1 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2 Chung-Ju Wu <jasonwucj@gmail.com>
4 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6 * config/nds32/nds32.opt (mvh): New option.
8 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9 Chung-Ju Wu <jasonwucj@gmail.com>
11 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
13 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
14 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18 Chung-Ju Wu <jasonwucj@gmail.com>
20 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
22 * config/nds32/nds32-multiple.md (strlensi): New pattern.
23 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
25 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
26 Kito Cheng <kito.cheng@gmail.com>
27 Chung-Ju Wu <jasonwucj@gmail.com>
29 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
30 UNSPEC_FFMISM and UNSPEC_FLMISM.
31 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
32 for ffb, ffmism and flmism.
33 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
34 (unspec_ffmism): Ditto.
35 (unspec_flmism): Ditto.
36 (nds32_expand_builtin_impl): Check if string extension is available.
37 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
38 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
40 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
43 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
46 * lra-assigns.c (assign_by_spills): Return a flag of reload
47 assignment failure. Do not process the reload assignment
48 failures. Do not spill other reload pseudos if they has the same
50 (lra_assign): Add a return arg. Set up from the result of
51 assign_by_spills call.
52 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
53 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
54 usage_insns if it is not NULL.
55 (spill_hard_reg_in_range): New function.
56 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
57 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
59 (lra_assign): Change prototype.
60 * lra.c (lra): Add code to deal with fails by splitting hard reg
63 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
66 * config/i386/i386.opt: Replace Enforcment with Enforcement.
68 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
71 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
72 (dw_val_node): Add val_symbolic_view.
73 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
74 (symview_upper_bound): New.
75 (new_line_info_table): Initialize symviews_since_reset.
76 (dwarf2out_source_line): Count symviews_since_reset and set
78 (dw_val_equal_p): Handle symview.
79 (add_AT_symview): New.
80 (print_dw_val): Handle symview.
81 (attr_checksum, attr_checksum_ordered): Likewise.
82 (same_dw_val_p, size_of_die): Likewise.
83 (value_format, output_die): Likewise.
84 (add_high_low_attributes): Use add_AT_symview for entry_view.
85 (dwarf2out_finish): Reset symview_upper_bound, clear
88 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
91 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
92 Add strict argument and use it.
93 (rs6000_split_multireg_move): Update for new strict argument.
94 (mem_operand_gpr): Disallow all non-offsettable addresses.
95 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
97 2018-03-09 Jakub Jelinek <jakub@redhat.com>
100 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
101 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
102 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
105 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
106 decl, use remap_type if we want to use the type.
108 2018-03-09 Martin Sebor <msebor@redhat.com>
110 PR tree-optimization/84526
111 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
113 (builtin_access::generic_overlap): Be prepared to handle non-array
116 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
118 PR rtl-optimization/84682
119 * lra-constraints.c (process_address_1): Check is_address flag
120 for address constraints.
121 (process_alt_operands): Likewise.
122 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
123 preprocess_constraints.
124 * recog.h (preprocess_constraints): Add oploc parameter.
126 * recog.c (preprocess_constraints): Test address_operand for
127 CT_ADDRESS constraints.
129 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
132 * lra-assigns.c (assign_by_spills): Return a flag of reload
133 assignment failure. Do not process the reload assignment
134 failures. Do not spill other reload pseudos if they has the same
136 (lra_assign): Add a return arg. Set up from the result of
137 assign_by_spills call.
138 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
139 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
140 usage_insns if it is not NULL.
141 (spill_hard_reg_in_range): New function.
142 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
143 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
145 (lra_assign): Change prototype.
146 * lra.c (lra): Add code to deal with fails by splitting hard reg
149 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
152 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
153 Accept complain bool parameter. Only emit errors if it is true.
154 (arm_parse_cpu_option_name): Likewise.
155 (arm_target_thumb_only): Adjust callers of the above.
156 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
157 prototype to take a default true bool parameter.
158 (arm_parse_arch_option_name): Likewise.
160 2018-03-09 David Malcolm <dmalcolm@redhat.com>
161 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
165 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
166 * configure: Regenerate.
167 * configure.ac ("linker --version-script option"): New.
168 ("linker soname option"): New.
170 2018-03-09 Richard Biener <rguenther@suse.de>
172 PR tree-optimization/84775
173 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
174 immediate uses of predicate stmts and mark them modified.
177 PR tree-optimization/84178
178 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
180 (version_loop_for_if_conversion): Delay update_ssa call.
181 (tree_if_conversion): Delay update_ssa until after predicate
184 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
187 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
188 when the function accesses prior frames.
190 2018-03-08 Jakub Jelinek <jakub@redhat.com>
193 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
194 gen_llsym, otherwise call maybe_gen_llsym.
197 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
198 has ',' character inside of it.
200 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
203 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
204 as clobbering CC_REGNUM.
206 2018-03-08 Richard Biener <rguenther@suse.de>
209 * tree-scalar-evolution.c: Include tree-into-ssa.h.
210 (follow_copies_to_constant): Do not follow SSA names registered
213 2018-03-08 Richard Biener <rguenther@suse.de>
215 PR tree-optimization/84178
216 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
218 (version_loop_for_if_conversion): Delay update_ssa call.
219 (tree_if_conversion): Delay update_ssa until after predicate
222 2018-03-08 David Malcolm <dmalcolm@redhat.com>
224 PR tree-optimization/84178
225 * tree-if-conv.c (release_bb_predicate): Remove the
226 the assertion that the stmts have NULL use_ops.
227 Discard the statements, asserting that they haven't
228 yet been added to a BB.
230 2018-03-08 Richard Biener <rguenther@suse.de>
232 PR tree-optimization/84746
233 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
234 (phi_translate): Pass in destination ANTIC_OUT set.
235 (phi_translate_1): Likewise. For a simplified result lookup
236 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
237 (phi_translate_set): Adjust.
238 (do_pre_regular_insertion): Likewise.
239 (do_pre_partial_partial_insertion): Likewise.
241 2018-03-08 Martin Liska <mliska@suse.cz>
243 PR gcov-profile/84735
244 * doc/gcov.texi: Document usage of profile files.
245 * gcov-io.h: Document changes in the format.
247 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
251 * dwarf2out.c (struct dw_line_info_table): Update comments for
253 (FORCE_RESET_NEXT_VIEW): New.
254 (FORCE_RESETTING_VIEW_P): New.
255 (RESETTING_VIEW_P): Check for -1 too.
256 (ZERO_VIEW_P): Likewise.
257 (new_line_info_table): Force-reset next view.
258 (dwarf2out_begin_function): Likewise.
259 (dwarf2out_source_line): Simplify zero_view_p initialization.
260 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
261 view directly. Omit view when omitting .loc at line 0.
263 2018-03-08 Jakub Jelinek <jakub@redhat.com>
265 PR tree-optimization/84740
266 * tree-switch-conversion.c (process_switch): Call build_constructors
267 only if info.phi_count is non-zero.
269 PR tree-optimization/84739
270 * tree-tailcall.c (find_tail_calls): Check call arguments against
271 DECL_ARGUMENTS (current_function_decl) rather than
272 DECL_ARGUMENTS (func) when checking for tail recursion.
274 2018-03-07 Jakub Jelinek <jakub@redhat.com>
276 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
277 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
278 Volker Reichelt's entry and add entries for people that perform
279 GCC fuzzy testing and report numerous bugs.
281 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
284 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
285 readonly data in sdata, if that is disabled.
286 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
287 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
288 -mreadonly-in-sdata option.
290 2018-03-07 Martin Sebor <msebor@redhat.com>
292 PR tree-optimization/84468
293 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
294 basic block when looking for nul assignment.
296 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
299 * except.h (output_function_exception_table): Adjust prototype.
300 * except.c (output_function_exception_table): Remove FNNAME parameter
301 and add SECTION parameter. Ouput one part of the table at a time.
302 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
303 the first part of the exception table and emit unwind directives.
304 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
305 (i386_pe_seh_cold_init): Likewise.
306 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
307 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
308 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
309 (ix86_output_call_insn): Emit a nop in one more case for SEH.
310 * config/i386/winnt.c: Include except.h.
311 (struct seh_frame_state): Add reg_offset, after_prologue and
312 in_cold_section fields.
313 (i386_pe_seh_end_prologue): Set seh->after_prologue.
314 (i386_pe_seh_cold_init): New function.
315 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
316 to seh->in_cold_section.
317 (seh_emit_push): Record the offset of the push.
318 (seh_emit_save): Record the offet of the save.
319 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
320 Test seh->after_prologue to disregard the epilogue.
321 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
322 (i386_pe_end_cold_function): New function.
324 2018-03-07 Jakub Jelinek <jakub@redhat.com>
327 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
328 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
331 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
333 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
334 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
337 * multiple_target.c: Include tree-inline.h and intl.h.
338 (expand_target_clones): Diagnose and fail if node->definition and
339 !tree_versionable_function_p (node->decl).
341 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
343 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
345 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
346 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
347 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
349 2018-03-06 Jakub Jelinek <jakub@redhat.com>
352 * combine.c (try_combine): Use reg_or_subregno instead of handling
353 just paradoxical SUBREGs and REGs.
355 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
357 * config/arc/arc.c (arc_finalize_pic): Remove function.
358 (arc_must_save_register): We use single base PIC register, remove
359 checks to save/restore the PIC register.
360 (arc_expand_prologue): Likewise.
361 * config/arc/arc-protos.h (arc_set_default_type_attributes):
363 (arc_verify_short): Likewise.
364 (arc_attr_type): Likewise.
365 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
366 (walk_stores): Likewise.
367 (arc_address_cost): Make it static.
368 (arc_verify_short): Likewise.
369 (branch_dest): Likewise.
370 (arc_attr_type): Likewise.
371 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
372 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
373 (arc_final_prescan_insn): Remove inserting the nops due to
374 hardware hazards. It is done in reorg step.
375 (insn_length_variant_t): Remove.
376 (insn_length_parameters_t): Likewise.
377 (arc_insn_length_parameters): Likewise.
378 (arc_get_insn_variants): Likewise.
379 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
381 2018-03-06 Jakub Jelinek <jakub@redhat.com>
384 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
387 PR tree-optimization/84687
388 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
390 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
392 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
394 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
395 Rename to ppc_speculation_barrier.
396 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
397 __builtin_ppc_speculation_barrier.
399 2018-03-05 Jakub Jelinek <jakub@redhat.com>
402 * combine.c (combine_simplify_rtx): Don't try to simplify if
403 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
406 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
408 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
409 to 32 bytes when compiling for POWER9.
411 2018-03-05 Jakub Jelinek <jakub@redhat.com>
414 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
415 regparm >= 3 with no arg reg available also for calls with
416 flag_force_indirect_call. Pass decl to ix86_function_regparm.
419 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
421 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
423 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
426 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
428 2018-03-05 Richard Biener <rguenther@suse.de>
430 PR tree-optimization/84486
431 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
432 When inserting a __builtin_assume_aligned call set the LHS
433 SSA name alignment info accordingly.
435 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
437 PR tree-optimization/84114
438 * config/aarch64/aarch64.c (aarch64_reassociation_width)
439 Avoid reassociation of FLOAT_MODE addition.
441 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
443 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
444 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
445 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
446 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
447 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
448 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
449 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
451 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
452 __builtin_ia32_wbinvd): New builtins.
453 (SPECIAL_ARGS2): New.
454 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
455 (SPECIAL_ARGS2): New.
456 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
457 (ix86_valid_target_attribute_inner_p): Ditto.
458 (ix86_init_mmx_sse_builtins): Add special_args2.
459 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
460 TARGET_WBNOINVD_P): New.
461 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
462 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
463 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
464 * config/i386/immintrin.h (_wbinvd): New intrinsic.
465 * config/i386/pconfigintrin.h: New file.
466 * config/i386/wbnoinvdintrin.h: Ditto.
467 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
468 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
470 2018-03-05 Richard Biener <rguenther@suse.de>
472 PR tree-optimization/84670
473 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
475 (BB_VISITED_WITH_VISITED_SUCCS): New define.
476 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
477 (compute_antic_aux): Only assert the number of values in ANTIC_IN
478 doesn't grow if all successors (recursively) were visited at least
481 2018-03-05 Richard Biener <rguenther@suse.de>
483 PR tree-optimization/84650
484 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
485 if executed in the loop pipeline.
487 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
489 * doc/configfiles.texi (Configuration Files): Move info about
490 conditionalizing $target-protos.h to...
491 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
492 differs from $target-protos.h.
494 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
495 Chung-Ju Wu <jasonwucj@gmail.com>
497 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
498 * config/nds32/nds32-multiple.md (setmemsi): Define.
499 * config/nds32/nds32-memory-manipulation.c
500 (nds32_gen_dup_4_byte_to_word_value): New.
501 (emit_setmem_word_loop): New.
502 (emit_setmem_byte_loop): New.
503 (nds32_expand_setmem_loop): New.
504 (nds32_expand_setmem_loop_v3m): New.
505 (nds32_expand_setmem_unroll): New.
506 (nds32_expand_setmem): New.
508 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
509 Chung-Ju Wu <jasonwucj@gmail.com>
511 * config/nds32/nds32-memory-manipulation.c
512 (nds32_emit_load_store): New.
513 (nds32_emit_post_inc_load_store): New.
514 (nds32_emit_mem_move): New.
515 (nds32_emit_mem_move_block): New.
516 (nds32_expand_movmemsi_loop_unknown_size): New.
517 (nds32_expand_movmemsi_loop_known_size): New.
518 (nds32_expand_movmemsi_loop): New.
519 (nds32_expand_movmemsi_unroll): New.
520 (nds32_expand_movmemqi): Rename ...
521 (nds32_expand_movmemsi): ... to this.
522 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
523 (movmemsi): ... to this.
524 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
525 (nds32_expand_movmemsi): ... to this.
527 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
528 Monk Chiang <sh.chiang04@gmail.com>
529 Chung-Ju Wu <jasonwucj@gmail.com>
531 * config/nds32/nds32-protos.h
532 (nds32_expand_load_multiple): New arguments.
533 (nds32_expand_store_multiple): Ditto.
534 (nds32_valid_multiple_load_store): Rename ...
535 (nds32_valid_multiple_load_store_p): ... to this.
536 * config/nds32/nds32-memory-manipulation.c
537 (nds32_expand_load_multiple): Refine implementation.
538 (nds32_expand_store_multiple): Ditto.
539 * config/nds32/nds32-multiple.md
540 (load_multiple): Update nds32_expand_load_multiple interface.
541 (store_multiple): Update nds32_expand_store_multiple interface.
542 * config/nds32/nds32-predicates.c
543 (nds32_valid_multiple_load_store): Rename ...
544 (nds32_valid_multiple_load_store_p): ... to this and refine
546 * config/nds32/predicates.md
547 (nds32_load_multiple_and_update_address_operation): New predicate.
548 (nds32_store_multiple_and_update_address_operation): New predicate.
550 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
551 Chung-Ju Wu <jasonwucj@gmail.com>
553 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
554 (combo): New attribute.
555 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
557 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
559 * config/nds32/nds32.opt: Change -mcmodel= default value.
561 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
562 Monk Chiang <sh.chiang04@gmail.com>
563 Chung-Ju Wu <jasonwucj@gmail.com>
565 * config/nds32/constants.md (unspec_element): New enum.
566 * config/nds32/constraints.md (Umw): New constraint.
567 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
568 * config/nds32/nds32-intrinsic.md: Likewise.
569 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
570 (nds32_valid_smw_lwm_base_p): New.
571 (nds32_output_smw_single_word): New.
572 (nds32_output_lmw_single_word): New.
573 (nds32_expand_unaligned_load): New.
574 (nds32_expand_unaligned_store): New.
575 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
576 (nds32_output_smw_single_word): Declare.
577 (nds32_output_lmw_single_word): Declare.
578 (nds32_expand_unaligned_load): Declare.
579 (nds32_expand_unaligned_store): Declare.
580 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
581 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
582 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
583 NDS32_BUILTIN_UASTORE_DW.
584 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
587 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
588 Kito Cheng <kito.cheng@gmail.com>
589 Chung-Ju Wu <jasonwucj@gmail.com>
591 * config/nds32/nds32-intrinsic.c
592 (nds32_expand_builtin_null_ftype_reg): Delete.
593 (nds32_expand_builtin_reg_ftype_imm): Ditto.
594 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
595 (nds32_read_argument): New.
596 (nds32_legitimize_target): Ditto.
597 (nds32_legitimize_argument): Ditto.
598 (nds32_check_constant_argument): Ditto.
599 (nds32_expand_unop_builtin): Ditto.
600 (nds32_expand_unopimm_builtin): Ditto.
601 (nds32_expand_binop_builtin): Ditto.
602 (nds32_builtin_decl_impl): Ditto.
603 (builtin_description): Ditto.
604 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
605 (nds32_init_builtins_impl): Ditto.
606 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
607 (nds32_builtin_decl): New.
608 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
609 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
611 2018-03-02 Jeff Law <law@redhat.com>
613 * reorg.c (stop_search_p): Handle DEBUG_INSN.
614 (redundant_insn, fill_simple_delay_slots): Likewise.
615 (fill_slots_from_thread): Likewise.
616 * resource.c (mark_referenced_resources): Likewise.
617 (mark_set_resources, find_dead_or_set_registers): Likewise.
619 2018-03-02 Jakub Jelinek <jakub@redhat.com>
621 * substring-locations.h (format_warning_va): Formatting fix for
623 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
625 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
626 * substring-locations.c: Include intl.h.
627 (format_warning_va): Turned into small wrapper around
628 format_warning_n_va, renamed to ...
629 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
630 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
632 (format_warning_at_substring_n): New function.
633 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
634 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
635 format_warning_at_substring with just a shorter name instead of
636 const function pointer.
