immediate_hs): New function.
7107 (aarch64_advsimd_valid_immediate): Likewise.
7108 (aarch64_simd_valid_immediate): Remove mode and inverse
7109 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7110 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7111 and aarch64_float_const_representable_p on the result.
7112 (aarch64_output_simd_mov_immediate): Remove mode argument.
7113 Update call to aarch64_simd_valid_immediate and use of
7114 simd_immediate_info.
7115 (aarch64_output_scalar_simd_mov_immediate): Update call
7118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7119 Alan Hayward <alan.hayward@arm.com>
7120 David Sherwood <david.sherwood@arm.com>
7122 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7123 (mode_nunits): Likewise CONST_MODE_NUNITS.
7124 * machmode.def (ADJUST_NUNITS): Document.
7125 * genmodes.c (mode_data::need_nunits_adj): New field.
7126 (blank_mode): Update accordingly.
7127 (adj_nunits): New variable.
7128 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7130 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7131 listed in adj_nunits.
7132 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7133 listed in adj_nunits. Don't emit case statements for such modes.
7134 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7135 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7136 nothing if adj_nunits is nonnull.
7137 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7138 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7139 (emit_mode_fbit): Update use of print_maybe_const_decl.
7140 (emit_move_size): Likewise. Treat the array as non-const
7142 (emit_mode_adjustments): Handle adj_nunits.
7144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7146 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7147 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7148 (VECTOR_MODES): Use it.
7149 (make_vector_modes): Take the prefix as an argument.
7151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7152 Alan Hayward <alan.hayward@arm.com>
7153 David Sherwood <david.sherwood@arm.com>
7155 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7156 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7157 for MODE_VECTOR_BOOL.
7158 * machmode.def (VECTOR_BOOL_MODE): Document.
7159 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7160 (make_vector_bool_mode): New function.
7161 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7163 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7164 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7166 * stor-layout.c (int_mode_for_mode): Likewise.
7167 * tree.c (build_vector_type_for_mode): Likewise.
7168 * varasm.c (output_constant_pool_2): Likewise.
7169 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7170 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7171 for MODE_VECTOR_BOOL.
7172 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7173 of mode class checks.
7174 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7175 instead of a list of mode class checks.
7176 (expand_vector_scalar_condition): Likewise.
7177 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7180 Alan Hayward <alan.hayward@arm.com>
7181 David Sherwood <david.sherwood@arm.com>
7183 * machmode.h (mode_size): Change from unsigned short to
7185 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7186 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7187 or if measurement_type is not polynomial.
7188 (fixed_size_mode::includes_p): Check for constant-sized modes.
7189 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7190 return a poly_uint16 rather than an unsigned short.
7191 (emit_mode_size): Change the type of mode_size from unsigned short
7192 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7193 (emit_mode_adjustments): Cope with polynomial vector sizes.
7194 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7196 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7198 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7199 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7200 * caller-save.c (setup_save_areas): Likewise.
7201 (replace_reg_with_saved_mem): Likewise.
7202 * calls.c (emit_library_call_value_1): Likewise.
7203 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7204 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7205 (gen_lowpart_for_combine): Likewise.
7206 * convert.c (convert_to_integer_1): Likewise.
7207 * cse.c (equiv_constant, cse_insn): Likewise.
7208 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7209 (cselib_subst_to_values): Likewise.
7210 * dce.c (word_dce_process_block): Likewise.
7211 * df-problems.c (df_word_lr_mark_ref): Likewise.
7212 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7213 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7214 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7215 (rtl_for_decl_location): Likewise.
7216 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7217 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7218 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7219 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7220 (expand_expr_real_1): Likewise.
7221 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7222 (pad_below): Likewise.
7223 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7224 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7225 * ira.c (get_subreg_tracking_sizes): Likewise.
7226 * ira-build.c (ira_create_allocno_objects): Likewise.
7227 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7228 (ira_sort_regnos_for_alter_reg): Likewise.
7229 * ira-costs.c (record_operand_costs): Likewise.
7230 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7231 (resolve_simple_move): Likewise.
7232 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7233 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7234 (lra_constraints): Likewise.
7235 (CONST_POOL_OK_P): Reject variable-sized modes.
7236 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7237 (add_pseudo_to_slot, lra_spill): Likewise.
7238 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7239 * optabs-query.c (get_best_extraction_insn): Likewise.
7240 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7241 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7242 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7243 * recog.c (offsettable_address_addr_space_p): Likewise.
7244 * regcprop.c (maybe_mode_change): Likewise.
7245 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7246 * regrename.c (build_def_use): Likewise.
7247 * regstat.c (dump_reg_info): Likewise.
7248 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7249 (find_reloads, find_reloads_subreg_address): Likewise.
7250 * reload1.c (eliminate_regs_1): Likewise.
7251 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7252 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7253 (simplify_binary_operation_1, simplify_subreg): Likewise.
7254 * targhooks.c (default_function_arg_padding): Likewise.
7255 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7256 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7257 (verify_gimple_assign_ternary): Likewise.
7258 * tree-inline.c (estimate_move_cost): Likewise.
7259 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7260 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7261 (get_address_cost_ainc): Likewise.
7262 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7263 (vect_supportable_dr_alignment): Likewise.
7264 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7265 (vectorizable_reduction): Likewise.
7266 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7267 (vectorizable_operation, vectorizable_load): Likewise.
7268 * tree.c (build_same_sized_truth_vector_type): Likewise.
7269 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7270 * var-tracking.c (emit_note_insn_var_location): Likewise.
7271 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7272 (ADDR_VEC_ALIGN): Likewise.
7274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7275 Alan Hayward <alan.hayward@arm.com>
7276 David Sherwood <david.sherwood@arm.com>
7278 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7280 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7281 or if measurement_type is polynomial.
7282 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7283 * combine.c (make_extraction): Likewise.
7284 * dse.c (find_shift_sequence): Likewise.
7285 * dwarf2out.c (mem_loc_descriptor): Likewise.
7286 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7287 (extract_bit_field, extract_low_bits): Likewise.
7288 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7289 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7290 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7291 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7292 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7293 * reload.c (find_reloads): Likewise.
7294 * reload1.c (alter_reg): Likewise.
7295 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7296 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7297 * tree-if-conv.c (predicate_mem_writes): Likewise.
7298 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7299 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7300 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7301 * valtrack.c (dead_debug_insert_temp): Likewise.
7302 * varasm.c (mergeable_constant_section): Likewise.
7303 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7306 Alan Hayward <alan.hayward@arm.com>
7307 David Sherwood <david.sherwood@arm.com>
7309 * expr.c (expand_assignment): Cope with polynomial mode sizes
7310 when assigning to a CONCAT.
7312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7313 Alan Hayward <alan.hayward@arm.com>
7314 David Sherwood <david.sherwood@arm.com>
7316 * machmode.h (mode_precision): Change from unsigned short to
7318 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7320 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7321 or if measurement_type is not polynomial.
7322 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7323 in which the mode is already known to be a scalar_int_mode.
7324 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7325 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7327 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7328 for GET_MODE_PRECISION.
7329 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7330 for GET_MODE_PRECISION.
7331 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7333 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7334 (expand_field_assignment, make_extraction): Likewise.
7335 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7336 (get_last_value): Likewise.
7337 * convert.c (convert_to_integer_1): Likewise.
7338 * cse.c (cse_insn): Likewise.
7339 * expr.c (expand_expr_real_1): Likewise.
7340 * lra-constraints.c (simplify_operand_subreg): Likewise.
7341 * optabs-query.c (can_atomic_load_p): Likewise.
7342 * optabs.c (expand_atomic_load): Likewise.
7343 (expand_atomic_store): Likewise.
7344 * ree.c (combine_reaching_defs): Likewise.
7345 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7346 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7347 * tree.h (type_has_mode_precision_p): Likewise.
7348 * ubsan.c (instrument_si_overflow): Likewise.
7350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7351 Alan Hayward <alan.hayward@arm.com>
7352 David Sherwood <david.sherwood@arm.com>
7354 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7355 polynomial numbers of units.
7356 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7357 (valid_vector_subparts_p): New function.
7358 (build_vector_type): Remove temporary shim and take the number
7359 of units as a poly_uint64 rather than an int.
7360 (build_opaque_vector_type): Take the number of units as a
7361 poly_uint64 rather than an int.
7362 * tree.c (build_vector_from_ctor): Handle polynomial
7363 TYPE_VECTOR_SUBPARTS.
7364 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7365 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7366 (build_vector_from_val): If the number of units is variable,
7367 use build_vec_duplicate_cst for constant operands and
7368 VEC_DUPLICATE_EXPR otherwise.
7369 (make_vector_type): Remove temporary is_constant ().
7370 (build_vector_type, build_opaque_vector_type): Take the number of
7371 units as a poly_uint64 rather than an int.
7372 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7374 * cfgexpand.c (expand_debug_expr): Likewise.
7375 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7376 (store_constructor, expand_expr_real_1): Likewise.
7377 (const_scalar_mask_from_tree): Likewise.
7378 * fold-const-call.c (fold_const_reduction): Likewise.
7379 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7380 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7381 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7382 (fold_relational_const): Likewise.
7383 (native_interpret_vector): Likewise. Change the size from an
7384 int to an unsigned int.
7385 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7386 TYPE_VECTOR_SUBPARTS.
7387 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7388 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7389 duplicating a non-constant operand into a variable-length vector.
7390 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7391 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7392 * ipa-icf.c (sem_variable::equals): Likewise.
7393 * match.pd: Likewise.
7394 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7395 * print-tree.c (print_node): Likewise.
7396 * stor-layout.c (layout_type): Likewise.
7397 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7398 * tree-cfg.c (verify_gimple_comparison): Likewise.
7399 (verify_gimple_assign_binary): Likewise.
7400 (verify_gimple_assign_ternary): Likewise.
7401 (verify_gimple_assign_single): Likewise.
7402 * tree-pretty-print.c (dump_generic_node): Likewise.
7403 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7404 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7405 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7406 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7407 (vect_shift_permute_load_chain): Likewise.
7408 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7409 (expand_vector_condition, optimize_vector_constructor): Likewise.
7410 (lower_vec_perm, get_compute_type): Likewise.
7411 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7412 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7413 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7414 (vect_recog_mask_conversion_pattern): Likewise.
7415 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7416 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7417 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7418 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7419 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7420 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7421 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7422 (supportable_widening_operation): Likewise.
7423 (supportable_narrowing_operation): Likewise.
7424 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7426 * varasm.c (output_constant): Likewise.
7428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7429 Alan Hayward <alan.hayward@arm.com>
7430 David Sherwood <david.sherwood@arm.com>
7432 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7433 so that both the length == 3 and length != 3 cases set up their
7434 own permute vectors. Add comments explaining why we know the
7435 number of elements is constant.
7436 (vect_permute_load_chain): Likewise.
7438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7439 Alan Hayward <alan.hayward@arm.com>
7440 David Sherwood <david.sherwood@arm.com>
7442 * machmode.h (mode_nunits): Change from unsigned char to
7444 (ONLY_FIXED_SIZE_MODES): New macro.
7445 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7446 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7447 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7449 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7450 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7451 or if measurement_type is not polynomial.
7452 * genmodes.c (ZERO_COEFFS): New macro.
7453 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7455 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7456 Use ZERO_COEFFS when emitting initializers.
7457 * data-streamer.h (bp_pack_poly_value): New function.
7458 (bp_unpack_poly_value): Likewise.
7459 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7460 for GET_MODE_NUNITS.
7461 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7462 for GET_MODE_NUNITS.
7463 * tree.c (make_vector_type): Remove temporary shim and make
7464 the real function take the number of units as a poly_uint64
7466 (build_vector_type_for_mode): Handle polynomial nunits.