637 (fmtwarn_n): New function.
638 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
639 appropriate, get rid of all the fmtstr temporaries, move conditionals
640 with G_() wrapped string literals directly into fmtwarn arguments,
641 cast dir.len to (int), formatting fixes.
643 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
645 * doc/invoke.texi: Remove "Cilk Plus" references.
647 2018-03-02 Jakub Jelinek <jakub@redhat.com>
648 Richard Biener <rguenther@suse.de>
651 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
652 for error or warning attributes if CALL_FROM_THUNK_P is set.
655 2018-03-02 Jakub Jelinek <jakub@redhat.com>
658 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
659 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
662 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
663 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
665 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
666 instead of -1U in last predictors element's probability member.
668 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
671 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
672 arguments if they are comparable.
674 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
676 PR tree-optimization/84634
677 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
678 masks and masked_loop_p with a single loop_masks, making sure it's
679 null for bb vectorization.
681 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
683 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
684 (vect_analyze_data_ref_access): Use loop->safe_len rather than
685 loop->force_vectorize to check whether there is no alias.
687 2018-03-02 Jakub Jelinek <jakub@redhat.com>
690 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
692 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
694 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
695 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
696 instead of a loop around prev_real_insn.
697 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
701 * config/i386/i386.c (ix86_print_operand): Use conditional
702 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
705 2018-03-02 Richard Biener <rguenther@suse.de>
707 PR tree-optimization/84427
708 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
709 (bitmap_set_subtract_values): Rewrite to handle multiple
712 (prune_clobbered_mems): Likewise.
713 (phi_translate): Take edge instead of pred/phiblock.
714 (phi_translate_1): Likewise.
715 (phi_translate_set): Likewise. Insert all translated
716 exprs for a value into the set, keeping possibly multiple
717 expressions per value.
718 (compute_antic_aux): Adjust for phi_translate changes.
719 When intersecting union the expressions and prune those
720 not in the final value set, keeping possibly multiple
721 expressions per value. Do not use value-insertion
722 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
723 all expressions. Add verification that the value-sets
724 only shrink during iteration.
725 (compute_partial_antic_aux): Adjust for the phi_translate changes.
726 (do_pre_regular_insertion): Likewise.
727 (do_pre_partial_partial_insertion): Likewise.
729 2018-03-02 Richard Biener <rguenther@suse.de>
732 * config/darwin.c (saved_debug_info_level): New static global.
733 (darwin_asm_lto_start): Disable debug info generation for LTO out.
734 (darwin_asm_lto_end): Restore debug info generation settings.
736 2018-03-01 Martin Liska <mliska@suse.cz>
739 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
742 2018-03-01 Richard Biener <rguenther@suse.de>
745 * dwarf2out.c (gen_variable_die): Properly handle late VLA
746 type annotation with LTO when debug was disabled at compile-time.
748 2018-03-01 Matthew Fortune <mfortune@gmail.com>
750 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
752 (mips_final_postscan_insn): Likewise.
754 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
756 PR rtl-optimization/84528
757 * alias.c (init_alias_target): Add commentary.
758 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
759 a unique base value if the frame pointer is not eliminated
760 to the stack pointer.
762 2018-03-01 Tom de Vries <tom@codesourcery.com>
764 PR rtl-optimization/83327
765 * lra-int.h (hard_regs_spilled_into): Declare.
766 * lra.c (hard_regs_spilled_into): Define.
767 (init_reg_info): Init hard_regs_spilled_into.
768 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
769 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
770 (process_bb_lives): Handle hard_regs_spilled_into.
771 (lra_create_live_ranges_1): Before doing liveness propagation, clear
772 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
774 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
776 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
777 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
778 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
779 * config/rs6000/aix72.h: New file.
781 2018-02-28 Jakub Jelinek <jakub@redhat.com>
783 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
784 instead of warning_at with conditional singular and plural messages
788 * stor-layout.c (update_alignment_for_field): For
789 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
790 && !DECL_PACKED (field), do the alignment update, just use
791 only desired_align instead of MAX (type_align, desired_align)
793 (place_field): Don't do known_align < desired_align handling
794 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
795 is non-NULL, instead do it after rli->prev_field handling and
796 only if not within a bitfield word. For DECL_PACKED (field)
797 use type_align of BITS_PER_UNIT.
799 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
801 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
802 superfluous parentheses and trailing spaces.
804 2018-02-28 Richard Biener <rguenther@suse.de>
806 PR tree-optimization/84584
807 * graphite-scop-detection.c (scop_detection::add_scop): Discard
808 SCoPs with fake exit edge.
810 2018-02-28 Martin Liska <mliska@suse.cz>
813 * timevar.c (timer::print): Fix format to properly print 100%
816 2018-02-28 Richard Biener <rguenther@suse.de>
819 * genmatch.c (capture_info::walk_match): Do not mark
820 captured expressions without operands as expr_p given
821 they act more like predicates and should be subject to
822 "lost tail" side-effect preserving.
824 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
826 PR rtl-optimization/81611
827 * auto-inc-dec.c (attempt_change): Move dead note from
828 mem_insn if it's the next use of regno
829 (find_address): Take address use of reg holding
830 non-incremented value. Add parm to limit search to the named
832 (merge_in_block): Attempt to use a mem insn that is the next
833 use of the original regno.
835 2018-02-27 Martin Sebor <msebor@redhat.com>
838 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
839 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
841 2018-02-27 Martin Sebor <msebor@redhat.com>
844 * diagnostic-core.h (warning_n, error_n, inform_n): Change
845 n argument to unsigned HOST_WIDE_INT.
846 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
847 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
848 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
849 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
851 2018-02-27 Richard Biener <rguenther@suse.de>
853 PR tree-optimization/84512
854 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
855 Do not use the estimate returned from record_stmt_cost for
856 the scalar iteration cost but sum properly using add_stmt_cost.
858 2018-02-27 Richard Biener <rguenther@suse.de>
860 PR tree-optimization/84466
861 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
862 Adjust last change to less strictly validate use operands.
864 2018-02-27 Martin Liska <mliska@suse.cz>
866 PR gcov-profile/84548
867 * gcov.c (process_file): Allow partial overlap and consider it
868 also as group functions.
869 (output_lines): Properly calculate range of lines for a group.
871 2018-02-27 Martin Liska <mliska@suse.cz>
873 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
874 'ggc' suffixes. Change first column width.
875 (timer::print): Fix formatting of the column.
877 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
879 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
880 preserve inline entry blocks for the sake of debug inline
881 entry point markers alone.
882 (remove_unused_locals): Suggest in comments a better place to
883 force the preservation of inline entry blocks that are
884 otherwise unused, but do not preserve them.
886 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
888 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
890 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
893 * config/i386/constraints.md (Bs): Replace
894 ix86_indirect_branch_register with
895 TARGET_INDIRECT_BRANCH_REGISTER.
897 * config/i386/i386.md (indirect_jump): Likewise.
898 (tablejump): Likewise.
899 (*sibcall_memory): Likewise.
900 (*sibcall_value_memory): Likewise.
901 Peepholes of indirect call and jump via memory: Likewise.
902 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
903 (*sibcall_value_GOT_32): Likewise.
904 * config/i386/predicates.md (indirect_branch_operand): Likewise.
905 (GOT_memory_operand): Likewise.
906 (call_insn_operand): Likewise.
907 (sibcall_insn_operand): Likewise.
908 (GOT32_symbol_operand): Likewise.
909 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
911 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
913 PR rtl-optimization/83496
914 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
915 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
916 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
917 redundant insn, if any.
918 (relax_delay_slots): Likewise.
919 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
921 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
923 PR tree-optimization/83965
924 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
925 that grouped statements are part of a reduction chain. Return
926 true if the statement is not marked as a reduction itself but
928 (vect_recog_dot_prod_pattern): Don't check whether the statement
929 is part of a group here.
930 (vect_recog_sad_pattern): Likewise.
931 (vect_recog_widen_sum_pattern): Likewise.
933 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
936 * final.c (rest_of_clean_state): Also look for calls inside sequences.
938 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
941 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
943 (ix86_output_indirect_function_return): New prototype.
944 (ix86_split_simple_return_pop_internal): Likewise.
945 * config/i386/i386.c (indirect_return_via_cx): New.
946 (indirect_return_via_cx_bnd): Likewise.
947 (indirect_thunk_name): Handle return va CX_REG.
948 (output_indirect_thunk_function): Create alias for
949 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
950 (ix86_output_indirect_jmp): Remove the bool argument.
951 (ix86_output_indirect_function_return): New function.
952 (ix86_split_simple_return_pop_internal): Likewise.
953 * config/i386/i386.md (*indirect_jump): Don't pass false
954 to ix86_output_indirect_jmp.
955 (*tablejump_1): Likewise.
956 (simple_return_pop_internal): Change it to define_insn_and_split.
957 Call ix86_split_simple_return_pop_internal to split it for
959 (simple_return_indirect_internal): Call
960 ix86_output_indirect_function_return instead of
961 ix86_output_indirect_jmp.
963 2018-02-26 Jakub Jelinek <jakub@redhat.com>
966 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
967 memset and value initialization afterwards.
969 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
971 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
973 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
976 * common/config/aarch64/aarch64-common.c
977 (aarch_option_optimization_table[]): Switch
978 off fomit-frame-pointer
980 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
981 Chung-Ju Wu <jasonwucj@gmail.com>
983 * config/nds32/nds32-multiple.md (load_multiple): Disallow
985 (store_multiple): Ditto.
987 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
989 * config.gcc: Add --with-cpu support for nds32 target.
990 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
991 * config/nds32/nds32.opt: Add -mcpu= option.
993 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
995 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
996 isel=yes): Warn for these deprecated options.
998 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1000 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1001 ISA_2_5_MASKS_EMBEDDED.
1003 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1005 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1006 p->max as pointers rather than using iterative_hash_expr.
1008 2018-02-23 Carl Love <cel@us.ibm.com>
1010 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1011 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1013 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1014 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1015 P8V_BUILTIN_VEC_VUNSIGNED2.
1017 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1020 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1021 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1023 * lra-constraints.c (curr_insn_transform): Set up
1024 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1026 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1027 * lra-eliminations.c (spill_pseudos): Ditto.
1028 (process_insn_for_elimination): Ditto.
1029 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1030 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1031 LRA_NON_CLOBBERED_ALT.
1033 2018-02-22 Martin Sebor <msebor@redhat.com>
1035 PR tree-optimization/84480
1036 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1037 to maybe_diag_stxncpy_trunc. Call it.
1038 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1039 from gimple_fold_builtin_strcpy. Print inlining stack.
1040 (handle_builtin_stxncpy): Print inlining stack.
1041 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1043 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1046 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1047 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1048 and -fcheck-pointer-bounds are used together.
1049 (indirect_thunk_prefix): New enum.
1050 (indirect_thunk_need_prefix): New function.
1051 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1052 "_nt" instead of "_bnd" for NOTRACK prefix.
1053 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1054 (output_indirect_thunk_function): Likewise.
1056 (ix86_code_end): Update output_indirect_thunk_function calls.
1057 (ix86_output_indirect_branch_via_reg): Replace
1058 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1059 (ix86_output_indirect_branch_via_push): Likewise.
1060 (ix86_output_function_return): Likewise.
1061 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1062 incompatible with -fcf-protection=branch and
1063 -fcheck-pointer-bounds.
1065 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1068 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1069 Change gcc_assert call to output_operand_lossage.
1071 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1073 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1075 2018-02-22 DJ Delorie <dj@redhat.com>
1076 Sebastian Perta <sebastian.perta@renesas.com>
1077 Oleg Endo <olegendo@gcc.gnu.org>
1079 * config/rx/rx.c (rx_rtx_costs): New function.
1080 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1082 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1084 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1086 2018-02-22 Martin Liska <mliska@suse.cz>
1089 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1090 Add "native" as a possible value.
1092 2018-02-22 Martin Liska <mliska@suse.cz>
1095 * config/i386/i386.c (ix86_option_override_internal):
1096 Add "native" as a possible value for -march and -mtune.
1098 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1101 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1102 to all type variants.
1104 PR tree-optimization/84503
1105 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1106 width as info->bitpos + info->bitsize - start.
1107 (merged_store_group::merge_overlapping): Simplify width computation.
1108 (check_no_overlap): New function.
1109 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1110 start + width and last_order of the group, fail if check_no_overlap
1112 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1113 to group if check_no_overlap fails.
1115 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1117 * config/rs6000/altivec.md: Delete contraint arguments to
1118 define_expand, define_split, and define_peephole2, and in
1119 define_insn_and_split if always unused.
1120 * config/rs6000/darwin.md: Ditto.
1121 * config/rs6000/dfp.md: Ditto.
1122 * config/rs6000/rs6000.md: Ditto.
1123 * config/rs6000/sync.md: Ditto.
1124 * config/rs6000/vector.md: Ditto.
1125 * config/rs6000/vsx.md: Ditto.
1127 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1129 * config/rs6000/altivec.md: Write output control strings as braced
1130 blocks instead of double-quoted strings.
1131 * config/rs6000/darwin.md: Ditto.
1132 * config/rs6000/rs6000.md: Ditto.
1133 * config/rs6000/vector.md: Ditto.
1134 * config/rs6000/vsx.md: Ditto.
1136 2018-02-21 Jason Merrill <jason@redhat.com>
1138 PR c++/84314 - ICE with templates and fastcall attribute.
1139 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1141 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1143 * ipa-cp.c (determine_versionability): Fix comment typos.
1145 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1148 * ipa-cp.c (determine_versionability): Do not version functions caling
1151 2018-02-21 Martin Liska <mliska@suse.cz>
1154 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1155 Add "native" as a possible value.
1156 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1157 the macro when native cpu detection is available.
1159 2018-02-21 Martin Liska <mliska@suse.cz>
1162 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1163 Add "native" as a possible value.
1164 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1165 when native cpu detection is available.
1167 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1168 Martin Sebor <msebor@redhat.com>
1170 PR tree-optimization/84478
1171 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1173 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1174 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1175 support which is conservatively correct, for 2 only stay conservative
1176 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1177 argument to the 2 argument get_range_strlen, adjust 6 arg
1178 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1180 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1181 (gimple_fold_builtin_strlen): Pass true as last argument to
1184 2018-02-20 Martin Sebor <msebor@redhat.com>
1187 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1188 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1189 (builtin_memref::builtin_memref): Factor out parts into
1190 set_base_and_offset and call it.
1192 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1195 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1196 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1197 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1198 search at the associated MODE_INT.
1200 2018-02-20 Jeff Law <law@redhat.com>
1203 PR tree-optimization/81592
1205 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1206 for range data rather than using global data.
1207 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1208 range data rather than using global data.
1209 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1210 pass it to children as needed.
1211 (struct directive::fmtresult): Similarly.
1212 (struct directive::set_width): Similarly.
1213 (struct directive::set_precision): Similarly.
1214 (format_integer, format_directive, parse_directive): Similarly.
1215 (format_none): Accept unnamed vr_values parameter.
1216 (format_percent, format_floating, format_character): Similarly.
1217 (format_string, format_plain): Similarly.
1218 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1219 the EVRP range analyzer for range data rather than using global data.
1220 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1221 gimple-ssa-evrp-analyze.h
1222 (class sprintf_dom_walker): Add after_dom_children member function.
1223 Add evrp_range_analyzer member.
1224 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1225 range analyzer as needed.
1226 (sprintf_dom_walker::after_dom_children): New member function.
1227 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1229 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1230 (evrp_range_analyzer::pop_to_marker): Likewise.
1232 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1234 PR tree-optimization/84419
1235 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1236 with the required type if its current type is compatible but
1239 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1242 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1243 after vectorization.
1245 2018-02-20 Martin Liska <mliska@suse.cz>
1248 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1249 possible values if we don't have a hint.
1251 2018-02-20 Martin Liska <mliska@suse.cz>
1255 * final.c (shorten_branches): Build align_tab array with one
1257 * opts.c (finish_options): Add alignment option limit check.
1258 (MAX_CODE_ALIGN): Likewise.
1259 (MAX_CODE_ALIGN_VALUE): Likewise.
1260 * doc/invoke.texi: Document maximum allowed option value for
1261 all -falign-* options.
1263 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1266 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1267 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1268 * var-tracking.c (emit_note_insn_var_location): Remove all references
1269 to NOTE_INSN_CALL_ARG_LOCATION.
1270 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1271 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1272 Use copy_rtx_if_shared.
1273 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1274 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1275 (dwarf2out_var_location): Remove handling of
1276 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1278 * final.c (final_scan_insn): Remove all references to
1279 NOTE_INSN_CALL_ARG_LOCATION.
1280 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1281 before dumping final insns.
1282 * except.c (emit_note_eh_region_end): Remove all references to
1283 NOTE_INSN_CALL_ARG_LOCATION.
1284 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1285 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1286 * config/arc/arc.c (hwloop_optimize): Likewise.
1287 * config/arm/arm.c (create_fix_barrier): Likewise.
1288 * config/s390/s390.c (s390_chunkify_start): Likewise.
1289 * config/sh/sh.c (find_barrier): Likewise.
1290 * config/i386/i386.c (rest_of_insert_endbranch,
1291 ix86_seh_fixup_eh_fallthru): Likewise.
1292 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1293 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1294 * config/frv/frv.c (frv_function_prologue): Likewise.
1295 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1297 (note_outside_basic_block_p): Remove all references to
1298 NOTE_INSN_CALL_ARG_LOCATION.
1299 * gengtype.c (adjust_field_rtx_def): Likewise.