7467 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7468 * emit-rtl.c (const_vec_series_p_1): Likewise.
7469 (gen_rtx_CONST_VECTOR): Likewise.
7470 * fold-const.c (test_vec_duplicate_folding): Likewise.
7471 * genrecog.c (validate_pattern): Likewise.
7472 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7473 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7474 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7475 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7476 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7477 * rtlanal.c (subreg_get_info): Likewise.
7478 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7479 (vect_grouped_load_supported): Likewise.
7480 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7481 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7482 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7483 (simplify_const_unary_operation, simplify_binary_operation_1)
7484 (simplify_const_binary_operation, simplify_ternary_operation)
7485 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7486 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7487 instead of CONST_VECTOR_NUNITS.
7488 * varasm.c (output_constant_pool_2): Likewise.
7489 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7490 explicit-encoded elements in the XVEC for variable-length vectors.
7492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7494 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7497 Alan Hayward <alan.hayward@arm.com>
7498 David Sherwood <david.sherwood@arm.com>
7500 * coretypes.h (fixed_size_mode): Declare.
7501 (fixed_size_mode_pod): New typedef.
7502 * builtins.h (target_builtins::x_apply_args_mode)
7503 (target_builtins::x_apply_result_mode): Change type to
7504 fixed_size_mode_pod.
7505 * builtins.c (apply_args_size, apply_result_size, result_vector)
7506 (expand_builtin_apply_args_1, expand_builtin_apply)
7507 (expand_builtin_return): Update accordingly.
7509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7511 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7512 * cselib.c (cselib_hash_rtx): Likewise.
7513 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7514 CONST_VECTOR encoding.
7516 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7517 Jeff Law <law@redhat.com>
7520 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7521 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7522 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7523 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7526 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7527 explicitly probe *sp in a noreturn function if there were any callee
7528 register saves or frame pointer is needed.
7530 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7533 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7534 BLKmode for ternary, binary or unary expressions.
7537 * var-tracking.c (delete_vta_debug_insn): New inline function.
7538 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7539 insns from get_insns () to NULL instead of each bb separately.
7540 Use delete_vta_debug_insn. No longer static.
7541 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7542 delete_vta_debug_insns callers.
7543 * rtl.h (delete_vta_debug_insns): Declare.
7544 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7545 instead of variable_tracking_main.
7547 2018-01-03 Martin Sebor <msebor@redhat.com>
7549 PR tree-optimization/83603
7550 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7551 arguments past the endof the argument list in functions declared
7552 without a prototype.
7553 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7554 Avoid checking when arguments are null.
7556 2018-01-03 Martin Sebor <msebor@redhat.com>
7559 * doc/extend.texi (attribute const): Fix a typo.
7560 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7561 issuing -Wsuggest-attribute for void functions.
7563 2018-01-03 Martin Sebor <msebor@redhat.com>
7565 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7566 offset_int::from instead of wide_int::to_shwi.
7567 (maybe_diag_overlap): Remove assertion.
7568 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7569 * gimple-ssa-sprintf.c (format_directive): Same.
7570 (parse_directive): Same.
7571 (sprintf_dom_walker::compute_format_length): Same.
7572 (try_substitute_return_value): Same.
7574 2018-01-03 Jeff Law <law@redhat.com>
7577 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7578 non-constant residual for zero at runtime and avoid probing in
7579 that case. Reorganize code for trailing problem to mirror handling
7582 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7584 PR tree-optimization/83501
7585 * tree-ssa-strlen.c (get_string_cst): New.
7586 (handle_char_store): Call get_string_cst.
7588 2018-01-03 Martin Liska <mliska@suse.cz>
7590 PR tree-optimization/83593
7591 * tree-ssa-strlen.c: Include tree-cfg.h.
7592 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7593 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7594 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7596 (strlen_dom_walker::before_dom_children): Call
7597 gimple_purge_dead_eh_edges. Dump tranformation with details
7599 (strlen_dom_walker::before_dom_children): Update call by adding
7600 new argument cleanup_eh.
7601 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7603 2018-01-03 Martin Liska <mliska@suse.cz>
7606 * cif-code.def (VARIADIC_THUNK): New enum value.
7607 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7610 2018-01-03 Jan Beulich <jbeulich@suse.com>
7612 * sse.md (mov<mode>_internal): Tighten condition for when to use
7613 vmovdqu<ssescalarsize> for TI and OI modes.
7615 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7617 Update copyright years.
7619 2018-01-03 Martin Liska <mliska@suse.cz>
7622 * ipa-visibility.c (function_and_variable_visibility): Skip
7623 functions with noipa attribure.
7625 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7627 * gcc.c (process_command): Update copyright notice dates.
7628 * gcov-dump.c (print_version): Ditto.
7629 * gcov.c (print_version): Ditto.
7630 * gcov-tool.c (print_version): Ditto.
7631 * gengtype.c (create_file): Ditto.
7632 * doc/cpp.texi: Bump @copying's copyright year.
7633 * doc/cppinternals.texi: Ditto.
7634 * doc/gcc.texi: Ditto.
7635 * doc/gccint.texi: Ditto.
7636 * doc/gcov.texi: Ditto.
7637 * doc/install.texi: Ditto.
7638 * doc/invoke.texi: Ditto.
7640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7642 * vector-builder.h (vector_builder::m_full_nelts): Change from
7643 unsigned int to poly_uint64.
7644 (vector_builder::full_nelts): Update prototype accordingly.
7645 (vector_builder::new_vector): Likewise.
7646 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7647 (vector_builder::operator ==): Likewise.
7648 (vector_builder::finalize): Likewise.
7649 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7650 Take the number of elements as a poly_uint64 rather than an
7652 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7653 from unsigned int to poly_uint64.
7654 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7655 (vec_perm_indices::new_vector): Likewise.
7656 (vec_perm_indices::length): Likewise.
7657 (vec_perm_indices::nelts_per_input): Likewise.
7658 (vec_perm_indices::input_nelts): Likewise.
7659 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7660 number of elements per input as a poly_uint64 rather than an
7661 unsigned int. Use the original encoding for variable-length
7662 vectors, rather than clamping each individual element.
7663 For the second and subsequent elements in each pattern,
7664 clamp the step and base before clamping their sum.
7665 (vec_perm_indices::series_p): Handle polynomial element counts.
7666 (vec_perm_indices::all_in_range_p): Likewise.
7667 (vec_perm_indices_to_tree): Likewise.
7668 (vec_perm_indices_to_rtx): Likewise.
7669 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7670 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7671 (tree_vector_builder::new_binary_operation): Handle polynomial
7672 element counts. Return false if we need to know the number
7673 of elements at compile time.
7674 * fold-const.c (fold_vec_perm): Punt if the number of elements
7675 isn't known at compile time.
7677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7679 * vec-perm-indices.h (vec_perm_builder): Change element type
7680 from HOST_WIDE_INT to poly_int64.
7681 (vec_perm_indices::element_type): Update accordingly.
7682 (vec_perm_indices::clamp): Handle polynomial element_types.
7683 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7684 (vec_perm_indices::all_in_range_p): Likewise.
7685 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7687 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7688 polynomial vec_perm_indices element types.
7689 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7690 * fold-const.c (fold_vec_perm): Likewise.
7691 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7692 * tree-vect-generic.c (lower_vec_perm): Likewise.
7693 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7694 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7695 element type to HOST_WIDE_INT.
7697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7698 Alan Hayward <alan.hayward@arm.com>
7699 David Sherwood <david.sherwood@arm.com>
7701 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7702 rather than an int. Use plus_constant.
7703 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7704 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7707 Alan Hayward <alan.hayward@arm.com>
7708 David Sherwood <david.sherwood@arm.com>
7710 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7711 a HOST_WIDE_INT to a poly_int64.
7713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7714 Alan Hayward <alan.hayward@arm.com>
7715 David Sherwood <david.sherwood@arm.com>
7717 * calls.c (load_register_parameters): Cope with polynomial
7718 mode sizes. Require a constant size for BLKmode parameters
7719 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7720 forces a parameter to be padded at the lsb end in order to
7721 fill a complete number of words, require the parameter size
7722 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
7729 from unsigned int to poly_uint64_pod.
7730 (alter_reg): Treat mode sizes as polynomial.
7732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7733 Alan Hayward <alan.hayward@arm.com>
7734 David Sherwood <david.sherwood@arm.com>
7736 * reload.c (complex_word_subreg_p): New function.
7737 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
7744 operands whose sizes aren't ordered.
7745 (match_reload): Refer to this check here.
7747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7748 Alan Hayward <alan.hayward@arm.com>
7749 David Sherwood <david.sherwood@arm.com>
7751 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7752 that the mode size is in the set {1, 2, 4, 8, 16}.
7754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7755 Alan Hayward <alan.hayward@arm.com>
7756 David Sherwood <david.sherwood@arm.com>
7758 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7759 Use plus_constant instead of gen_rtx_PLUS.
7761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7762 Alan Hayward <alan.hayward@arm.com>
7763 David Sherwood <david.sherwood@arm.com>
7765 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7766 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7767 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7768 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7769 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7770 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7771 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7772 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7773 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7774 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7776 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7777 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7778 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7779 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7780 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7781 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7782 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7783 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7784 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7785 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7787 * expr.c (emit_move_resolve_push): Treat the input and result
7788 of PUSH_ROUNDING as a poly_int64.
7789 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7790 (emit_push_insn): Likewise.
7791 * lra-eliminations.c (mark_not_eliminable): Likewise.
7792 * recog.c (push_operand): Likewise.
7793 * reload1.c (elimination_effects): Likewise.
7794 * rtlanal.c (nonzero_bits1): Likewise.
7795 * calls.c (store_one_arg): Likewise. Require the padding to be
7796 known at compile time.
7798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7799 Alan Hayward <alan.hayward@arm.com>
7800 David Sherwood <david.sherwood@arm.com>
7802 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7803 Use plus_constant instead of gen_rtx_PLUS.
7805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7806 Alan Hayward <alan.hayward@arm.com>
7807 David Sherwood <david.sherwood@arm.com>
7809 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
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_1): Use tree_to_poly_uint64
7817 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7818 via stack temporaries. Treat the mode size as polynomial too.
7820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7821 Alan Hayward <alan.hayward@arm.com>
7822 David Sherwood <david.sherwood@arm.com>
7824 * expr.c (expand_expr_real_2): When handling conversions involving
7825 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7826 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7827 as a poly_uint64 too.
7829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7830 Alan Hayward <alan.hayward@arm.com>
7831 David Sherwood <david.sherwood@arm.com>
7833 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7836 Alan Hayward <alan.hayward@arm.com>
7837 David Sherwood <david.sherwood@arm.com>
7839 * combine.c (can_change_dest_mode): Handle polynomial
7840 REGMODE_NATURAL_SIZE.
7841 * expmed.c (store_bit_field_1): Likewise.
7842 * expr.c (store_constructor): Likewise.
7843 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7844 and polynomial REGMODE_NATURAL_SIZE.
7845 (gen_lowpart_common): Likewise.
7846 * reginfo.c (record_subregs_of_mode): Likewise.
7847 * rtlanal.c (read_modify_subreg_p): Likewise.
7849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7850 Alan Hayward <alan.hayward@arm.com>
7851 David Sherwood <david.sherwood@arm.com>
7853 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7854 numbers of elements.
7856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7857 Alan Hayward <alan.hayward@arm.com>
7858 David Sherwood <david.sherwood@arm.com>
7860 * match.pd: Cope with polynomial numbers of vector elements.