1300 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1302 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1303 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1306 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1309 PR tree-optimization/84452
1310 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1311 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1314 2018-02-19 Martin Liska <mliska@suse.cz>
1317 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1319 2018-02-19 Martin Liska <mliska@suse.cz>
1320 Richard Sandiford <richard.sandiford@linaro.org>
1322 PR tree-optimization/82491
1323 * gimple-fold.c (get_base_constructor): Make earlier bail out
1326 2018-02-19 Carl Love <cel@us.ibm.com>
1328 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1329 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1331 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1332 P8V_BUILTIN_VEC_NEG.
1334 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1336 * config/rl78/rl78.md (movdf): New define expand.
1338 2018-02-19 Martin Liska <mliska@suse.cz>
1341 * doc/invoke.texi: Fix typo.
1342 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1344 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1346 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1347 handle rs6000_single_float and rs6000_double_float specially for
1350 2018-02-16 Jeff Law <law@redhat.com>
1352 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1355 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1358 * ipa-inline.c (inline_small_functions): Fix a typo.
1360 2018-02-16 Nathan Sidwell <nathan@acm.org>
1362 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1364 2018-02-16 Carl Love <cel@us.ibm.com>
1366 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1367 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1368 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1369 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1370 expansion to P8V_BUILTIN_VEC_FLOAT2.
1372 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1374 PR rtl-optimization/70023
1375 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1376 src_regno into account.
1378 2018-02-16 Carl Love <cel@us.ibm.com>
1380 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1381 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1382 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1383 * config/rs6000/rs6000.c: Remove case statements for
1384 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1385 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1386 and P9V_BUILTIN_VEC_VINSERT4B.
1387 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1388 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1389 * config/rs6000/vsx.md:
1390 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1393 2018-02-16 Carl Love <cel@us.ibm.com>
1395 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1397 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1399 * config/rs6000/rs6000-c.c: Add the definitions for
1400 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1401 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1402 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1403 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1404 definition for insert4b and define insn *insert3b_internal.
1405 * doc/extend.texi: Add documentation for vec_extract4b.
1407 2018-02-16 Nathan Sidwell <nathan@acm.org>
1409 * doc/extend.texi (Backwards Compatibility): Mention friend
1410 injection. Note for-scope is deprecated.
1411 * doc/invoke.texi (-ffriend-injection): Deprecate.
1413 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1415 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1416 that moved to I2, also allow destinations that are a paradoxical
1417 subreg (instead of a normal reg).
1419 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1422 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1425 2018-02-16 Richard Biener <rguenther@suse.de>
1427 PR tree-optimization/84037
1428 PR tree-optimization/84016
1430 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1431 Adjust vec_construct for the fact we need additional higher latency
1432 128bit inserts for AVX256 and AVX512 vector builds.
1433 (ix86_add_stmt_cost): Scale vector construction cost for
1436 2018-02-16 Richard Biener <rguenther@suse.de>
1438 PR tree-optimization/84417
1439 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1440 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1441 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1443 2018-02-16 Martin Liska <mliska@suse.cz>
1446 * internal-fn.def (ASAN_CHECK): Set proper flags.
1447 (ASAN_MARK): Likewise.
1449 2018-02-16 Julia Koval <julia.koval@intel.com>
1451 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1452 from PTA_CANNONLAKE.
1454 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1457 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1458 Use ++iter rather than iter++ for std::list iterators.
1459 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1460 defer deleting them until all nodes in the forest are processed. Do
1461 free even leaf nodes. Change to_process into auto_vec.
1464 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1465 * vec.h (vec_default_construct): Use memset instead of placement new
1466 if BROKEN_VALUE_INITIALIZATION is defined.
1467 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1468 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1471 PR rtl-optimization/83723
1472 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1473 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1474 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1476 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1478 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1480 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1482 PR rtl-optimization/81443
1483 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1484 from inner REGs to paradoxical SUBREGs.
1486 2018-02-16 Richard Biener <rguenther@suse.de>
1488 PR tree-optimization/84399
1489 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1490 For operands we can analyze at their definition make sure we can
1491 analyze them at each use as well.
1493 2018-02-16 Richard Biener <rguenther@suse.de>
1495 PR tree-optimization/84190
1496 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1497 volatile accesses if the decl isn't volatile.
1499 2018-02-15 Jason Merrill <jason@redhat.com>
1501 PR c++/84314 - ICE with templates and fastcall attribute.
1502 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1503 TYPE_CANONICAL on an existing type.
1505 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1507 PR tree-optimization/84383
1508 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1509 dstoff nor call operand_equal_p if dstbase is NULL.
1511 PR tree-optimization/84334
1512 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1513 also a CONSTANT_CLASS_P, punt.
1515 2018-02-14 Jim Wilson <jimw@sifive.com>
1517 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1518 first SMALL_OPERAND check. New local min_second_step. Move assert
1519 to where locals are set. Add TARGET_RVC support.
1520 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1522 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1524 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1526 2018-02-14 Martin Sebor <msebor@redhat.com>
1528 PR tree-optimization/83698
1529 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1530 arrays constrain the offset range to their bounds.
1531 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1532 (builtin_access::overlap): Avoid setting the size of overlap if it's
1534 (maybe_diag_overlap): Also consider arrays when deciding what values
1535 of offsets to include in diagnostics.
1537 2018-02-14 Martin Sebor <msebor@redhat.com>
1540 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1541 that correspond to the kind of a declaration.
1543 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1546 * config/pa/pa.md: Load address of PIC label using the linkage table
1547 if the label is nonlocal.
1549 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1551 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1552 warning message if user requests -maltivec=be.
1553 * doc/invoke.texi: Document deprecation of -maltivec=be.
1555 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1558 * config/rs6000/rs6000-c.c: Update definitions for
1559 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1560 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1562 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1565 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1566 add _get_ssp intrinsics. Remove argument from
1567 __builtin_ia32_rdssp[d|q].
1568 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1569 * config/i386/i386-builtin.def: Remove argument from
1570 __builtin_ia32_rdssp[d|q].
1571 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1572 ix86_expand_special_args_builtin for _rdssp[d|q].
1573 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1574 Clear register before usage.
1575 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1576 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1578 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1580 PR tree-optimization/84357
1581 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1582 operand 1 of an ARRAY_REF too.
1584 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1587 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1588 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1590 (set_of_reg): New struct.
1591 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1592 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1593 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1595 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1596 Split into bitclr, bitset, bitinvert patterns if appropriate.
1597 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1598 use rx_fuse_in_memory_bitop.
1599 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1600 to named insn, correct maximum insn length.
1602 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1605 * machmode.def: Define a complex mode for PARTIAL_INT.
1606 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1608 * doc/rtl.texi: Document CSPImode.
1609 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1611 (msp430_hard_regno_nregs_with_padding): Likewise.
1613 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1616 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1618 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1620 PR rtl-optimization/84169
1621 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1622 we generated a parallel as new i3 and we split that to new i2 and i3
1623 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1624 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1625 those to i2, not i1. Partially rewrite this scan code.
1627 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1630 * stor-layout.c (place_field): For variable length fields, adjust
1631 offset_align afterwards not just based on the field's alignment,
1632 but also on the size.
1635 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1636 of exps and logs in the use_exp2 case.
1638 2018-02-13 Jeff Law <law@redhat.com>
1640 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1643 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1646 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1650 * common.opt (gas-loc-support, gas-locview-support): New.
1651 (ginline-points, ginternal-reset-location-views): New.
1652 * doc/invoke.texi: Document them. Use @itemx where intended.
1653 (gvariable-location-views): Adjust.
1654 * target.def (reset_location_view): New.
1655 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1656 (TARGET_RESET_LOCATION_VIEW): New.
1657 * doc/tm.texi: Rebuilt.
1658 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1659 (dwarf2out_default_as_locview_support): New.
1660 (output_asm_line_debug_info): Use option variables.
1661 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1662 (output_loc_list): Likewise.
1663 (add_high_low_attributes): Check option variables.
1664 Don't output entry view attribute in strict mode.
1665 (gen_inlined_subroutine_die): Check option variables.
1666 (dwarf2out_inline_entry): Likewise.
1667 (init_sections_and_labels): Likewise.
1668 (dwarf2out_early_finish): Likewise.
1669 (maybe_reset_location_view): New, from...
1670 (dwarf2out_var_location): ... here. Call it.
1671 * debug.h (dwarf2out_default_as_loc_support): Declare.
1672 (dwarf2out_default_as_locview_support): Declare.
1673 * hooks.c (hook_int_rtx_insn_0): New.
1674 * hooks.h (hook_int_rtx_insn_0): Declare.
1675 * toplev.c (process_options): Take -gas-loc-support and
1676 -gas-locview-support from dwarf2out. Enable
1677 -gvariable-location-views by default only with locview
1678 assembler support. Enable -ginternal-reset-location-views by
1679 default only if the target defines the corresponding hook.
1680 Enable -ginline-points by default if location views are
1681 enabled; force it disabled if statement frontiers are
1683 * tree-inline.c (expand_call_inline): Check option variables.
1684 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1686 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1688 PR tree-optimization/84321
1689 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1690 handling. Also check whether the anti-range contains any values
1691 that satisfy the mask; switch to a VR_RANGE if not.
1693 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1696 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1698 2018-02-13 Martin Jambor <mjambor@suse.cz>
1701 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1702 of call statements, also set location of a load to a temporary.
1704 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1706 * config/rl78/rl78.c (add_vector_labels): New function.
1707 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1708 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1709 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1710 which checks that no arguments are passed.
1711 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1712 * doc/extend.texi: Documentation for the new attribute.
1714 2018-02-13 Andreas Schwab <schwab@suse.de>
1716 * config/riscv/linux.h (CPP_SPEC): Define.
1718 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1721 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1722 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1723 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1724 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1725 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1726 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1727 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1728 temporarily for AES and PCLMUL builtins.
1730 PR tree-optimization/84339
1731 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1732 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1736 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1737 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1738 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1740 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1742 * omp-simd-clone.h: New file.
1743 * omp-simd-clone.c: Include omp-simd-clone.h.
1744 (expand_simd_clones): No longer static.
1745 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1746 cgraph.h and omp-simd-clone.h.
1747 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1748 (vect_recog_widen_shift_pattern): Formatting fix.
1749 (vect_pattern_recog_1): Don't check optab for calls.
1752 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1753 operands[2] into a REG before using gen_lowpart on it.
1755 2018-02-12 Jeff Law <law@redhat.com>
1758 * config/sh/sh.c (find_barrier): Consider a sibling call
1761 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1762 successfully back substituting a reg.
1764 2018-02-12 Richard Biener <rguenther@suse.de>
1766 PR tree-optimization/84037
1767 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1768 parameter, move visited init to caller.
1769 (vect_slp_analyze_operations): Separate cost from validity
1770 check, initialize visited once for all instances.
1771 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1773 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1775 (vect_model_promotion_demotion_cost): Likewise.
1776 (vectorizable_bswap): Guard cost modeling with !slp_node
1777 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1779 (vectorizable_call): Likewise.
1780 (vectorizable_conversion): Likewise.
1781 (vectorizable_assignment): Likewise.
1782 (vectorizable_shift): Likewise.
1783 (vectorizable_operation): Likewise.
1784 (vectorizable_store): Likewise.
1785 (vectorizable_load): Likewise.
1786 (vectorizable_condition): Likewise.
1787 (vectorizable_comparison): Likewise.
1789 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1792 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1793 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1794 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1796 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1799 * params.def (inline-min-speedup): Increase from 8 to 15.
1800 (max-inline-insns-auto): Decrease from 40 to 30.
1801 * ipa-split.c (consider_split): Add some buffer for function to
1802 be considered inlining candidate.
1803 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1806 2018-02-12 Richard Biener <rguenther@suse.de>
1808 PR tree-optimization/84037
1809 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1810 matched stmts if we cannot swap the non-matched ones.
1812 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1814 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1815 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1816 _mm_maskz_scalef_round_ss): New intrinsics.
1817 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1818 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1819 __builtin_ia32_scalefss_round): Remove.
1820 (__builtin_ia32_scalefsd_mask_round,
1821 __builtin_ia32_scalefss_mask_round): New intrinsics.
1822 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1823 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1824 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1825 "<round_constraint>")): Changed to ...
1826 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1827 "<round_scalar_constraint>")): ... this.
1828 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1829 %0, %1, %2<round_op3>}"): Changed to ...
1830 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1831 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1832 %2<round_scalar_mask_op3>}"): ... this.
1833 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1835 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1837 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1838 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1839 (_mm_maskz_sqrt_round_ss): New intrinsics.
1840 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1841 (__builtin_ia32_sqrtsd_mask_round)
1842 (__builtin_ia32_sqrtss_mask_round): New builtins.
1843 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1844 (__builtin_ia32_sqrtss_round): Remove.
1845 (__builtin_ia32_sqrtsd_mask_round)
1846 (__builtin_ia32_sqrtss_mask_round): New builtins.
1847 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1848 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1849 ((match_operand:VF_128 1 "vector_operand"
1850 "xBm,<round_constraint>")): Changed to ...
1851 ((match_operand:VF_128 1 "vector_operand"
1852 "xBm,<round_scalar_constraint>")): ... this.
1853 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1854 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1855 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1856 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1857 %<iptr>1<round_scalar_mask_op3>}): ... this.
1858 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1859 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1861 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1864 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1865 Cast vec_cmpeq result to correct type.
1866 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1867 Cast vec_cmpgt result to correct type.
1869 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1871 * final.c (final_scan_insn_1): Renamed from...
1872 (final_scan_insn): ... this. New wrapper, to recover
1873 seen from the outermost call in recursive ones.
1874 * config/sparc/sparc.c (output_return): Drop seen from call.
1875 (output_sibcall): Likewise.
1876 * config/visium/visium.c (output_branch): Likewise.
1878 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1880 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1883 2018-02-10 Alan Modra <amodra@gmail.com>
1886 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1887 Specify LR as an input.
1889 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1892 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1893 remove_member_access_dummy_vars): New functions.
1894 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1895 lower_omp_1, execute_lower_omp): Use them.
1897 PR rtl-optimization/84308
1898 * shrink-wrap.c (spread_components): Release todo vector.
1900 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1902 PR rtl-optimization/57193
1903 * ira-color.c (struct allocno_color_data): Add member
1904 conflict_allocno_hard_prefs.
1905 (update_conflict_allocno_hard_prefs): New.
1906 (bucket_allocno_compare_func): Add a preference based on
1907 conflict_allocno_hard_prefs.
1908 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1909 (color_allocnos): Remove a dead code. Initiate
1910 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1912 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1915 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1916 constraint from =wa to wa. Avoid a subreg on the output operand,
1917 instead use a pseudo and subreg it in a move.
1918 (p9_xxbrd_<mode>): Changed to ...
1919 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1920 (p9_xxbrd_v2df): New expander.
1921 (p9_xxbrw_<mode>): Changed to ...
1922 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1923 (p9_xxbrw_v4sf): New expander.
1925 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1927 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1929 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1932 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1933 multiply in 32-bit mode.
1934 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1935 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1938 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1940 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1941 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1942 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1943 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1945 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1948 * dwarf2out.c (is_trivial_indirect_ref): New function.
1949 (dwarf2out_late_global_decl): Do not generate a location
1950 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1951 and that are not defined in the current unit.
1953 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1955 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1956 instead of a libcall for UNORDERED.
1958 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1961 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1962 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1964 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1967 * config/s390/s390.c (s390_set_current_function): Invoke
1968 s390_indirect_branch_settings also if fndecl didn't change.
1970 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1972 * config/rs6000/rs6000.md (blockage): Set length to zero.
1974 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1976 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1978 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1981 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1982 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1986 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1987 PARALLEL incoming that failed vt_get_decl_and_offset check.
1990 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1991 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1993 (get_variable_section): For decls in named .bss* sections pass true as
1994 second argument to bss_initializer_p.
1996 2018-02-09 Marek Polacek <polacek@redhat.com>
1997 Jakub Jelinek <jakub@redhat.com>
2000 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2001 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2002 Sync some changes from cxx_fold_indirect_ref.
2004 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2006 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2008 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2009 (BLOCK_INLINE_ENTRY_LABEL): New.
2010 (dwarf2out_var_location): Disregard inline entry markers.
2011 (inline_entry_data): New struct.
2012 (inline_entry_data_hasher): New hashtable type.
2013 (inline_entry_data_hasher::hash): New.
2014 (inline_entry_data_hasher::equal): New.
2015 (inline_entry_data_table): New variable.
2016 (add_high_low_attributes): Add DW_AT_entry_pc and
2017 DW_AT_GNU_entry_view attributes if a pending entry is found
2018 in inline_entry_data_table. Add old entry_pc attribute only
2019 if debug nonbinding markers are disabled.
2020 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2021 markers are enabled.
2022 (block_within_block_p, dwarf2out_inline_entry): New.
2023 (dwarf2out_finish): Check that no entries remained in
2024 inline_entry_data_table.
2025 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2026 (final_scan_insn, notice_source_line): Likewise.
2027 (rest_of_clean_state): Skip inline entry markers.
2028 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2030 * gimple.c (gimple_build_debug_inline_entry): New.
2031 * gimple.h (enum gimple_debug_subcode): Add
2032 GIMPLE_DEBUG_INLINE_ENTRY.
2033 (gimple_build_debug_inline_entry): Declare.
2034 (gimple_debug_inline_entry_p): New.
2035 (gimple_debug_nonbind_marker_p): Adjust.
2036 * insn-notes.def (INLINE_ENTRY): New.
2037 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2038 inline entry marker notes.
2039 (print_insn): Likewise.
2040 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2041 (INSN_DEBUG_MARKER_KIND): Likewise.
2042 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2043 * tree-inline.c (expand_call_inline): Build and insert
2044 debug_inline_entry stmt.
2045 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2046 inline entry blocks early, if nonbind markers are enabled.