7862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7863 Alan Hayward <alan.hayward@arm.com>
7864 David Sherwood <david.sherwood@arm.com>
7866 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7867 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
7874 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7875 (ipa_simd_modify_function_body): Likewise.
7877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7878 Alan Hayward <alan.hayward@arm.com>
7879 David Sherwood <david.sherwood@arm.com>
7881 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7882 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7883 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7884 (expand_vector_condition, vector_element): Likewise.
7885 (subparts_gt): New function.
7886 (get_compute_type): Use subparts_gt.
7887 (count_type_subparts): Delete.
7888 (expand_vector_operations_1): Use subparts_gt instead of
7889 count_type_subparts.
7891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7892 Alan Hayward <alan.hayward@arm.com>
7893 David Sherwood <david.sherwood@arm.com>
7895 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7896 (vect_compile_time_alias): ...this new function. Do the calculation
7897 on poly_ints rather than trees.
7898 (vect_prune_runtime_alias_test_list): Update call accordingly.
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_build_slp_tree_1): Handle polynomial
7906 (vect_schedule_slp_instance): Likewise.
7908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7909 Alan Hayward <alan.hayward@arm.com>
7910 David Sherwood <david.sherwood@arm.com>
7912 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7913 constant and extern definitions for variable-length vectors.
7914 (vect_get_constant_vectors): Note that the number of units
7915 is known to be constant.
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 (vectorizable_conversion): Treat the number
7922 of units as polynomial. Choose between WIDE and NARROW based
7925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7926 Alan Hayward <alan.hayward@arm.com>
7927 David Sherwood <david.sherwood@arm.com>
7929 * tree-vect-stmts.c (simd_clone_subparts): New function.
7930 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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 (vectorizable_call): Treat the number of
7937 vectors as polynomial. Use build_index_vector for
7940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7941 Alan Hayward <alan.hayward@arm.com>
7942 David Sherwood <david.sherwood@arm.com>
7944 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7945 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7946 for variable-length vectors.
7947 (vectorizable_mask_load_store): Treat the number of units as
7948 polynomial, asserting that it is constant if the condition has
7949 already been enforced.
7950 (vectorizable_store, vectorizable_load): Likewise.
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_live_operation): Treat the number
7957 of units as polynomial. Punt if we can't tell at compile time
7958 which vector contains the final result.
7960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7961 Alan Hayward <alan.hayward@arm.com>
7962 David Sherwood <david.sherwood@arm.com>
7964 * tree-vect-loop.c (vectorizable_induction): Treat the number
7965 of units as polynomial. Punt on SLP inductions. Use an integer
7966 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7967 cast of such a series for variable-length floating-point
7970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7971 Alan Hayward <alan.hayward@arm.com>
7972 David Sherwood <david.sherwood@arm.com>
7974 * tree.h (build_index_vector): Declare.
7975 * tree.c (build_index_vector): New function.
7976 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7977 of units as polynomial, forcibly converting it to a constant if
7978 vectorizable_reduction has already enforced the condition.
7979 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7980 to create a {1,2,3,...} vector.
7981 (vectorizable_reduction): Treat the number of units as polynomial.
7982 Choose vectype_in based on the largest scalar element size rather
7983 than the smallest number of units. Enforce the restrictions
7986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7987 Alan Hayward <alan.hayward@arm.com>
7988 David Sherwood <david.sherwood@arm.com>
7990 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7991 number of units as polynomial.
7993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7994 Alan Hayward <alan.hayward@arm.com>
7995 David Sherwood <david.sherwood@arm.com>
7997 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7998 * target.def (autovectorize_vector_sizes): Return the vector sizes
7999 by pointer, using vector_sizes rather than a bitmask.
8000 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8001 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8002 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8004 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8005 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8006 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8007 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8008 * omp-general.c (omp_max_vf): Likewise.
8009 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8010 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8011 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8012 * tree-vect-slp.c (vect_slp_bb): Likewise.
8013 * doc/tm.texi: Regenerate.
8014 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8016 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8017 the vector size as a poly_uint64 rather than an unsigned int.
8018 (current_vector_size): Change from an unsigned int to a poly_uint64.
8019 (get_vectype_for_scalar_type): Update accordingly.
8020 * tree.h (build_truth_vector_type): Take the size and number of
8021 units as a poly_uint64 rather than an unsigned int.
8022 (build_vector_type): Add a temporary overload that takes
8023 the number of units as a poly_uint64 rather than an unsigned int.
8024 * tree.c (make_vector_type): Likewise.
8025 (build_truth_vector_type): Take the number of units as a poly_uint64
8026 rather than an unsigned int.
8028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8029 Alan Hayward <alan.hayward@arm.com>
8030 David Sherwood <david.sherwood@arm.com>
8032 * target.def (get_mask_mode): Take the number of units and length
8033 as poly_uint64s rather than unsigned ints.
8034 * targhooks.h (default_get_mask_mode): Update accordingly.
8035 * targhooks.c (default_get_mask_mode): Likewise.
8036 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8037 * doc/tm.texi: Regenerate.
8039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8040 Alan Hayward <alan.hayward@arm.com>
8041 David Sherwood <david.sherwood@arm.com>
8043 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8044 * omp-general.c (omp_max_vf): Likewise.
8045 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8046 (expand_omp_simd): Handle polynomial safelen.
8047 * omp-low.c (omplow_simd_context): Add a default constructor.
8048 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8049 (lower_rec_simd_input_clauses): Update accordingly.
8050 (lower_rec_input_clauses): Likewise.
8052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8053 Alan Hayward <alan.hayward@arm.com>
8054 David Sherwood <david.sherwood@arm.com>
8056 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8057 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8058 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8059 (vect_analyze_slp_cost): Likewise.
8060 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8061 (vect_model_load_cost): Likewise.
8063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8064 Alan Hayward <alan.hayward@arm.com>
8065 David Sherwood <david.sherwood@arm.com>
8067 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8068 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8069 from an unsigned int * to a poly_uint64_pod *.
8070 (calculate_unrolling_factor): New function.
8071 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8074 Alan Hayward <alan.hayward@arm.com>
8075 David Sherwood <david.sherwood@arm.com>
8077 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8078 from an unsigned int to a poly_uint64.
8079 (_loop_vec_info::slp_unrolling_factor): Likewise.
8080 (_loop_vec_info::vectorization_factor): Change from an int
8082 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8083 (vect_get_num_vectors): New function.
8084 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8085 (vect_get_num_copies): Use vect_get_num_vectors.
8086 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8087 to an unsigned int *.
8088 (vect_analyze_data_refs): Change min_vf from an int * to a
8090 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8091 than an unsigned HOST_WIDE_INT.
8092 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8093 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8094 to an unsigned int *.
8095 (vect_analyze_data_ref_dependences): Likewise.
8096 (vect_compute_data_ref_alignment): Handle polynomial vf.
8097 (vect_enhance_data_refs_alignment): Likewise.
8098 (vect_prune_runtime_alias_test_list): Likewise.
8099 (vect_shift_permute_load_chain): Likewise.
8100 (vect_supportable_dr_alignment): Likewise.
8101 (dependence_distance_ge_vf): Take the vectorization factor as a
8102 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8103 (vect_analyze_data_refs): Change min_vf from an int * to a
8105 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8106 vfm1 as a poly_uint64 rather than an int. Make the same change
8107 for the returned bound_scalar.
8108 (vect_gen_vector_loop_niters): Handle polynomial vf.
8109 (vect_do_peeling): Likewise. Update call to
8110 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8111 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8113 * tree-vect-loop.c (vect_determine_vectorization_factor)
8114 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8115 (vect_get_known_peeling_cost): Likewise.
8116 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8117 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8118 (vect_transform_loop): Likewise. Use the lowest possible VF when
8119 updating the upper bounds of the loop.
8120 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8122 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8123 polynomial unroll factors.
8124 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8125 (vect_make_slp_decision): Likewise.
8126 (vect_supported_load_permutation_p): Likewise, and polynomial
8128 (vect_analyze_slp_cost): Handle polynomial vf.
8129 (vect_slp_analyze_node_operations): Likewise.
8130 (vect_slp_analyze_bb_1): Likewise.
8131 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8132 than an unsigned HOST_WIDE_INT.
8133 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8134 (vectorizable_load): Handle polynomial vf.
8135 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8137 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8140 Alan Hayward <alan.hayward@arm.com>
8141 David Sherwood <david.sherwood@arm.com>
8143 * match.pd: Handle bit operations involving three constants
8144 and try to fold one pair.
8146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8148 * tree-vect-loop-manip.c: Include gimple-fold.h.
8149 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8150 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8151 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8152 Add a path that uses a step of VF instead of 1, but disable it
8154 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8155 and niters_no_overflow parameters. Update calls to
8156 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8157 Create a new SSA name if the latter choses to use a ste other
8158 than zero, and return it via niters_vector_mult_vf_var.
8159 * tree-vect-loop.c (vect_transform_loop): Update calls to
8160 vect_do_peeling, vect_gen_vector_loop_niters and
8161 slpeel_make_loop_iterate_ntimes.
8162 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8163 (vect_gen_vector_loop_niters): Update declarations after above changes.
8165 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8167 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8168 128-bit round to integer instructions.
8169 (ceil<mode>2): Likewise.
8170 (btrunc<mode>2): Likewise.
8171 (round<mode>2): Likewise.
8173 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8175 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8176 unaligned VSX load/store on P8/P9.
8177 (expand_block_clear): Allow the use of unaligned VSX
8178 load/store on P8/P9.
8180 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8182 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8184 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8185 swap associated with both a load and a store.
8187 2018-01-02 Andrew Waterman <andrew@sifive.com>
8189 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8190 * config/riscv/riscv.md (clear_cache): Use it.
8192 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8194 * web.c: Remove out-of-date comment.
8196 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8198 * expr.c (fixup_args_size_notes): Check that any existing
8199 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8200 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8201 (emit_single_push_insn): ...here.
8203 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8205 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8206 (const_vector_encoded_nelts): New function.
8207 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8208 (const_vector_int_elt, const_vector_elt): Declare.
8209 * emit-rtl.c (const_vector_int_elt_1): New function.
8210 (const_vector_elt): Likewise.
8211 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8212 of CONST_VECTOR_ELT.
8214 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8216 * expr.c: Include rtx-vector-builder.h.
8217 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8218 directly on the tree encoding.
8219 (const_vector_from_tree): Likewise.
8220 * optabs.c: Include rtx-vector-builder.h.
8221 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8222 sequence of "u" values.
8223 * vec-perm-indices.c: Include rtx-vector-builder.h.
8224 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8225 directly on the vec_perm_indices encoding.
8227 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8229 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8230 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8231 * rtx-vector-builder.h: New file.
8232 * rtx-vector-builder.c: Likewise.
8233 * rtl.h (rtx_def::u2): Add a const_vector field.
8234 (CONST_VECTOR_NPATTERNS): New macro.
8235 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8236 (CONST_VECTOR_DUPLICATE_P): Likewise.
8237 (CONST_VECTOR_STEPPED_P): Likewise.
8238 (CONST_VECTOR_ENCODED_ELT): Likewise.
8239 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8240 (unwrap_const_vec_duplicate): Likewise.
8241 (const_vec_series_p): Check for a non-duplicated vector encoding.
8242 Say that the function only returns true for integer vectors.
8243 * emit-rtl.c: Include rtx-vector-builder.h.
8244 (gen_const_vec_duplicate_1): Delete.
8245 (gen_const_vector): Call gen_const_vec_duplicate instead of
8246 gen_const_vec_duplicate_1.
8247 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8248 (gen_const_vec_duplicate): Use rtx_vector_builder.
8249 (gen_const_vec_series): Likewise.