2047 (dump_scope_block): Dump fragment info.
2048 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2049 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2050 (gimple_build_debug_inline_entry): New.
2051 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2052 Enable/disable inline entry points too.
2053 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2054 (DEBUG_INSN): Describe inline entry markers.
2056 * common.opt (gvariable-location-views): New.
2057 (gvariable-location-views=incompat5): New.
2058 * config.in: Rebuilt.
2059 * configure: Rebuilt.
2060 * configure.ac: Test assembler for view support.
2061 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2062 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2063 * dwarf2out.c (var_loc_view): New typedef.
2064 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2065 (dwarf2out_locviews_in_attribute): New.
2066 (dwarf2out_locviews_in_loclist): New.
2067 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2068 (enum dw_line_info_opcode): Add LI_adv_address.
2069 (struct dw_line_info_table): Add view.
2070 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2071 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2072 (zero_view_p): New variable.
2073 (ZERO_VIEW_P): New macro.
2074 (output_asm_line_debug_info): New.
2075 (struct var_loc_node): Add view.
2076 (add_AT_view_list, AT_loc_list): New.
2077 (add_var_loc_to_decl): Add view param. Test it against last.
2078 (new_loc_list): Add view params. Record them.
2079 (AT_loc_list_ptr): Handle loc and view lists.
2080 (view_list_to_loc_list_val_node): New.
2081 (print_dw_val): Handle dw_val_class_view_list.
2082 (size_of_die): Likewise.
2083 (value_format): Likewise.
2084 (loc_list_has_views): New.
2085 (gen_llsym): Set vl_symbol too.
2086 (maybe_gen_llsym, skip_loc_list_entry): New.
2087 (dwarf2out_maybe_output_loclist_view_pair): New.
2088 (output_loc_list): Output view list or entries too.
2089 (output_view_list_offset): New.
2090 (output_die): Handle dw_val_class_view_list.
2091 (output_dwarf_version): New.
2092 (output_compilation_unit_header): Use it.
2093 (output_skeleton_debug_sections): Likewise.
2094 (output_rnglists, output_line_info): Likewise.
2095 (output_pubnames, output_aranges): Update version comments.
2096 (output_one_line_info_table): Output view numbers in asm comments.
2097 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2098 Call maybe_gen_llsym.
2099 (loc_list_from_tree_1): Adjust.
2100 (add_AT_location_description): Create view list attribute if
2101 needed, check it's absent otherwise.
2102 (convert_cfa_to_fb_loc_list): Adjust.
2103 (maybe_emit_file): Call output_asm_line_debug_info for test.
2104 (dwarf2out_var_location): Reset views as needed. Precompute
2105 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2106 attribute. Set view.
2107 (new_line_info_table): Reset next view.
2108 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2109 (dwarf2out_source_line): Likewise. Output view resets and labels to
2110 the assembler, or select appropriate line info opcodes.
2111 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2112 (optimize_string_length): Catch it. Adjust.
2113 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2114 dw_val_class_view_list, and remove it if no longer needed.
2115 (hash_loc_list): Hash view numbers.
2116 (loc_list_hasher::equal): Compare them.
2117 (optimize_location_lists): Check whether a view list symbol is
2118 needed, and whether the locview attribute is present, and
2119 whether they match. Remove the locview attribute if no longer
2121 (index_location_lists): Call skip_loc_list_entry for test.
2122 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2123 Use output_dwarf_version.
2124 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2125 (struct dw_val_node): Add val_view_list.
2126 * final.c (SEEN_NEXT_VIEW): New.
2127 (set_next_view_needed): New.
2128 (clear_next_view_needed): New.
2129 (maybe_output_next_view): New.
2130 (final_start_function): Rename to...
2131 (final_start_function_1): ... this. Take pointer to FIRST,
2132 add SEEN parameter. Emit param bindings in the initial view.
2133 (final_start_function): Reintroduce SEEN-less interface.
2134 (final): Rename to...
2135 (final_1): ... this. Take SEEN parameter. Output final pending
2136 next view at the end.
2137 (final): Reintroduce seen-less interface.
2138 (final_scan_insn): Output pending next view before switching
2139 sections or ending a block. Mark the next view as needed when
2140 outputting variable locations. Notify debug backend of section
2141 changes, and of location view changes.
2142 (rest_of_handle_final): Adjust.
2143 * toplev.c (process_options): Autodetect value for debug variable
2144 location views option. Warn on incompat5 without -gdwarf-5.
2145 * doc/invoke.texi (gvariable-location-views): New.
2146 (gvariable-location-views=incompat5): New.
2147 (gno-variable-location-views): New.
2149 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2151 PR tree-optimization/84136
2152 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2153 that the result of find_edge is non-NULL.
2155 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2158 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2159 storing integer register in SImode. Fix cost of 256 and 512
2160 byte aligned SSE register store.
2162 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2164 * config/i386/i386.c (ix86_multiplication_cost): Fix
2165 multiplication cost for TARGET_AVX512DQ.
2167 2018-02-08 Marek Polacek <polacek@redhat.com>
2169 PR tree-optimization/84238
2170 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2173 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2175 PR tree-optimization/84265
2176 * tree-vect-stmts.c (vectorizable_store): Don't treat
2177 VMAT_CONTIGUOUS accesses as grouped.
2178 (vectorizable_load): Likewise.
2180 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2182 PR tree-optimization/81635
2183 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2184 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2185 (test_round_for_mask): New functions.
2186 (wide_int_cc_tests): Call test_round_for_mask.
2187 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2188 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2189 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2190 range returned by get_range_info.
2192 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2195 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2196 * symtab.c: Include builtins.h
2197 (symtab_node::output_to_lto_symbol_table_p): Move here
2198 from lto-streamer-out.c:output_symbol_p.
2199 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2200 (output_symbol_p): Move all logic to symtab.c
2201 (produce_symtab): Update.
2203 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2205 * config/s390/s390-opts.h (enum indirect_branch): Define.
2206 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2207 (s390_indirect_branch_via_thunk)
2208 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2209 (enum s390_indirect_branch_type): Define.
2210 * config/s390/s390.c (struct s390_frame_layout, struct
2211 machine_function): Remove.
2212 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2213 (indirect_branch_table_label_no, indirect_branch_table_name):
2215 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2216 (enum s390_indirect_branch_option): Define.
2217 (s390_return_addr_from_memory): New function.
2218 (s390_handle_string_attribute): New function.
2219 (s390_attribute_table): Add new attribute handler.
2220 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2221 (s390_indirect_branch_via_thunk): New function.
2222 (s390_indirect_branch_via_inline_thunk): New function.
2223 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2224 sibling call optimization for non z10 compiles.
2225 (s390_emit_call): Force indirect branch target to be a single
2226 register. Add r1 clobber for non-z10 compiles.
2227 (s390_emit_epilogue): Emit return jump via return_use expander.
2228 (s390_reorg): Handle JUMP_INSNs as execute targets.
2229 (s390_option_override_internal): Perform validity checks for the
2230 new command line options.
2231 (s390_indirect_branch_attrvalue): New function.
2232 (s390_indirect_branch_settings): New function.
2233 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2234 (s390_output_indirect_thunk_function): New function.
2235 (s390_code_end): Implement target hook.
2236 (s390_case_values_threshold): Implement target hook.
2237 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2239 * config/s390/s390.h (struct s390_frame_layout)
2240 (struct machine_function): Move here from s390.c.
2241 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2242 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2243 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2244 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2245 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2246 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2247 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2248 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2249 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2250 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2251 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2252 (mnemonic attribute): Add values which aren't recognized
2254 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2255 pattern for branch conversion. Fix mnemonic attribute.
2256 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2257 indirect branch via thunk if requested.
2258 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2259 ("*indirect_jump"): Disable for branch conversion using out of
2261 ("indirect_jump_via_thunk<mode>_z10")
2262 ("indirect_jump_via_thunk<mode>")
2263 ("indirect_jump_via_inlinethunk<mode>_z10")
2264 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2265 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2266 ("casesi_jump_via_inlinethunk<mode>_z10")
2267 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2268 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2269 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2270 ("*indirect2_jump"): Disable for branch conversion.
2271 ("casesi_jump"): Turn into expander and expand patterns for branch
2273 ("return_use"): New expander.
2274 ("*return"): Emit return via thunk and rename it to ...
2275 ("*return<mode>"): ... this one.
2276 * config/s390/s390.opt: Add new options and and enum for the
2279 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2281 * lra-constraints.c (match_reload): Unconditionally use
2282 gen_lowpart_SUBREG, rather than selecting between that
2283 and equivalent gen_rtx_SUBREG code.
2285 2018-02-08 Richard Biener <rguenther@suse.de>
2287 PR tree-optimization/84233
2288 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2289 changed flag instead of boguously re-using phi_inserted.
2291 2018-02-08 Martin Jambor <mjambor@suse.cz>
2293 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2294 static local variables.
2296 2018-02-08 Richard Biener <rguenther@suse.de>
2298 PR tree-optimization/84278
2299 * tree-vect-stmts.c (vectorizable_store): When looking for
2300 smaller vector types to perform grouped strided loads/stores
2301 make sure the mode is supported by the target.
2302 (vectorizable_load): Likewise.
2304 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2306 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2307 Increase LDP/STP opportunities by adding adjacent callee-saves.
2309 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2311 PR rtl-optimization/84068
2312 PR rtl-optimization/83459
2313 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2315 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2317 PR tree-optimization/84224
2318 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2319 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2322 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2325 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2326 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2327 count, remove one USE.
2329 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2331 * doc/install.texi (Configuration): Document the
2332 --with-long-double-format={ibm,ieee} PowerPC configuration
2336 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2337 Convert from define_expand to be define_insn_and_split. Rework
2338 float/double/_Float128 conversions to QI/HI/SImode to work with
2339 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2340 conversions to QI/HImode types did a store and then a load to
2341 truncate the value. For conversions to VSX registers, don't split
2342 the insn, instead emit the code directly. Use the code iterator
2343 any_fix to combine signed and unsigned conversions.
2344 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2345 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2346 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2347 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2348 (fix_<mode>di2_hw): Likewise.
2349 (fixuns_<mode>di2_hw): Likewise.
2350 (fix_<mode>si2_hw): Likewise.
2351 (fixuns_<mode>si2_hw): Likewise.
2352 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2353 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2354 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2355 fix<uns>_trunc<SFDF:mode>si2_p8.
2356 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2358 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2359 (fix<uns>_<mode>_mem): Likewise.
2360 (fctiw<u>z_<mode>_mem): Likewise.
2361 (fix<uns>_<mode>_mem): Likewise.
2362 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2363 the register allocator from doing a direct move to the GPRs to do
2364 a store, and instead use the ISA 3.0 store byte/half-word from
2365 vector register instruction. For IEEE 128-bit floating point,
2366 also optimize stores of 32-bit ints.
2367 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2369 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2371 * genextract.c (push_pathstr_operand): New function to support
2373 (walk_rtx): Call push_pathstr_operand.
2374 (print_path): Support [a-zA-Z].
2376 2018-02-07 Richard Biener <rguenther@suse.de>
2378 PR tree-optimization/84037
2379 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2380 (cse_and_gimplify_to_preheader): Declare.
2381 (vect_get_place_in_interleaving_chain): Likewise.
2382 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2384 (_loop_vec_info::~_loop_vec_info): Delete it.
2385 (cse_and_gimplify_to_preheader): New function.
2386 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2387 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2388 (vectorizable_load): Likewise. For grouped stores always base
2389 the IV on the first element.
2390 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2391 condition before gimplifying.
2393 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2395 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2396 *DIV_EXPR and *MOD_EXPR.
2398 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2401 * config/i386/i386.c (ix86_option_override_internal): Mask out
2402 the CF_SET bit when checking -fcf-protection.
2404 2018-02-07 Tom de Vries <tom@codesourcery.com>
2407 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2410 2018-02-07 Richard Biener <rguenther@suse.de>
2412 PR tree-optimization/84204
2413 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2416 PR tree-optimization/84205
2417 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2418 special-case isl_ast_op_zdiv_r.
2420 PR tree-optimization/84223
2421 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2422 Only add conditions from within the region.
2423 (gather_bbs::after_dom_children): Adjust.
2425 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2428 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2429 * config/avr/avr.md: Only post-reload split REG-REG moves if
2430 either register is GENERAL_REG_P.
2432 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2434 PR tree-optimization/84235
2435 * tree-ssa-scopedtables.c
2436 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2437 if the subtraction is performed in floating point type where NaNs are
2438 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2439 build 1. Formatting fix.
2441 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2444 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2445 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2446 and skip it regardless of bb boundaries. Use CALL_P macro,
2447 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2448 unnecessarily, formatting fix.
2450 2018-02-06 Michael Collison <michael.collison@arm.com>
2452 * config/arm/thumb2.md:
2453 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2454 (*thumb_mov_notscc): Ditto.
2456 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2459 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2460 unsigned_fix, not "s".
2462 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2464 * configure.ac (gcc_fn_eh_frame_ro): New function.
2465 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2466 correct .eh_frame permissions.
2467 * configure: Regenerate.
2469 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2471 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2474 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2476 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2477 Display warning message for -mno-speculate-indirect-jumps.
2479 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2481 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2483 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2485 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2487 PR tree-optimization/84225
2488 * tree-eh.c (find_trapping_overflow): Only call
2489 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2491 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2494 * config/i386/i386.c: Reimplement the check of possible options
2495 -mibt/-mshstk conbination. Change error messages.
2496 * doc/invoke.texi: Fix a typo: remove extra '='.
2498 2018-02-06 Marek Polacek <polacek@redhat.com>
2500 PR tree-optimization/84228
2501 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2503 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2506 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2507 emitted arch directives.
2508 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2509 __ARM_FEATURE_COPROC before changing architectures.
2511 2018-02-06 Richard Biener <rguenther@suse.de>
2513 * config/i386/i386.c (print_reg): Fix typo.
2514 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2516 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2518 * configure: Regenerate.
2520 2018-02-05 Martin Sebor <msebor@redhat.com>
2522 PR tree-optimization/83369
2523 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2526 2018-02-05 Martin Liska <mliska@suse.cz>
2528 * doc/invoke.texi: Cherry-pick upstream r323995.
2530 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2532 * ira.c (ira_init_register_move_cost): Adjust comment.
2534 2018-02-05 Martin Liska <mliska@suse.cz>
2536 PR gcov-profile/84137
2537 * doc/gcov.texi: Fix typo in documentation.
2539 2018-02-05 Martin Liska <mliska@suse.cz>
2541 PR gcov-profile/83879
2542 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2543 using dlopen functionality.
2545 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2547 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2548 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2549 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2550 _mm_maskz_range_round_ss): New intrinsics.
2551 (__builtin_ia32_rangesd128_round)
2552 (__builtin_ia32_rangess128_round): Remove.
2553 (__builtin_ia32_rangesd128_mask_round,
2554 __builtin_ia32_rangess128_mask_round): New builtins.
2555 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2556 __builtin_ia32_rangess128_round): Remove.
2557 (__builtin_ia32_rangesd128_mask_round,
2558 __builtin_ia32_rangess128_mask_round): New builtins.
2559 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2560 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2561 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2562 "<round_saeonly_constraint>")): Changed to ...
2563 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2564 "<round_saeonly_scalar_constraint>")): ... this.
2565 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2566 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2567 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2568 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2569 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2571 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2573 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2575 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2576 Remove all values except native, 8540 and 8548.
2578 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2580 * config/i386/i386.c (ix86_output_function_return): Pass
2581 INVALID_REGNUM, instead of -1, as invalid register number to
2582 indirect_thunk_name and output_indirect_thunk.
2584 2018-02-02 Julia Koval <julia.koval@intel.com>
2586 * config.gcc: Add -march=icelake.
2587 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2588 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2589 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2590 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2591 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2592 (processor_target_table): Add icelake.
2593 (ix86_option_override_internal): Handle new PTAs.
2594 (get_builtin_code_for_version): Handle icelake.
2595 (M_INTEL_COREI7_ICELAKE): New.
2596 (fold_builtin_cpu): Handle icelake.
2597 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2598 * doc/invoke.texi: Add -march=icelake.
2600 2018-02-02 Julia Koval <julia.koval@intel.com>
2602 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2603 to wide_int_bitmask.
2604 * wide-int-bitmask.h: New.
2606 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2609 * config/i386/i386.md: Replace Pmode with word_mode in
2610 builtin_setjmp_setup and builtin_longjmp to support x32.
2612 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2616 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2618 (rs6000_supported_cpu_names): New static variable.
2619 (linux_cpu_translation_table): Likewise.
2620 (elf_platform) <cpu>: Define new static variable and use it.
2621 Translate kernel AT_PLATFORM name to canonical name if needed.
2622 Error if platform name is unknown.
2624 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2627 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2629 2018-02-01 Jeff Law <law@redhat.com>
2632 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2633 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2634 the scratch if RELEASE_VIA_POP is false.
2635 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2636 If we have to save a temporary register, decrement SIZE appropriately.
2637 Pass new arguments to release_scratch_register_on_entry.
2638 (ix86_adjust_stack_and_probe): Likewise.
2639 (ix86_emit_probe_stack_range): Pass new arguments to
2640 release_scratch_register_on_entry.
2642 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2644 PR rtl-optimization/84157
2645 * combine.c (change_zero_ext): Use REG_P predicate in
2646 front of HARD_REGISTER_P predicate.
2648 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2650 * config/avr/avr.c (avr_option_override): Move disabling of
2651 -fdelete-null-pointer-checks to...
2652 * common/config/avr/avr-common.c (avr_option_optimization_table):
2655 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2657 PR tree-optimization/81635
2658 * tree-data-ref.c (split_constant_offset_1): For types that
2659 wrap on overflow, try to use range info to prove that wrapping
2662 2018-02-01 Renlin Li <renlin.li@arm.com>
2665 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2667 (aarch64_register_move_cost): Likewise.
2668 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2670 (REG_CLASS_NAMES): Likewise.
2671 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2672 TAILCALL_ADDR_REGS. Remove IP registers.
2673 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2675 2018-02-01 Richard Biener <rguenther@suse.de>
2677 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2678 for specifying RPO order and allow NULL for that.
2679 * domwalk.c (dom_walker::dom_walker): Likewise.
2680 (dom_walker::walk): Handle NULL RPO order.
2681 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2683 (rewrite_update_dom_walker): Likewise.
2684 (mark_def_dom_walker): Likewise.
2686 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2688 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2689 (aarch64_maybe_expand_sve_subreg_move): Declare.
2690 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2691 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2693 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2694 that are semantically a reverse operation.
2695 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2696 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2697 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2699 (aarch64_can_change_mode_class): For big-endian, forbid changes
2700 between two SVE modes if they have different element sizes.
2702 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2704 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2705 the TImode handling for big-endian targets.
2707 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2709 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2710 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2712 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2713 Remove BSWAP handing for big-endian targets and use the form of
2714 LD1RQ appropariate for the mode.
2716 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2718 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2719 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2722 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2725 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2726 check for operands that need to go through aarch64_sve_reload_be.
2728 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2730 PR tree-optimization/81661
2731 PR tree-optimization/84117
2732 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2733 * tree-eh.c: Include gimplify.h.
2734 (find_trapping_overflow, replace_trapping_overflow,
2735 rewrite_to_non_trapping_overflow): New functions.
2736 * tree-vect-loop.c: Include tree-eh.h.
2737 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2738 * tree-data-ref.c: Include tree-eh.h.
2739 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2741 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2743 PR rtl-optimization/84123
2744 * combine.c (change_zero_ext): Check if hard register satisfies
2745 can_change_dest_mode before calling gen_lowpart_SUBREG.
2747 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2750 * ira.c (ira_init_register_move_cost): Remove assert.
2752 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2754 PR rtl-optimization/84071
2755 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2756 * doc/tm.texi: Regenerate.
2758 2018-01-31 Richard Biener <rguenther@suse.de>
2760 PR tree-optimization/84132
2761 * tree-data-ref.c (analyze_miv_subscript): Properly
2762 check whether evolution_function_is_affine_multivariate_p
2763 before calling gcd_of_steps_may_divide_p.
2765 2018-01-31 Julia Koval <julia.koval@intel.com>
2768 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2769 * config/i386/i386.md (rdpid_rex64) New.
2770 (rdpid): Make 32bit only.
2772 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2775 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2776 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2778 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2781 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2783 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2785 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2787 PR rtl-optimization/84071
2788 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2789 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2791 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2793 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2794 (arc_attribute_table): Add 'aux' attribute.
2795 (arc_in_small_data_p): Consider aux like variables.
2796 (arc_is_aux_reg_p): New function.
2797 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2798 (arc_get_aux_arg): New function.
2799 (prepare_move_operands): Handle aux-register access.
2800 (arc_handle_aux_attribute): New function.
2801 * doc/extend.texi (ARC Variable attributes): Add subsection.
2803 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2805 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2806 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2807 (arc_attribute_table): Add 'uncached' attribute.
2808 (arc_print_operand): Print '.di' flag for uncached memory
2810 (arc_in_small_data_p): Do not consider for small data the uncached
2812 (arc_is_uncached_mem_p): New function.
2813 * config/arc/predicates.md (compact_store_memory_operand): Check
2814 for uncached memory accesses.
2815 (nonvol_nonimm_operand): Likewise.
2816 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2818 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2821 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2822 falign-loops=): Add Optimization flag.
2824 2018-01-30 Jeff Law <law@redhat.com>
2827 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2828 INT_REGISTERS_SAVED. Check it prior to calling
2829 get_scratch_register_on_entry.
2830 (ix86_adjust_stack_and_probe): Similarly.
2831 (ix86_emit_probe_stack_range): Similarly.
2832 (ix86_expand_prologue): Corresponding changes.
2834 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2837 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2838 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2840 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2843 * lra-constraints.c (curr_insn_transform): Process AND in the
2846 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2848 PR rtl-optimization/83986
2849 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2850 dependence against last_pending_memory_flush in addition to
2853 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2855 PR tree-optimization/81611
2856 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2859 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2862 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2865 2018-01-30 Richard Biener <rguenther@suse.de>
2866 Jakub Jelinek <jakub@redhat.com>
2868 PR tree-optimization/84111
2869 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2870 inner loops added during recursion, as they don't have up-to-date
2873 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2876 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2877 (can_inline_edge_by_limits_p): ... here.
2878 (can_early_inline_edge_p, check_callers,
2879 update_caller_keys, update_callee_keys, recursive_inlining,
2880 add_new_edges_to_heap, speculation_useful_p,
2881 inline_small_functions,
2882 inline_small_functions, flatten_function,
2883 inline_to_all_callers_1): Update.
2885 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2887 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2890 2018-01-30 Richard Biener <rguenther@suse.de>
2892 PR tree-optimization/83008
2893 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2894 invariant and constant vector uses in stmts when they need
2897 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2900 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2901 * configure: Regenerate.
2903 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2905 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2907 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2908 Use gen_rtx_REG rather than gen_lowpart.
2910 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2912 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2913 rather than 0 when creating partial subregs.
2915 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2917 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2920 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2923 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2924 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2925 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2926 flags. This restores the settings used before the 2017-07-24.
2927 Turning off pre increment/decrement/modify allows IVOPTS to
2928 optimize DF/SF loops where the index is an int.
2930 2018-01-29 Richard Biener <rguenther@suse.de>
2931 Kelvin Nilsen <kelvin@gcc.gnu.org>
2934 * tree-vect-stmts.c (vectorizable_call): Don't call
2935 targetm.vectorize_builtin_md_vectorized_function if callee is
2938 2018-01-22 Carl Love <cel@us.ibm.com>
2940 * doc/extend.tex: Fix typo in second arg in
2941 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2943 2018-01-29 Richard Biener <rguenther@suse.de>
2945 PR tree-optimization/84086
2946 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2947 (flush_ssaname_freelist): When SSA names were released reset
2948 the SCEV hash table.
2950 2018-01-29 Richard Biener <rguenther@suse.de>
2952 PR tree-optimization/84057
2953 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2954 removed paths when removing edges.
2956 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2958 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2959 -mfunction-return=@var{choice}.
2961 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2964 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2966 (layout::print_source_line): Likewise.
2967 (test_get_line_width_without_trailing_whitespace): Add test cases.
2969 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2972 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2975 2018-01-26 Jim Wilson <jimw@sifive.com>
2977 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2979 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2982 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2984 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2985 and CMP + SUB-immediate -> SUBS.
2987 2018-01-26 Martin Sebor <msebor@redhat.com>
2989 PR tree-optimization/83896
2990 * tree-ssa-strlen.c (get_string_len): Rename...
2991 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2992 Avoid assuming length is constant.
2993 (handle_char_store): Use HOST_WIDE_INT for string length.
2995 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2998 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2999 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3001 2018-01-26 Richard Biener <rguenther@suse.de>
3003 PR rtl-optimization/84003
3004 * dse.c (record_store): Only record redundant stores when
3005 the earlier store aliases at least all accesses the later one does.
3007 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3009 PR rtl-optimization/83985
3010 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3011 REG_CFA_RESTORE insns.
3012 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3013 REG_CFA_RESTORE insns here.
3016 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3017 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3019 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3021 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3022 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3023 (arc_init): Likewise.
3024 (arc_override_options): Likewise.
3025 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3027 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3029 * config/arc/arc.h (TARGET_DBNZ): Define.
3030 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3031 properly set the tune attribute.
3032 (dbnz): Use TARGET_DBNZ guard.
3033 * config/arc/arc.opt (mtune): Add core3 option.
3035 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3037 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3038 recognize new pic like addresses.
3039 (arc_delegitimize_address): Clean up.
3041 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3043 * config/arc/arc-arches.def: Option mrf16 valid for all
3045 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3046 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3047 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3048 * config/arc/arc-tables.opt: Regenerate.
3049 * config/arc/arc.c (arc_conditional_register_usage): Handle
3050 reduced register file case.
3051 (arc_file_start): Set must have build attributes.
3052 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3054 * config/arc/arc.opt (mrf16): Add new option.
3055 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3056 * config/arc/genmultilib.awk: Handle new mrf16 option.
3057 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3058 * config/arc/t-multilib: Regenerate.
3059 * doc/invoke.texi (ARC Options): Document mrf16 option.
3061 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3063 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3064 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3065 (arc_attribute_table): Add 'secure_call' attribute.
3066 (arc_print_operand): Print secure call operand.
3067 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3069 (arc_is_secure_call_p): New function. * config/arc/arc.md
3070 (call_i): Add support for sjli instruction.
3071 (call_value_i): Likewise.
3072 * config/arc/constraints.md (Csc): New constraint.
3074 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3075 John Eric Martin <John.Martin@emmicro-us.com>
3077 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3078 * config/arc/arc.c (_arc_jli_section): New struct.
3079 (arc_jli_section): New type.
3080 (rc_jli_sections): New static variable.
3081 (arc_handle_jli_attribute): New function.
3082 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3083 (arc_file_end): New function.
3084 (TARGET_ASM_FILE_END): Define.
3085 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3086 (arc_add_jli_section): New function.
3087 (jli_call_scan): Likewise.
3088 (arc_reorg): Call jli_call_scan.
3089 (arc_output_addsi): Remove 'S' from printing asm operand.
3090 (arc_is_jli_call_p): New function.
3091 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3093 (movhi_insn): Likewise.
3094 (movsi_insn): Likewise.
3095 (movsi_set_cc_insn): Likewise.
3096 (loadqi_update): Likewise.
3097 (load_zeroextendqisi_update): Likewise.
3098 (load_signextendqisi_update): Likewise.
3099 (loadhi_update): Likewise.
3100 (load_zeroextendhisi_update): Likewise.
3101 (load_signextendhisi_update): Likewise.
3102 (loadsi_update): Likewise.
3103 (loadsf_update): Likewise.
3104 (movsicc_insn): Likewise.
3105 (bset_insn): Likewise.
3106 (bxor_insn): Likewise.
3107 (bclr_insn): Likewise.
3108 (bmsk_insn): Likewise.
3109 (bicsi3_insn): Likewise.
3110 (cmpsi_cc_c_insn): Likewise.
3111 (movsi_ne): Likewise.
3112 (movsi_cond_exec): Likewise.
3113 (clrsbsi2): Likewise.
3124 (call_i): Remove 'S' asm letter, add jli instruction.
3125 (call_value_i): Likewise.
3126 * config/arc/arc.op (mjli-always): New option.
3127 * config/arc/constraints.md (Cji): New constraint.
3128 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3130 (subsf3_fpx): Likewise.
3131 (mulsf3_fpx): Likewise.
3132 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3134 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3136 * doc/invoke.texi (ARC): Document mjli-always option.
3138 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3140 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3141 avoid addition with 0 and use incw and decw where possible.
3143 2018-01-26 Richard Biener <rguenther@suse.de>
3145 PR tree-optimization/81082
3146 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3147 association if it requires casting to unsigned.
3148 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3149 from fold_plusminus_mult_expr to catch important cases late when
3150 range info is available.
3152 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3154 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3155 * configure.ac (hidden_linkonce): New test.
3156 * configure: Regenerate.
3157 * config.in: Regenerate.
3159 2018-01-26 Julia Koval <julia.koval@intel.com>
3161 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3162 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3163 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3164 _mm_mask_bitshuffle_epi64_mask): Fix type.
3165 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3166 USI_FTYPE_V4DI_V4DI_USI): Remove.
3167 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3168 __builtin_ia32_vpshufbitqmb256_mask,
3169 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3170 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3171 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3173 2018-01-26 Alan Modra <amodra@gmail.com>
3176 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3177 UNSPEC_VBPERMQ. Sort other unspecs.
3179 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3181 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3183 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3186 * predict.c (drop_profile): Do not push/pop cfun; update also
3188 (handle_missing_profiles): Fix logic looking for zero profiles.
3190 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3193 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3194 on functions with #pragma omp declare simd or functions with simd
3196 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3197 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3198 Remove trailing \n from warning_at calls.
3200 2018-01-25 Tom de Vries <tom@codesourcery.com>
3203 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3204 for neutered workers.
3206 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3209 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3210 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3212 2018-01-24 Jeff Law <law@redhat.com>
3215 * i386.c (get_probe_interval): Move to earlier point.
3216 (ix86_compute_frame_layout): If -fstack-clash-protection and
3217 the frame is larger than the probe interval, then use pushes
3218 to save registers rather than reg->mem moves.
3219 (ix86_expand_prologue): Remove conditional for int_registers_saved
3222 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3225 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3226 min/max for never referenced object.
3228 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3231 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3233 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3234 attributes from DECL_ATTRIBUTES (decl) without affecting
3235 DECL_ATTRIBUTES (current_function_decl).
3236 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3237 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3239 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3241 PR tree-optimization/83979
3242 * fold-const.c (fold_comparison): Use constant_boolean_node
3243 instead of boolean_{true,false}_node.
3245 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3247 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3250 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3252 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3253 Simplify the clause that sets the length attribute.
3254 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3255 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3256 clause that sets the length attribute.
3257 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3259 2018-01-24 Tom de Vries <tom@codesourcery.com>
3262 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3263 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3264 Add strict parameter.
3265 (prevent_branch_around_nothing): Insert dummy insn between branch to
3266 label and label with no ptx insn inbetween.
3267 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3269 2018-01-24 Tom de Vries <tom@codesourcery.com>
3272 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3273 for neutered threads in warp.
3274 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3276 2018-01-24 Richard Biener <rguenther@suse.de>
3278 PR tree-optimization/83176
3279 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3282 2018-01-24 Richard Biener <rguenther@suse.de>
3284 PR tree-optimization/82819
3285 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3286 code generating pluses that are no-ops in the target precision.
3288 2018-01-24 Richard Biener <rguenther@suse.de>
3291 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3293 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3295 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3296 to merge probabilities.
3297 * predict.c (probably_never_executed): Also mark as cold functions
3298 with global 0 profile and guessed local profile.
3299 * profile-count.c (profile_probability::combine_with_count): New
3301 * profile-count.h (profile_probability::operator*,
3302 profile_probability::operator*=, profile_probability::operator/,
3303 profile_probability::operator/=): Reduce precision to adjusted
3304 and set value to guessed on contradictory divisions.
3305 (profile_probability::combine_with_freq): Remove.
3306 (profile_probability::combine_wiht_count): Declare.
3307 (profile_count::force_nonzero):: Set to adjusted.
3308 (profile_count::probability_in):: Set quality to adjusted.
3309 * tree-ssa-tail-merge.c (replace_block_by): Use
3312 2018-01-23 Andrew Waterman <andrew@sifive.com>
3313 Jim Wilson <jimw@sifive.com>
3315 * config/riscv/riscv.c (riscv_stack_boundary): New.
3316 (riscv_option_override): Set riscv_stack_boundary. Handle
3317 riscv_preferred_stack_boundary_arg.
3318 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3319 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3320 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3321 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3322 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3323 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3325 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3328 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3329 of struct ix86_frame.
3330 (ix86_expand_epilogue): Likewise. Add a local variable for
3331 the reg_save_offset field in struct ix86_frame.
3333 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3335 PR tree-optimization/82604
3336 * tree-loop-distribution.c (enum partition_kind): New enum item
3337 PKIND_PARTIAL_MEMSET.
3338 (partition_builtin_p): Support above new enum item.
3339 (generate_code_for_partition): Ditto.
3340 (compute_access_range): Differentiate cases that equality can be
3341 proven at all loops, the innermost loops or no loops.
3342 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3343 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3344 (finalize_partitions, distribute_loop): Don't fuse partition of
3345 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3346 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3349 2018-01-23 Martin Liska <mliska@suse.cz>
3351 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3352 order to ignore the predictor.
3353 (PRED_POLYMORPHIC_CALL): Likewise.
3354 (PRED_RECURSIVE_CALL): Likewise.
3356 2018-01-23 Martin Liska <mliska@suse.cz>
3358 * tree-profile.c (tree_profiling): Print function header to
3359 aware reader which function we are working on.
3360 * value-prof.c (gimple_find_values_to_profile): Do not print
3361 not interesting value histograms.
3363 2018-01-23 Martin Liska <mliska@suse.cz>
3365 * profile-count.h (enum profile_quality): Add
3366 profile_uninitialized as the first value. Do not number values
3367 as they are zero based.
3368 (profile_count::verify): Update sanity check.
3369 (profile_probability::verify): Likewise.
3371 2018-01-23 Nathan Sidwell <nathan@acm.org>
3373 * doc/invoke.texi (ffor-scope): Deprecate.
3375 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3377 PR tree-optimization/83510
3378 * domwalk.c (set_all_edges_as_executable): New function.
3379 (dom_walker::dom_walker): Convert bool param
3380 "skip_unreachable_blocks" to enum reachability. Move setup of
3381 edge flags to set_all_edges_as_executable and only do it when
3382 reachability is REACHABLE_BLOCKS.
3383 * domwalk.h (enum dom_walker::reachability): New enum.
3384 (dom_walker::dom_walker): Convert bool param
3385 "skip_unreachable_blocks" to enum reachability.
3386 (set_all_edges_as_executable): New decl.
3387 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3388 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3390 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3391 but converting true to REACHABLE_BLOCKS.
3392 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3394 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3395 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3396 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3398 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3399 if check_all_array_refs will be called.