8250 (gen_rtx_CONST_VECTOR): Likewise.
8251 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8252 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8253 Build a new vector rather than modifying a CONST_VECTOR in-place.
8254 (handle_special_swappables): Update call accordingly.
8255 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8256 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8257 Build a new vector rather than modifying a CONST_VECTOR in-place.
8258 (handle_special_swappables): Update call accordingly.
8260 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8262 * simplify-rtx.c (simplify_const_binary_operation): Use
8263 CONST_VECTOR_ELT instead of XVECEXP.
8265 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8267 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8268 the selector elements to be different from the data elements
8269 if the selector is a VECTOR_CST.
8270 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8271 ssizetype for the selector.
8273 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8275 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8276 before testing each element individually.
8277 * tree-vect-generic.c (lower_vec_perm): Likewise.
8279 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8281 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8282 * selftest-run-tests.c (selftest::run_tests): Call it.
8283 * vector-builder.h (vector_builder::operator ==): New function.
8284 (vector_builder::operator !=): Likewise.
8285 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8286 (vec_perm_indices::all_from_input_p): New function.
8287 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8288 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8289 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8290 instead of reading the VECTOR_CST directly. Detect whether both
8291 vector inputs are the same before constructing the vec_perm_indices,
8292 and update the number of inputs argument accordingly. Use the
8293 utility functions added above. Only construct sel2 if we need to.
8295 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8297 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8298 the broadcast of the low byte.
8299 (expand_mult_highpart): Use an explicit encoding for the permutes.
8300 * optabs-query.c (can_mult_highpart_p): Likewise.
8301 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8302 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8303 (vectorizable_bswap): Likewise.
8304 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8305 explicit encoding for the power-of-2 permutes.
8306 (vect_permute_store_chain): Likewise.
8307 (vect_grouped_load_supported): Likewise.
8308 (vect_permute_load_chain): Likewise.
8310 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8312 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8313 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8314 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8315 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8316 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8317 (vect_gen_perm_mask_any): Likewise.
8319 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8321 * int-vector-builder.h: New file.
8322 * vec-perm-indices.h: Include int-vector-builder.h.
8323 (vec_perm_indices): Redefine as an int_vector_builder.
8324 (auto_vec_perm_indices): Delete.
8325 (vec_perm_builder): Redefine as a stand-alone class.
8326 (vec_perm_indices::vec_perm_indices): New function.
8327 (vec_perm_indices::clamp): Likewise.
8328 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8329 (vec_perm_indices::new_vector): New function.
8330 (vec_perm_indices::new_expanded_vector): Update for new
8331 vec_perm_indices class.
8332 (vec_perm_indices::rotate_inputs): New function.
8333 (vec_perm_indices::all_in_range_p): Operate directly on the
8334 encoded form, without computing elided elements.
8335 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8336 encoding. Update for new vec_perm_indices class.
8337 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8338 the given vec_perm_builder.
8339 (expand_vec_perm_var): Update vec_perm_builder constructor.
8340 (expand_mult_highpart): Use vec_perm_builder instead of
8341 auto_vec_perm_indices.
8342 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8343 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8344 or double series encoding as appropriate.
8345 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8346 vec_perm_indices instead of auto_vec_perm_indices.
8347 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8348 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8349 (vect_permute_store_chain): Likewise.
8350 (vect_grouped_load_supported): Likewise.
8351 (vect_permute_load_chain): Likewise.
8352 (vect_shift_permute_load_chain): Likewise.
8353 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8354 (vect_transform_slp_perm_load): Likewise.
8355 (vect_schedule_slp_instance): Likewise.
8356 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8357 (vectorizable_mask_load_store): Likewise.
8358 (vectorizable_bswap): Likewise.
8359 (vectorizable_store): Likewise.
8360 (vectorizable_load): Likewise.
8361 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8362 vec_perm_indices instead of auto_vec_perm_indices. Use
8363 tree_to_vec_perm_builder to read the vector from a tree.
8364 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8365 vec_perm_builder instead of a vec_perm_indices.
8366 (have_whole_vector_shift): Use vec_perm_builder and
8367 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8368 truncation to calc_vec_perm_mask_for_shift.
8369 (vect_create_epilog_for_reduction): Likewise.
8370 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8371 from auto_vec_perm_indices to vec_perm_indices.
8372 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8373 instead of changing individual elements.
8374 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8375 the vector in d.perm.
8376 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8377 from auto_vec_perm_indices to vec_perm_indices.
8378 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8379 instead of changing individual elements.
8380 (arm_vectorize_vec_perm_const): Use new_vector to install
8381 the vector in d.perm.
8382 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8383 Update vec_perm_builder constructor.
8384 (rs6000_expand_interleave): Likewise.
8385 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8386 (rs6000_expand_interleave): Likewise.
8388 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8390 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8391 to qimode could truncate the indices.
8392 * optabs.c (expand_vec_perm_var): Likewise.
8394 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8396 * Makefile.in (OBJS): Add vec-perm-indices.o.
8397 * vec-perm-indices.h: New file.
8398 * vec-perm-indices.c: Likewise.
8399 * target.h (vec_perm_indices): Replace with a forward class
8401 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8402 * optabs.h: Include vec-perm-indices.h.
8403 (expand_vec_perm): Delete.
8404 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8405 (expand_vec_perm_const): Declare.
8406 * target.def (vec_perm_const_ok): Replace with...
8407 (vec_perm_const): ...this new hook.
8408 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8409 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8410 * doc/tm.texi: Regenerate.
8411 * optabs.def (vec_perm_const): Delete.
8412 * doc/md.texi (vec_perm_const): Likewise.
8413 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8414 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8415 expand_vec_perm for constant permutation vectors. Assert that
8416 the mode of variable permutation vectors is the integer equivalent
8417 of the mode that is being permuted.
8418 * optabs-query.h (selector_fits_mode_p): Declare.
8419 * optabs-query.c: Include vec-perm-indices.h.