3401 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3403 * tree.c (selftest::test_location_wrappers): Add more test
3406 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3408 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3409 (selftest::test_bit_in_range): Likewise.
3411 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3414 * doc/sourcebuild.texi (vect_float): Say that the selector
3415 only describes the situation when -funsafe-math-optimizations is on.
3416 (vect_float_strict): Document.
3418 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3420 PR tree-optimization/83965
3421 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3422 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3423 instead of checking only for a reduction.
3424 (vect_recog_widen_sum_pattern): Likewise.
3426 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3428 * predict.c (probably_never_executed): Only use precise profile info.
3429 (compute_function_frequency): Skip after inlining hack since we now
3430 have quality checking.
3432 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3434 * profile-count.h (profile_probability::very_unlikely,
3435 profile_probability::unlikely, profile_probability::even): Set
3436 precision to guessed.
3438 2018-01-23 Richard Biener <rguenther@suse.de>
3440 PR tree-optimization/83963
3441 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3442 Properly terminate dominator walk when crossing the exit edge not
3443 when visiting its source block.
3445 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3448 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3449 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3451 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3453 PR tree-optimization/83957
3454 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3455 semicolon after for body surrounded by braces.
3457 PR tree-optimization/83081
3458 * profile-count.h (profile_probability::split): New method.
3459 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3460 Use profile_probability::split.
3461 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3462 when splitting a single conditional jump into 2.
3464 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3466 PR tree-optimization/69452
3467 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3470 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3472 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3473 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3474 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3476 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3478 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3479 * config/rl78/rl78.md (movdi): New define_expand.
3480 * config/rl78/rl78.c (rl78_split_movdi): New function.
3482 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3485 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3487 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3488 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3489 128-bit to produce an UNSPEC move to get the double word with the
3490 signbit and then a shift directly to do signbit.
3491 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3492 implementation with a new version that just does either a direct
3493 move or a regular move. Move memory interface to separate insns.
3494 Move insns so they are next to the expander.
3495 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3496 with signbit move. Split big and little endian case.
3497 (signbit<mode>2_dm_mem_le): Likewise.
3498 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3499 (signbit<mode>2_dm2): Likewise.
3501 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3503 * config/rl78/rl78.md (anddi3): New define_expand.
3505 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3507 * config/rl78/rl78.md (umindi3): New define_expand.
3509 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3511 * config/rl78/rl78.md (smindi3): New define_expand.
3513 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3515 * config/rl78/rl78.md (smaxdi3): New define_expand.
3517 2018-01-22 Carl Love <cel@us.ibm.com>
3519 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3520 LVX_V1TI): Add macro expansion.
3521 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3522 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3523 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3524 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3525 Change check to determine if the instruction is a byte reversing
3526 entry. Fix typo in comment.
3527 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3528 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3529 Add def_builtin calls for new builtins.
3530 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3531 Add define_insn expansion.
3533 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3535 * config/rl78/rl78.md (umaxdi3): New define_expand.
3537 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3539 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3540 for non-QImode registers.
3542 2018-01-22 Richard Biener <rguenther@suse.de>
3544 PR tree-optimization/83963
3545 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3546 including the loop exit block.
3547 (scop_detection::merge_sese): Likewise.
3548 (scop_detection::add_scop): Do it here instead.
3550 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3552 * doc/sourcebuild.texi (arm_softfloat): Document.
3554 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3557 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3558 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3559 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3561 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3562 David Edelsohn <dje.gcc@gmail.com>
3565 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3566 Change "crset eq" to "crset 2".
3567 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3568 (*call_indirect_aix<mode>_nospec): Likewise.
3569 (*call_value_indirect_aix<mode>_nospec): Likewise.
3570 (*call_indirect_elfv2<mode>_nospec): Likewise.
3571 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3572 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3573 change assembly output from . to $.
3574 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3575 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3576 (*tablejump<mode>_internal1_nospec): Likewise.
3578 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3581 * config/sh/sh_optimize_sett_clrt.cc:
3582 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3584 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3586 PR tree-optimization/83940
3587 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3588 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3589 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3590 use it to pass back the definition type.
3591 (vect_check_store_rhs): Likewise rhs_dt_out.
3592 (vect_build_gather_load_calls): Add a mask_dt argument and use
3593 it instead of a call to vect_is_simple_use.
3594 (vectorizable_store): Update calls to vect_check_load_store_mask
3595 and vect_check_store_rhs. Use the dt returned by the latter instead
3596 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3597 instead of calls to vect_is_simple_use. Pass the scalar rather
3598 than the vector operand to vect_is_simple_use when handling
3599 second and subsequent copies of an rhs value.
3600 (vectorizable_load): Update calls to vect_check_load_store_mask
3601 and vect_build_gather_load_calls. Use the cached mask_dt and
3602 gs_info.offset_dt instead of calls to vect_is_simple_use.
3604 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3607 * tree-emutls.c: Include gimplify.h.
3608 (lower_emutls_2): New function.
3609 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3610 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3611 it before further processing.
3614 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3615 UINTVAL (trueop1) instead of INTVAL (op1).
3617 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3621 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3622 INCOMING_FRAME_SP_OFFSET if not defined.
3623 (scan_trace): Add ENTRY argument. If true and
3624 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3625 emit a note to adjust the CFA offset.
3626 (create_cfi_notes): Adjust scan_trace callers.
3627 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3628 INCOMING_FRAME_SP_OFFSET in the CIE.
3629 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3630 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3632 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3633 * doc/tm.texi: Regenerated.
3635 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3637 PR rtl-optimization/83147
3638 * lra-constraints.c (remove_inheritance_pseudos): Use
3639 lra_substitute_pseudo_within_insn.
3641 2018-01-19 Tom de Vries <tom@codesourcery.com>
3642 Cesar Philippidis <cesar@codesourcery.com>
3645 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3647 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3650 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3651 spaces for function labels.
3653 2018-01-19 Martin Liska <mliska@suse.cz>
3655 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3656 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3657 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3658 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3659 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3660 (PRED_CONST_RETURN): Change from 69 to 65.
3661 (PRED_NULL_RETURN): Change from 91 to 71.
3662 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3663 (PRED_LOOP_GUARD): Change from 66 to 73.
3665 2018-01-19 Martin Liska <mliska@suse.cz>
3667 * predict.c (predict_insn_def): Add new assert.
3668 (struct branch_predictor): Change type to signed integer.
3669 (test_prediction_value_range): Amend test to cover
3671 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3672 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3673 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3674 (PRED_LOOP_IV_COMPARE): Likewise.
3675 * predict.h (PROB_UNINITIALIZED): Define new constant.
3677 2018-01-19 Martin Liska <mliska@suse.cz>
3679 * predict.c (dump_prediction): Add new format for
3680 analyze_brprob.py script which is enabled with -details
3682 * profile-count.h (precise_p): New function.
3684 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3686 PR tree-optimization/83922
3687 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3688 there are no statements that need masking.
3689 (vect_active_double_reduction_p): New function.
3690 (vect_analyze_loop_operations): Use it when handling phis that
3691 are not in the loop header.
3693 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3695 PR tree-optimization/83914
3696 * tree-vect-loop.c (vectorizable_induction): Don't convert
3697 init_expr or apply the peeling adjustment for inductions
3698 that are nested within the vectorized loop.
3700 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3702 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3705 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3709 * function.h (gimplify_parameters): Add gimple_seq * argument.
3710 * function.c: Include gimple.h and options.h.
3711 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3712 for the added local temporaries if needed.
3713 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3714 if there are any parameter cleanups, wrap whole body into a
3715 try/finally with the cleanups.
3717 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3720 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3721 Use GET_MODE_CLASS for scalar floating point.
3723 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3727 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3728 Fix call of call_cgraph_insertion_hooks.
3730 2018-01-18 Martin Sebor <msebor@redhat.com>
3732 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3734 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3737 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3740 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3743 * common.opt: (-ffile-prefix-map): New option.
3744 * opts.c (common_handle_option): Defer it.
3745 * opts-global.c (handle_common_deferred_options): Handle it.
3746 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3747 * file-prefix-map.h: New file.
3748 (remap_debug_filename, add_debug_prefix_map): ...here.
3749 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3750 * final.c (debug_prefix_map, add_debug_prefix_map
3751 remap_debug_filename): Move to...
3752 * file-prefix-map.c: New file.
3753 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3754 generalize, get rid of alloca(), use strrchr() instead of strchr().
3755 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3756 Implement in terms of add_prefix_map().
3757 (remap_macro_filename, remap_debug_filename): Implement in term of
3759 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3760 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3761 * dbxout.c: Include file-prefix-map.h.
3762 * varasm.c: Likewise.
3763 * vmsdbgout.c: Likewise.
3764 * xcoffout.c: Likewise.
3765 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3766 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3767 * doc/invoke.texi (-ffile-prefix-map): Document.
3768 (-fdebug-prefix-map): Update description.
3770 2018-01-18 Martin Liska <mliska@suse.cz>
3772 * config/i386/i386.c (indirect_thunk_name): Document that also
3774 (output_indirect_thunk): Document why both instructions
3775 (pause and lfence) are generated.
3777 2018-01-18 Richard Biener <rguenther@suse.de>
3779 PR tree-optimization/83887
3780 * graphite-scop-detection.c
3781 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3782 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3783 (scop_detection::merge_sese): Re-implement with a flood-fill
3784 algorithm that properly finds a SESE region if it exists.
3786 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3789 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3790 pointer_diff optimizations use view_convert instead of convert.
3792 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3794 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3795 Generate different code for -mno-speculate-indirect-jumps.
3796 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3797 (*call_indirect_aix<mode>): Disable for
3798 -mno-speculate-indirect-jumps.
3799 (*call_indirect_aix<mode>_nospec): New define_insn.
3800 (*call_value_indirect_aix<mode>): Disable for
3801 -mno-speculate-indirect-jumps.
3802 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3803 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3804 -mno-speculate-indirect-jumps.
3805 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3807 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3809 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3810 long double type, set the flags for noting the default long double
3811 type, even if we don't pass or return a long double type.
3813 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3816 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3819 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3821 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3822 support for merge[hl].
3823 (fold_mergehl_helper): New helper function.
3824 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3825 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3826 (altivec_vmrglw_direct): Add xxmrglw insn.
3828 2018-01-17 Andrew Waterman <andrew@sifive.com>
3830 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3831 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3833 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3836 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3837 call the lto_location_cache before reading the
3838 DECL_SOURCE_LOCATION of the types.
3840 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3841 Richard Sandiford <richard.sandiford@linaro.org>
3843 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3844 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3845 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3846 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3847 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3849 * config/aarch64/constraints.md (aarch64_movti_operand):
3851 * config/aarch64/predicates.md (Uti): Add new constraint.
3853 2018-01-17 Carl Love <cel@us.ibm.com>
3855 * config/rs6000/vsx.md (define_expand xl_len_r,
3856 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3857 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3859 (define_expand, define_insn): Move the shift left from the
3860 define_insn to the define_expand for lxvl and stxvl instructions.
3861 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3862 and XL_LEN_R definitions to PURE.
3864 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3866 * config/i386/i386.c (indirect_thunk_name): Declare regno
3867 as unsigned int. Compare regno with INVALID_REGNUM.
3868 (output_indirect_thunk): Ditto.
3869 (output_indirect_thunk_function): Ditto.
3870 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3871 in the call to output_indirect_thunk_function.
3873 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3876 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3877 rather than the size of inner_type to determine the stack slot size
3878 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3880 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3883 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3886 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3888 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3889 endian Linux systems to optionally enable multilibs for selecting
3890 the long double type if the user configured an explicit type.
3891 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3892 have no long double multilibs if not defined.
3893 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3894 warn if the user used -mabi={ieee,ibm}longdouble and we built
3895 multilibs for long double.
3896 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3897 appropriate multilib option.
3898 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3900 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3901 for building long double multilibs.
3902 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3904 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3906 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3909 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3911 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3914 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3917 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3920 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3922 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3923 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3925 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3927 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3928 different rtl trees depending on TARGET_64BIT.
3929 (rs6000_gen_lvx): Likewise.
3931 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3933 * config/visium/visium.md (nop): Tweak comment.
3934 (hazard_nop): Likewise.
3936 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3938 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3939 -mspeculate-indirect-jumps.
3940 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3941 for -mno-speculate-indirect-jumps.
3942 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3943 (*call_value_indirect_elfv2<mode>): Disable for
3944 -mno-speculate-indirect-jumps.
3945 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3946 (indirect_jump): Emit different RTL for
3947 -mno-speculate-indirect-jumps.
3948 (*indirect_jump<mode>): Disable for
3949 -mno-speculate-indirect-jumps.
3950 (*indirect_jump<mode>_nospec): New define_insn.
3951 (tablejump): Emit different RTL for
3952 -mno-speculate-indirect-jumps.
3953 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3954 (tablejumpsi_nospec): New define_expand.
3955 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3956 (tablejumpdi_nospec): New define_expand.
3957 (*tablejump<mode>_internal1): Disable for
3958 -mno-speculate-indirect-jumps.
3959 (*tablejump<mode>_internal1_nospec): New define_insn.
3960 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3963 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3965 * caller-save.c (insert_save): Drop unnecessary parameter. All
3968 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3969 Richard Biener <rguenth@suse.de>
3972 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3973 return early, inline manually is_gimple_sizepos. Make sure if we
3974 call gimplify_expr we don't end up with a gimple constant.
3975 * tree.c (variably_modified_type_p): Don't return true for
3976 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3977 * gimplify.h (is_gimple_sizepos): Remove.
3979 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3981 PR tree-optimization/83857
3982 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3983 vectorizable_live_operation for pure SLP statements.
3984 (vectorizable_live_operation): Handle PHIs.
3986 2018-01-16 Richard Biener <rguenther@suse.de>
3988 PR tree-optimization/83867
3989 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3990 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3992 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3995 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3996 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3997 If off is not INTEGER_CST, issue a may not be aligned warning
3998 rather than isn't aligned. Use isn%'t rather than isn't.
3999 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4001 <case MULT_EXPR>: Improve the case when bottom and one of the
4002 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4003 operand, in that case check if the other operand is multiple of
4004 bottom divided by the INTEGER_CST operand.
4006 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4009 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4010 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4011 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4012 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4013 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4014 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4015 (pa_function_arg_size): New function.
4017 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4019 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4020 in a separate statement.
4022 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4024 PR tree-optimization/83847
4025 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4026 group gathers and scatters.
4028 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4030 PR rtl-optimization/86620
4031 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4033 PR rtl-optimization/83213
4034 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4035 to last if both are JUMP_INSNs.
4037 PR tree-optimization/83843
4038 * gimple-ssa-store-merging.c
4039 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4040 store_immediate_info for bswap/nop orig_stores.
4042 2018-01-15 Andrew Waterman <andrew@sifive.com>
4044 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4046 <UDIV>: Increase cost if !TARGET_DIV.
4048 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4050 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4051 (define_attr "cr_logical_3op"): New.
4052 (cceq_ior_compare): Adjust.
4053 (cceq_ior_compare_complement): Adjust.
4054 (*cceq_rev_compare): Adjust.
4055 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4056 (is_cracked_insn): Adjust.
4057 (insn_must_be_first_in_group): Adjust.
4058 * config/rs6000/40x.md: Adjust.
4059 * config/rs6000/440.md: Adjust.
4060 * config/rs6000/476.md: Adjust.
4061 * config/rs6000/601.md: Adjust.
4062 * config/rs6000/603.md: Adjust.
4063 * config/rs6000/6xx.md: Adjust.
4064 * config/rs6000/7450.md: Adjust.
4065 * config/rs6000/7xx.md: Adjust.
4066 * config/rs6000/8540.md: Adjust.
4067 * config/rs6000/cell.md: Adjust.
4068 * config/rs6000/e300c2c3.md: Adjust.
4069 * config/rs6000/e500mc.md: Adjust.
4070 * config/rs6000/e500mc64.md: Adjust.
4071 * config/rs6000/e5500.md: Adjust.
4072 * config/rs6000/e6500.md: Adjust.
4073 * config/rs6000/mpc.md: Adjust.
4074 * config/rs6000/power4.md: Adjust.
4075 * config/rs6000/power5.md: Adjust.
4076 * config/rs6000/power6.md: Adjust.
4077 * config/rs6000/power7.md: Adjust.
4078 * config/rs6000/power8.md: Adjust.
4079 * config/rs6000/power9.md: Adjust.
4080 * config/rs6000/rs64.md: Adjust.
4081 * config/rs6000/titan.md: Adjust.
4083 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4085 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4086 ix86_indirect_branch_register logic.
4088 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4090 * config/i386/constraints.md (Bs): Update
4091 ix86_indirect_branch_register check. Don't check
4092 ix86_indirect_branch_register with GOT_memory_operand.
4094 * config/i386/predicates.md (GOT_memory_operand): Don't check
4095 ix86_indirect_branch_register here.
4096 (GOT32_symbol_operand): Likewise.
4098 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4100 * config/i386/predicates.md (constant_call_address_operand):
4101 Rewrite ix86_indirect_branch_register logic.
4102 (sibcall_insn_operand): Likewise.
4104 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4106 * config/i386/constraints.md (Bs): Replace
4107 ix86_indirect_branch_thunk_register with
4108 ix86_indirect_branch_register.
4110 * config/i386/i386.md (indirect_jump): Likewise.
4111 (tablejump): Likewise.
4112 (*sibcall_memory): Likewise.
4113 (*sibcall_value_memory): Likewise.
4114 Peepholes of indirect call and jump via memory: Likewise.
4115 * config/i386/i386.opt: Likewise.
4116 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4117 (GOT_memory_operand): Likewise.
4118 (call_insn_operand): Likewise.
4119 (sibcall_insn_operand): Likewise.
4120 (GOT32_symbol_operand): Likewise.
4122 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4125 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4126 type rather than type addr's type points to.
4127 (expand_omp_atomic_mutex): Likewise.
4128 (expand_omp_atomic): Likewise.
4130 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4133 * config/i386/i386.c (output_indirect_thunk_function): Use
4134 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4135 for __x86_return_thunk.
4137 2018-01-15 Richard Biener <rguenther@suse.de>
4140 * expmed.c (extract_bit_field_1): Fix typo.
4142 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4145 * config/arm/iterators.md (VF): New mode iterator.
4146 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4147 Remove integer-related logic from pattern.
4148 (neon_vabd<mode>_3): Likewise.
4150 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4153 * common.opt (fstrict-overflow): No longer an alias.
4154 (fwrapv-pointer): New option.
4155 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4156 also for pointer types based on flag_wrapv_pointer.
4157 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4158 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4159 opts->x_flag_wrapv got set.
4160 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4161 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4162 POINTER_TYPE_OVERFLOW_UNDEFINED.
4163 * match.pd: Likewise in address comparison pattern.
4164 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4166 2018-01-15 Richard Biener <rguenther@suse.de>
4169 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4170 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4171 Reset type names to their identifier if their TYPE_DECL doesn't
4172 have linkage (and thus is used for ODR and devirt).
4173 (save_debug_info_for_decl): Remove.
4174 (save_debug_info_for_type): Likewise.
4175 (add_tree_to_fld_list): Adjust.
4176 * tree-pretty-print.c (dump_generic_node): Make dumping of
4177 type names more robust.
4179 2018-01-15 Richard Biener <rguenther@suse.de>
4181 * BASE-VER: Bump to 8.0.1.
4183 2018-01-14 Martin Sebor <msebor@redhat.com>
4186 * builtins.c (check_access): Avoid warning when the no-warning bit
4189 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4191 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4192 * ira-color (allocno_hard_regs_compare): Likewise.
4194 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4197 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4198 Use .pushsection/.popsection.
4200 2018-01-14 Martin Sebor <msebor@redhat.com>
4203 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4205 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4207 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4208 entry from extra_headers.
4209 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4210 extra_headers, make the list bitwise identical to the i?86-*-* one.
4212 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4214 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4215 -mcmodel=large with -mindirect-branch=thunk,
4216 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4217 -mfunction-return=thunk-extern.
4218 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4219 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4220 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4222 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4224 * config/i386/i386.c (print_reg): Print the name of the full
4225 integer register without '%'.
4226 (ix86_print_operand): Handle 'V'.
4227 * doc/extend.texi: Document 'V' modifier.
4229 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4231 * config/i386/constraints.md (Bs): Disallow memory operand for
4232 -mindirect-branch-register.
4234 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4235 (GOT_memory_operand): Likewise.
4236 (call_insn_operand): Likewise.
4237 (sibcall_insn_operand): Likewise.
4238 (GOT32_symbol_operand): Likewise.
4239 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4240 for -mindirect-branch-register.
4241 (tablejump): Likewise.
4242 (*sibcall_memory): Likewise.
4243 (*sibcall_value_memory): Likewise.
4244 Disallow peepholes of indirect call and jump via memory for
4245 -mindirect-branch-register.
4246 (*call_pop): Replace m with Bw.
4247 (*call_value_pop): Likewise.
4248 (*sibcall_pop_memory): Replace m with Bs.
4249 * config/i386/i386.opt (mindirect-branch-register): New option.
4250 * doc/invoke.texi: Document -mindirect-branch-register option.
4252 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4254 * config/i386/i386-protos.h (ix86_output_function_return): New.
4255 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4256 set function_return_type.
4257 (indirect_thunk_name): Add ret_p to indicate thunk for function
4259 (output_indirect_thunk_function): Pass false to
4260 indirect_thunk_name.
4261 (ix86_output_indirect_branch_via_reg): Likewise.
4262 (ix86_output_indirect_branch_via_push): Likewise.
4263 (output_indirect_thunk_function): Create alias for function
4264 return thunk if regno < 0.
4265 (ix86_output_function_return): New function.
4266 (ix86_handle_fndecl_attribute): Handle function_return.
4267 (ix86_attribute_table): Add function_return.
4268 * config/i386/i386.h (machine_function): Add
4269 function_return_type.
4270 * config/i386/i386.md (simple_return_internal): Use
4271 ix86_output_function_return.
4272 (simple_return_internal_long): Likewise.
4273 * config/i386/i386.opt (mfunction-return=): New option.
4274 (indirect_branch): Mention -mfunction-return=.
4275 * doc/extend.texi: Document function_return function attribute.
4276 * doc/invoke.texi: Document -mfunction-return= option.
4278 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4280 * config/i386/i386-opts.h (indirect_branch): New.
4281 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4282 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4283 with local indirect jump when converting indirect call and jump.
4284 (ix86_set_indirect_branch_type): New.
4285 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4286 (indirectlabelno): New.
4287 (indirect_thunk_needed): Likewise.
4288 (indirect_thunk_bnd_needed): Likewise.
4289 (indirect_thunks_used): Likewise.
4290 (indirect_thunks_bnd_used): Likewise.
4291 (INDIRECT_LABEL): Likewise.
4292 (indirect_thunk_name): Likewise.
4293 (output_indirect_thunk): Likewise.
4294 (output_indirect_thunk_function): Likewise.
4295 (ix86_output_indirect_branch_via_reg): Likewise.
4296 (ix86_output_indirect_branch_via_push): Likewise.
4297 (ix86_output_indirect_branch): Likewise.
4298 (ix86_output_indirect_jmp): Likewise.
4299 (ix86_code_end): Call output_indirect_thunk_function if needed.
4300 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4302 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4303 (ix86_attribute_table): Add indirect_branch.
4304 * config/i386/i386.h (machine_function): Add indirect_branch_type
4305 and has_local_indirect_jump.
4306 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4308 (tablejump): Likewise.
4309 (*indirect_jump): Use ix86_output_indirect_jmp.
4310 (*tablejump_1): Likewise.
4311 (simple_return_indirect_internal): Likewise.
4312 * config/i386/i386.opt (mindirect-branch=): New option.
4313 (indirect_branch): New.
4316 (thunk-inline): Likewise.
4317 (thunk-extern): Likewise.
4318 * doc/extend.texi: Document indirect_branch function attribute.
4319 * doc/invoke.texi: Document -mindirect-branch= option.
4321 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4324 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4326 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4328 * ipa-inline.c (want_inline_small_function_p): Return false if
4329 inlining has already failed with CIF_FINAL_ERROR.
4330 (update_caller_keys): Call want_inline_small_function_p before
4332 (update_callee_keys): Likewise.
4334 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4336 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4338 (rs6000_quadword_masked_address_p): Likewise.
4339 (quad_aligned_load_p): Likewise.
4340 (quad_aligned_store_p): Likewise.
4341 (const_load_sequence_p): Add comment to describe the outer-most loop.
4342 (mimic_memory_attributes_and_flags): New function.
4343 (rs6000_gen_stvx): Likewise.
4344 (replace_swapped_aligned_store): Likewise.
4345 (rs6000_gen_lvx): Likewise.
4346 (replace_swapped_aligned_load): Likewise.
4347 (replace_swapped_load_constant): Capitalize argument name in
4348 comment describing this function.
4349 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4350 and stores that access quad-word aligned addresses and replace
4351 with stvx or lvx instructions when appropriate.
4352 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4353 New function prototype.
4354 (rs6000_quadword_masked_address_p): Likewise.
4355 (rs6000_gen_lvx): Likewise.
4356 (rs6000_gen_stvx): Likewise.
4357 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4358 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4359 when memory address is aligned.
4360 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4361 this split to select lvx instruction when memory address is aligned.
4362 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4363 instruction when memory address is aligned.
4364 (*vsx_le_perm_load_v16qi): Likewise.
4365 (four unnamed splitters): Modify to select the stvx instruction
4366 when memory is aligned.
4368 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4370 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4371 which appears in the queue multiple times.
4373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4374 Alan Hayward <alan.hayward@arm.com>
4375 David Sherwood <david.sherwood@arm.com>
4377 * tree-vectorizer.h (vec_lower_bound): New structure.
4378 (_loop_vec_info): Add check_nonzero and lower_bounds.
4379 (LOOP_VINFO_CHECK_NONZERO): New macro.
4380 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4381 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4382 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4383 fields. Make seg_len the distance travelled, not including the
4385 (dr_direction_indicator): Declare.
4386 (dr_zero_step_indicator): Likewise.
4387 (dr_known_forward_stride_p): Likewise.
4388 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4390 (runtime_alias_check_p): Allow runtime alias checks with
4392 (operator ==): Compare access_size and align.
4393 (prune_runtime_alias_test_list): Rework for new distinction between
4394 the access_size and seg_len.
4395 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4397 (get_segment_min_max): New function.
4398 (create_intersect_range_checks): Use it.
4399 (dr_step_indicator): New function.
4400 (dr_direction_indicator): Likewise.
4401 (dr_zero_step_indicator): Likewise.
4402 (dr_known_forward_stride_p): Likewise.
4403 * tree-loop-distribution.c (data_ref_segment_size): Return
4404 DR_STEP * (niters - 1).
4405 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4407 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4408 (vect_preserves_scalar_order_p): New function, split out from...
4409 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4410 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4411 (vect_vfa_access_size): New function.
4412 (vect_vfa_align): Likewise.
4413 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4414 (dump_lower_bound): New function.
4415 (vect_check_lower_bound): Likewise.
4416 (vect_small_gap_p): Likewise.
4417 (vectorizable_with_step_bound_p): Likewise.
4418 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4419 depencies if the vectorization factor is 1. Convert the checks
4420 for nonzero steps into checks on the bounds of DR_STEP. Try using
4421 a bunds check for variable steps if the minimum required step is
4422 relatively small. Update calls to the dr_with_seg_len
4423 constructor and to vect_compile_time_alias.
4424 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4426 (vect_loop_versioning): Call it.
4427 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4429 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4432 Alan Hayward <alan.hayward@arm.com>
4433 David Sherwood <david.sherwood@arm.com>
4435 * doc/sourcebuild.texi (vect_scatter_store): Document.
4436 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4438 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4440 * genopinit.c (main): Add supports_vec_scatter_store and
4441 supports_vec_scatter_store_cached to target_optabs.
4442 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4443 IFN_MASK_SCATTER_STORE.
4444 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4446 * internal-fn.h (internal_store_fn_p): Declare.
4447 (internal_fn_stored_value_index): Likewise.
4448 * internal-fn.c (scatter_store_direct): New macro.
4449 (expand_scatter_store_optab_fn): New function.
4450 (direct_scatter_store_optab_supported_p): New macro.
4451 (internal_store_fn_p): New function.
4452 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4453 IFN_MASK_SCATTER_STORE.
4454 (internal_fn_mask_index): Likewise.
4455 (internal_fn_stored_value_index): New function.
4456 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4458 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4459 * optabs-query.c (supports_vec_scatter_store_p): New function.
4460 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4461 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4462 true for scatter stores.
4463 (vect_gather_scatter_fn_p): Handle scatter stores too.
4464 (vect_check_gather_scatter): Consider using scatter stores if
4465 supports_vec_scatter_store_p.
4466 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4468 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4469 internal_fn_stored_value_index.
4470 (check_load_store_masking): Handle scatter stores too.
4471 (vect_get_store_rhs): Make public.
4472 (vectorizable_call): Use internal_store_fn_p.
4473 (vectorizable_store): Handle scatter store internal functions.
4474 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4475 when deciding whether the end of the group has been reached.
4476 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4477 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4478 (mask_scatter_store<mode>): New insns.
4480 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4481 Alan Hayward <alan.hayward@arm.com>
4482 David Sherwood <david.sherwood@arm.com>
4484 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4485 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4486 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4488 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4489 Use vect_truncate_gather_scatter_offset if we can't treat the
4490 operation as a normal gather load or scatter store.
4491 (get_group_load_store_type): Take the gather_scatter_info
4492 as argument. Try using a gather load or scatter store for
4493 single-element groups.
4494 (get_load_store_type): Update calls to get_group_load_store_type
4495 and vect_use_strided_gather_scatters_p.
4497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4498 Alan Hayward <alan.hayward@arm.com>
4499 David Sherwood <david.sherwood@arm.com>
4501 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4502 optional tree argument.
4503 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4505 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4506 but continue to use the current value as a fallback.
4507 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4508 to compare the updates.
4509 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4510 (get_load_store_type): Use it when handling a strided access.
4511 (vect_get_strided_load_store_ops): New function.
4512 (vect_get_data_ptr_increment): Likewise.
4513 (vectorizable_load): Handle strided gather loads. Always pass
4514 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4516 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4517 Alan Hayward <alan.hayward@arm.com>
4518 David Sherwood <david.sherwood@arm.com>
4520 * doc/md.texi (gather_load@var{m}): Document.
4521 (mask_gather_load@var{m}): Likewise.
4522 * genopinit.c (main): Add supports_vec_gather_load and
4523 supports_vec_gather_load_cached to target_optabs.
4524 * optabs-tree.c (init_tree_optimization_optabs): Use
4525 ggc_cleared_alloc to allocate target_optabs.
4526 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4527 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4529 * internal-fn.h (internal_load_fn_p): Declare.
4530 (internal_gather_scatter_fn_p): Likewise.
4531 (internal_fn_mask_index): Likewise.
4532 (internal_gather_scatter_fn_supported_p): Likewise.
4533 * internal-fn.c (gather_load_direct): New macro.
4534 (expand_gather_load_optab_fn): New function.
4535 (direct_gather_load_optab_supported_p): New macro.
4536 (direct_internal_fn_optab): New function.
4537 (internal_load_fn_p): Likewise.
4538 (internal_gather_scatter_fn_p): Likewise.
4539 (internal_fn_mask_index): Likewise.
4540 (internal_gather_scatter_fn_supported_p): Likewise.
4541 * optabs-query.c (supports_at_least_one_mode_p): New function.
4542 (supports_vec_gather_load_p): Likewise.
4543 * optabs-query.h (supports_vec_gather_load_p): Declare.
4544 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4545 and memory_type field.
4546 (NUM_PATTERNS): Bump to 15.
4547 * tree-vect-data-refs.c: Include internal-fn.h.
4548 (vect_gather_scatter_fn_p): New function.
4549 (vect_describe_gather_scatter_call): Likewise.
4550 (vect_check_gather_scatter): Try using internal functions for
4551 gather loads. Recognize existing calls to a gather load function.
4552 (vect_analyze_data_refs): Consider using gather loads if
4553 supports_vec_gather_load_p.
4554 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4555 (vect_get_gather_scatter_offset_type): Likewise.
4556 (vect_convert_mask_for_vectype): Likewise.
4557 (vect_add_conversion_to_patterm): Likewise.
4558 (vect_try_gather_scatter_pattern): Likewise.
4559 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4560 (vect_vect_recog_func_ptrs): Add it.
4561 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4562 internal_fn_mask_index and internal_gather_scatter_fn_p.
4563 (check_load_store_masking): Take the gather_scatter_info as an
4564 argument and handle gather loads.
4565 (vect_get_gather_scatter_ops): New function.
4566 (vectorizable_call): Check internal_load_fn_p.
4567 (vectorizable_load): Likewise. Handle gather load internal
4569 (vectorizable_store): Update call to check_load_store_masking.
4570 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4571 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4572 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4573 (aarch64_gather_scale_operand_d): New predicates.
4574 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4575 (mask_gather_load<mode>): New insns.
4577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4578 Alan Hayward <alan.hayward@arm.com>
4579 David Sherwood <david.sherwood@arm.com>
4581 * optabs.def (fold_left_plus_optab): New optab.
4582 * doc/md.texi (fold_left_plus_@var{m}): Document.
4583 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4584 * internal-fn.c (fold_left_direct): Define.
4585 (expand_fold_left_optab_fn): Likewise.
4586 (direct_fold_left_optab_supported_p): Likewise.
4587 * fold-const-call.c (fold_const_fold_left): New function.
4588 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4589 * tree-parloops.c (valid_reduction_p): New function.
4590 (gather_scalar_reductions): Use it.
4591 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4592 (vect_finish_replace_stmt): Declare.
4593 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4594 (needs_fold_left_reduction_p): New function, split out from...
4595 (vect_is_simple_reduction): ...here. Accept reductions that
4596 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4597 (vect_force_simple_reduction): Also store the reduction type in
4598 the assignment's STMT_VINFO_REDUC_TYPE.
4599 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4600 (merge_with_identity): New function.
4601 (vect_expand_fold_left): Likewise.
4602 (vectorize_fold_left_reduction): Likewise.
4603 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4604 scalar phi in place for it. Check for target support and reject
4605 cases that would reassociate the operation. Defer the transform
4606 phase to vectorize_fold_left_reduction.
4607 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4608 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4609 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4611 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4613 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4614 call to ifc_temp_var.
4616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4617 Alan Hayward <alan.hayward@arm.com>
4618 David Sherwood <david.sherwood@arm.com>
4620 * target.def (legitimize_address_displacement): Take the original
4621 offset as a poly_int.
4622 * targhooks.h (default_legitimize_address_displacement): Update
4624 * targhooks.c (default_legitimize_address_displacement): Likewise.
4625 * doc/tm.texi: Regenerate.
4626 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4627 as an argument, moving assert of ad->disp == ad->disp_term to...
4628 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4629 Try calling targetm.legitimize_address_displacement before expanding
4630 the address rather than afterwards, and adjust for the new interface.
4631 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4632 Match the new hook interface. Handle SVE addresses.
4633 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4638 * Makefile.in (OBJS): Add early-remat.o.
4639 * target.def (select_early_remat_modes): New hook.