8420 (selector_fits_mode_p): New function.
8421 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8422 is defined, instead of checking whether the vec_perm_const_optab
8423 exists. Use targetm.vectorize.vec_perm_const instead of
8424 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8425 fit in the vector mode before using a variable permute.
8426 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8427 vec_perm_indices instead of an rtx.
8428 (expand_vec_perm): Replace with...
8429 (expand_vec_perm_const): ...this new function. Take the selector
8430 as a vec_perm_indices rather than an rtx. Also take the mode of
8431 the selector. Update call to shift_amt_for_vec_perm_mask.
8432 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8433 Use vec_perm_indices::new_expanded_vector to expand the original
8434 selector into bytes. Check whether the indices fit in the vector
8435 mode before using a variable permute.
8436 (expand_vec_perm_var): Make global.
8437 (expand_mult_highpart): Use expand_vec_perm_const.
8438 * fold-const.c: Includes vec-perm-indices.h.
8439 * tree-ssa-forwprop.c: Likewise.
8440 * tree-vect-data-refs.c: Likewise.
8441 * tree-vect-generic.c: Likewise.
8442 * tree-vect-loop.c: Likewise.
8443 * tree-vect-slp.c: Likewise.
8444 * tree-vect-stmts.c: Likewise.
8445 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8447 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8448 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8449 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8450 (aarch64_vectorize_vec_perm_const): ...this new function.
8451 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8452 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8453 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8454 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8455 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8456 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8457 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8459 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8460 check for NEON modes.
8461 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8462 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8463 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8464 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8466 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8467 the old VEC_PERM_CONST conditions.
8468 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8469 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8470 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8471 (ia64_vectorize_vec_perm_const_ok): Merge into...
8472 (ia64_vectorize_vec_perm_const): ...this new function.
8473 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8474 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8475 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8476 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8477 * config/mips/mips.c (mips_expand_vec_perm_const)
8478 (mips_vectorize_vec_perm_const_ok): Merge into...
8479 (mips_vectorize_vec_perm_const): ...this new function.
8480 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8481 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8482 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8483 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8484 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8485 (rs6000_expand_vec_perm_const): Delete.
8486 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8488 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8489 (altivec_expand_vec_perm_const_le): Take each operand individually.
8490 Operate on constant selectors rather than rtxes.
8491 (altivec_expand_vec_perm_const): Likewise. Update call to
8492 altivec_expand_vec_perm_const_le.
8493 (rs6000_expand_vec_perm_const): Delete.
8494 (rs6000_vectorize_vec_perm_const_ok): Delete.
8495 (rs6000_vectorize_vec_perm_const): New function.
8496 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8497 an element count and rtx array.
8498 (rs6000_expand_extract_even): Update call accordingly.
8499 (rs6000_expand_interleave): Likewise.
8500 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8501 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8502 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8503 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8504 (rs6000_expand_vec_perm_const): Delete.
8505 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8506 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8507 (altivec_expand_vec_perm_const_le): Take each operand individually.
8508 Operate on constant selectors rather than rtxes.
8509 (altivec_expand_vec_perm_const): Likewise. Update call to
8510 altivec_expand_vec_perm_const_le.
8511 (rs6000_expand_vec_perm_const): Delete.
8512 (rs6000_vectorize_vec_perm_const_ok): Delete.
8513 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8514 reference to the SPE evmerge intructions.
8515 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8516 an element count and rtx array.
8517 (rs6000_expand_extract_even): Update call accordingly.
8518 (rs6000_expand_interleave): Likewise.
8519 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8520 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8522 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8524 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8526 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8527 vector mode and that that mode matches the mode of the data
8529 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8530 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8531 directly using expand_vec_perm_1 when forcing selectors into
8533 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8535 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8537 * optabs-query.h (can_vec_perm_p): Delete.
8538 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8539 * optabs-query.c (can_vec_perm_p): Split into...
8540 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8541 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8542 particular selector is valid.
8543 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8544 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8545 (vect_grouped_load_supported): Likewise.
8546 (vect_shift_permute_load_chain): Likewise.
8547 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8548 (vect_transform_slp_perm_load): Likewise.
8549 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8550 (vectorizable_bswap): Likewise.
8551 (vect_gen_perm_mask_checked): Likewise.
8552 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8553 implementations of variable permutation vectors into account
8554 when deciding which selector to use.
8555 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8556 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8557 with a false third argument.
8558 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8559 to test whether the constant selector is valid and can_vec_perm_var_p
8560 to test whether a variable selector is valid.
8562 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8564 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8565 * optabs-query.c (can_vec_perm_p): Likewise.
8566 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8567 instead of vec_perm_indices.
8568 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8569 (vect_gen_perm_mask_checked): Likewise,
8570 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8571 (vect_gen_perm_mask_checked): Likewise,
8573 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8575 * optabs-query.h (qimode_for_vec_perm): Declare.
8576 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8577 (qimode_for_vec_perm): ...this new function.
8578 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8580 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8582 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8583 does not have a conditional at the top.
8585 2018-01-02 Richard Biener <rguenther@suse.de>
8587 * ipa-inline.c (big_speedup_p): Fix expression.
8589 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8592 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8595 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8599 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8600 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8601 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8602 cond_taken_branch_cost 3->4.
8604 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8606 PR tree-optimization/83581
8607 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8608 TODO_cleanup_cfg if any changes have been made.
8611 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8612 convert_modes if target mode has the right side, but different mode
8616 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8617 last argument when extracting from CONCAT. If either from_real or
8618 from_imag is NULL, use expansion through memory. If result is not
8619 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8620 the parts directly to inner mode, if even that fails, use expansion
8624 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8625 check for bswap in mode rather than HImode and use that in expand_unop
8628 Copyright (C) 2018 Free Software Foundation, Inc.
8630 Copying and distribution of this file, with or without modification,
8631 are permitted in any medium without royalty provided the copyright
8632 notice and this notice are preserved.