4640 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4641 * doc/tm.texi: Regenerate.
4642 * targhooks.h (default_select_early_remat_modes): Declare.
4643 * targhooks.c (default_select_early_remat_modes): New function.
4644 * timevar.def (TV_EARLY_REMAT): New timevar.
4645 * passes.def (pass_early_remat): New pass.
4646 * tree-pass.h (make_pass_early_remat): Declare.
4647 * early-remat.c: New file.
4648 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4650 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4653 Alan Hayward <alan.hayward@arm.com>
4654 David Sherwood <david.sherwood@arm.com>
4656 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4657 vfm1 with a bound_epilog parameter.
4658 (vect_do_peeling): Update calls accordingly, and move the prologue
4659 call earlier in the function. Treat the base bound_epilog as 0 for
4660 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4661 this base when peeling for gaps.
4662 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4663 with fully-masked loops.
4664 (vect_estimate_min_profitable_iters): Handle the single peeled
4665 iteration in that case.
4667 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4668 Alan Hayward <alan.hayward@arm.com>
4669 David Sherwood <david.sherwood@arm.com>
4671 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4672 single-element interleaving even if the size is not a power of 2.
4673 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4674 accesses for single-element interleaving if the group size is
4677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4678 Alan Hayward <alan.hayward@arm.com>
4679 David Sherwood <david.sherwood@arm.com>
4681 * doc/md.texi (fold_extract_last_@var{m}): Document.
4682 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4683 * optabs.def (fold_extract_last_optab): New optab.
4684 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4685 * internal-fn.c (fold_extract_direct): New macro.
4686 (expand_fold_extract_optab_fn): Likewise.
4687 (direct_fold_extract_optab_supported_p): Likewise.
4688 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4689 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4690 EXTRACT_LAST_REDUCTION.
4691 (get_initial_def_for_reduction): Do not create an initial vector
4692 for EXTRACT_LAST_REDUCTION reductions.
4693 (vectorizable_reduction): Leave the scalar phi in place for
4694 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4695 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4696 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4697 transform phase to vectorizable_condition.
4698 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4700 (vect_finish_stmt_generation): ...here.
4701 (vect_finish_replace_stmt): New function.
4702 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4703 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4705 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4708 Alan Hayward <alan.hayward@arm.com>
4709 David Sherwood <david.sherwood@arm.com>
4711 * doc/md.texi (extract_last_@var{m}): Document.
4712 * optabs.def (extract_last_optab): New optab.
4713 * internal-fn.def (EXTRACT_LAST): New internal function.
4714 * internal-fn.c (cond_unary_direct): New macro.
4715 (expand_cond_unary_optab_fn): Likewise.
4716 (direct_cond_unary_optab_supported_p): Likewise.
4717 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4718 loops using EXTRACT_LAST.
4719 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4720 (extract_last_<mode>): ...this optab.
4721 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
4728 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4729 * doc/tm.texi: Regenerate.
4730 * targhooks.h (default_empty_mask_is_expensive): Declare.
4731 * targhooks.c (default_empty_mask_is_expensive): New function.
4732 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4733 if the target says that empty masks are expensive.
4734 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4736 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4739 Alan Hayward <alan.hayward@arm.com>
4740 David Sherwood <david.sherwood@arm.com>
4742 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4743 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4744 (vect_use_loop_mask_for_alignment_p): New function.
4745 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4746 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4747 niters_skip argument. Make sure that the first niters_skip elements
4748 of the first iteration are inactive.
4749 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4750 Update call to vect_set_loop_masks_directly.
4751 (get_misalign_in_elems): New function, split out from...
4752 (vect_gen_prolog_loop_niters): ...here.
4753 (vect_update_init_of_dr): Take a code argument that specifies whether
4754 the adjustment should be added or subtracted.
4755 (vect_update_init_of_drs): Likewise.
4756 (vect_prepare_for_masked_peels): New function.
4757 (vect_do_peeling): Skip prologue peeling if we're using a mask
4758 instead. Update call to vect_update_inits_of_drs.
4759 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4761 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4762 alignment. Do not include the number of peeled iterations in
4763 the minimum threshold in that case.
4764 (vectorizable_induction): Adjust the start value down by
4765 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4766 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4767 Take the number of skipped iterations into account when calculating
4769 * tree-vect-stmts.c (vect_gen_while_not): New function.
4771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4772 Alan Hayward <alan.hayward@arm.com>
4773 David Sherwood <david.sherwood@arm.com>
4775 * doc/sourcebuild.texi (vect_fully_masked): Document.
4776 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4778 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4780 (vect_analyze_loop_2): ...here. Don't check the vectorization
4781 factor against the number of loop iterations if the loop is
4784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4785 Alan Hayward <alan.hayward@arm.com>
4786 David Sherwood <david.sherwood@arm.com>
4788 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4789 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4790 (dump_groups): Update accordingly.
4791 (iv_use::mem_type): New member variable.
4792 (address_p): New function.
4793 (record_use): Add a mem_type argument and initialize the new
4795 (record_group_use): Add a mem_type argument. Use address_p.
4796 Remove obsolete null checks of base_object. Update call to record_use.
4797 (find_interesting_uses_op): Update call to record_group_use.
4798 (find_interesting_uses_cond): Likewise.
4799 (find_interesting_uses_address): Likewise.
4800 (get_mem_type_for_internal_fn): New function.
4801 (find_address_like_use): Likewise.
4802 (find_interesting_uses_stmt): Try find_address_like_use before
4803 calling find_interesting_uses_op.
4804 (addr_offset_valid_p): Use the iv mem_type field as the type
4805 of the addressed memory.
4806 (add_autoinc_candidates): Likewise.
4807 (get_address_cost): Likewise.
4808 (split_small_address_groups_p): Use address_p.
4809 (split_address_groups): Likewise.
4810 (add_iv_candidate_for_use): Likewise.
4811 (autoinc_possible_for_pair): Likewise.
4812 (rewrite_groups): Likewise.
4813 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4814 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4815 (get_alias_ptr_type_for_ptr_address): New function.
4816 (rewrite_use_address): Rewrite address uses in calls that were
4817 identified by find_address_like_use.
4819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4820 Alan Hayward <alan.hayward@arm.com>
4821 David Sherwood <david.sherwood@arm.com>
4823 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4825 * gimple-expr.h (is_gimple_addressable: Likewise.
4826 * gimple-expr.c (is_gimple_address): Likewise.
4827 * internal-fn.c (expand_call_mem_ref): New function.
4828 (expand_mask_load_optab_fn): Use it.
4829 (expand_mask_store_optab_fn): Likewise.
4831 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4832 Alan Hayward <alan.hayward@arm.com>
4833 David Sherwood <david.sherwood@arm.com>
4835 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4836 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4837 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4838 (cond_umax@var{mode}): Document.
4839 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4840 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4841 (cond_umin_optab, cond_umax_optab): New optabs.
4842 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4843 (COND_IOR, COND_XOR): New internal functions.
4844 * internal-fn.h (get_conditional_internal_fn): Declare.
4845 * internal-fn.c (cond_binary_direct): New macro.
4846 (expand_cond_binary_optab_fn): Likewise.
4847 (direct_cond_binary_optab_supported_p): Likewise.
4848 (get_conditional_internal_fn): New function.
4849 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4850 Cope with reduction statements that are vectorized as calls rather
4852 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4853 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4854 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4855 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4856 (UNSPEC_COND_EOR): New unspecs.
4857 (optab): Add mappings for them.
4858 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4859 (sve_int_op, sve_fp_op): New int attributes.
4861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4862 Alan Hayward <alan.hayward@arm.com>
4863 David Sherwood <david.sherwood@arm.com>
4865 * optabs.def (while_ult_optab): New optab.
4866 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4867 * internal-fn.def (WHILE_ULT): New internal function.
4868 * internal-fn.h (direct_internal_fn_supported_p): New override
4869 that takes two types as argument.
4870 * internal-fn.c (while_direct): New macro.
4871 (expand_while_optab_fn): New function.
4872 (convert_optab_supported_p): Likewise.
4873 (direct_while_optab_supported_p): New macro.
4874 * wide-int.h (wi::udiv_ceil): New function.
4875 * tree-vectorizer.h (rgroup_masks): New structure.
4876 (vec_loop_masks): New typedef.
4877 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4879 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4880 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4881 (vect_max_vf): New function.
4882 (slpeel_make_loop_iterate_ntimes): Delete.
4883 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4884 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4885 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4886 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4887 internal-fn.h, stor-layout.h and optabs-query.h.
4888 (vect_set_loop_mask): New function.
4889 (add_preheader_seq): Likewise.
4890 (add_header_seq): Likewise.
4891 (interleave_supported_p): Likewise.
4892 (vect_maybe_permute_loop_masks): Likewise.
4893 (vect_set_loop_masks_directly): Likewise.
4894 (vect_set_loop_condition_masked): Likewise.
4895 (vect_set_loop_condition_unmasked): New function, split out from
4896 slpeel_make_loop_iterate_ntimes.
4897 (slpeel_make_loop_iterate_ntimes): Rename to..
4898 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4899 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4900 (vect_do_peeling): Update call accordingly.
4901 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4903 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4904 mask_compare_type, can_fully_mask_p and fully_masked_p.
4905 (release_vec_loop_masks): New function.
4906 (_loop_vec_info): Use it to free the loop masks.
4907 (can_produce_all_loop_masks_p): New function.
4908 (vect_get_max_nscalars_per_iter): Likewise.
4909 (vect_verify_full_masking): Likewise.
4910 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4911 retries, and free the mask rgroups before retrying. Check loop-wide
4912 reasons for disallowing fully-masked loops. Make the final decision
4913 about whether use a fully-masked loop or not.
4914 (vect_estimate_min_profitable_iters): Do not assume that peeling
4915 for the number of iterations will be needed for fully-masked loops.
4916 (vectorizable_reduction): Disable fully-masked loops.
4917 (vectorizable_live_operation): Likewise.
4918 (vect_halve_mask_nunits): New function.
4919 (vect_double_mask_nunits): Likewise.
4920 (vect_record_loop_mask): Likewise.
4921 (vect_get_loop_mask): Likewise.
4922 (vect_transform_loop): Handle the case in which the final loop
4923 iteration might handle a partial vector. Call vect_set_loop_condition
4924 instead of slpeel_make_loop_iterate_ntimes.
4925 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4926 (check_load_store_masking): New function.
4927 (prepare_load_store_mask): Likewise.
4928 (vectorizable_store): Handle fully-masked loops.
4929 (vectorizable_load): Likewise.
4930 (supportable_widening_operation): Use vect_halve_mask_nunits for
4932 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4933 (vect_gen_while): New function.
4934 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4935 (aarch64_uqdec<mode>): New insn.
4937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4938 Alan Hayward <alan.hayward@arm.com>
4939 David Sherwood <david.sherwood@arm.com>
4941 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4942 (reduc_xor_scal_optab): New optabs.
4943 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4944 (reduc_xor_scal_@var{m}): Document.
4945 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4946 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4948 * fold-const-call.c (fold_const_call): Handle them.
4949 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4950 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4951 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4952 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4953 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4954 (UNSPEC_XORV): New unspecs.
4955 (optab): Add entries for them.
4956 (BITWISEV): New int iterator.
4957 (bit_reduc_op): New int attributes.
4959 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4960 Alan Hayward <alan.hayward@arm.com>
4961 David Sherwood <david.sherwood@arm.com>
4963 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4964 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4965 * optabs.def (vec_shl_insert_optab): New optab.
4966 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4967 (duplicate_and_interleave): Likewise.
4968 * tree-vect-loop.c: Include internal-fn.h.
4969 (neutral_op_for_slp_reduction): New function, split out from
4970 get_initial_defs_for_reduction.
4971 (get_initial_def_for_reduction): Handle option 2 for variable-length
4972 vectors by loading the neutral value into a vector and then shifting
4973 the initial value into element 0.
4974 (get_initial_defs_for_reduction): Replace the code argument with
4975 the neutral value calculated by neutral_op_for_slp_reduction.
4976 Use gimple_build_vector for constant-length vectors.
4977 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4978 but the first group_size elements have a neutral value.
4979 Use duplicate_and_interleave otherwise.
4980 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4981 Update call to get_initial_defs_for_reduction. Handle SLP
4982 reductions for variable-length vectors by creating one vector
4983 result for each scalar result, with the elements associated
4984 with other scalar results stubbed out with the neutral value.
4985 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4986 Require IFN_VEC_SHL_INSERT for double reductions on
4987 variable-length vectors, or SLP reductions that have
4988 a neutral value. Require can_duplicate_and_interleave_p
4989 support for variable-length unchained SLP reductions if there
4990 is no neutral value, such as for MIN/MAX reductions. Also require
4991 the number of vector elements to be a multiple of the number of
4992 SLP statements when doing variable-length unchained SLP reductions.
4993 Update call to vect_create_epilog_for_reduction.
4994 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4995 and remove initial values.
4996 (duplicate_and_interleave): Make public.
4997 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4998 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5000 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5001 Alan Hayward <alan.hayward@arm.com>
5002 David Sherwood <david.sherwood@arm.com>
5004 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5005 (can_duplicate_and_interleave_p): New function.
5006 (vect_get_and_check_slp_defs): Take the vector of statements
5007 rather than just the current one. Remove excess parentheses.
5008 Restriction rejectinon of vect_constant_def and vect_external_def
5009 for variable-length vectors to boolean types, or types for which
5010 can_duplicate_and_interleave_p is false.
5011 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5012 (duplicate_and_interleave): New function.
5013 (vect_get_constant_vectors): Use gimple_build_vector for
5014 constant-length vectors and suitable variable-length constant
5015 vectors. Use duplicate_and_interleave for other variable-length
5016 vectors. Don't defer the update when inserting new statements.
5018 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5019 Alan Hayward <alan.hayward@arm.com>
5020 David Sherwood <david.sherwood@arm.com>
5022 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5023 min_profitable_iters doesn't go negative.
5025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5026 Alan Hayward <alan.hayward@arm.com>
5027 David Sherwood <david.sherwood@arm.com>
5029 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5030 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5031 * optabs.def (vec_mask_load_lanes_optab): New optab.
5032 (vec_mask_store_lanes_optab): Likewise.
5033 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5034 (MASK_STORE_LANES): Likewise.
5035 * internal-fn.c (mask_load_lanes_direct): New macro.
5036 (mask_store_lanes_direct): Likewise.
5037 (expand_mask_load_optab_fn): Handle masked operations.
5038 (expand_mask_load_lanes_optab_fn): New macro.
5039 (expand_mask_store_optab_fn): Handle masked operations.
5040 (expand_mask_store_lanes_optab_fn): New macro.
5041 (direct_mask_load_lanes_optab_supported_p): Likewise.
5042 (direct_mask_store_lanes_optab_supported_p): Likewise.
5043 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5045 (vect_load_lanes_supported): Likewise.
5046 * tree-vect-data-refs.c (strip_conversion): New function.
5047 (can_group_stmts_p): Likewise.
5048 (vect_analyze_data_ref_accesses): Use it instead of checking
5049 for a pair of assignments.
5050 (vect_store_lanes_supported): Take a masked_p parameter.
5051 (vect_load_lanes_supported): Likewise.
5052 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5053 vect_store_lanes_supported and vect_load_lanes_supported.
5054 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5055 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5056 parameter. Don't allow gaps for masked accesses.
5057 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5058 and vect_load_lanes_supported.
5059 (get_load_store_type): Take a masked_p parameter and update
5060 call to get_group_load_store_type.
5061 (vectorizable_store): Update call to get_load_store_type.
5062 Handle IFN_MASK_STORE_LANES.
5063 (vectorizable_load): Update call to get_load_store_type.
5064 Handle IFN_MASK_LOAD_LANES.
5066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5067 Alan Hayward <alan.hayward@arm.com>
5068 David Sherwood <david.sherwood@arm.com>
5070 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5072 * config/aarch64/aarch64-protos.h
5073 (aarch64_sve_struct_memory_operand_p): Declare.
5074 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5075 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5076 (VPRED, vpred): Handle SVE structure modes.
5077 * config/aarch64/constraints.md (Utx): New constraint.
5078 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5079 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5080 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5081 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5082 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5083 structure modes. Split into pieces after RA.
5084 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5085 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5087 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5088 SVE structure modes.
5089 (aarch64_classify_address): Likewise.
5090 (sizetochar): Move earlier in file.
5091 (aarch64_print_operand): Handle SVE register lists.
5092 (aarch64_array_mode): New function.
5093 (aarch64_sve_struct_memory_operand_p): Likewise.
5094 (TARGET_ARRAY_MODE): Redefine.
5096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5097 Alan Hayward <alan.hayward@arm.com>
5098 David Sherwood <david.sherwood@arm.com>
5100 * target.def (array_mode): New target hook.
5101 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5102 * doc/tm.texi: Regenerate.
5103 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5104 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5105 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5107 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5110 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5111 Alan Hayward <alan.hayward@arm.com>
5112 David Sherwood <david.sherwood@arm.com>
5114 * fold-const.c (fold_binary_loc): Check the argument types
5115 rather than the result type when testing for a vector operation.
5117 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5119 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5120 * doc/tm.texi: Regenerate.
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 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5127 (sve): Document new AArch64 extension.
5128 * doc/md.texi (w): Extend the description of the AArch64
5129 constraint to include SVE vectors.
5130 (Upl, Upa): Document new AArch64 predicate constraints.
5131 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5133 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5134 (msve-vector-bits=): New option.
5135 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5136 SVE when these are disabled.
5137 (sve): New extension.
5138 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5139 modes. Adjust their number of units based on aarch64_sve_vg.
5140 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5141 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5142 aarch64_addr_query_type.
5143 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5144 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5145 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5146 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5147 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5148 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5149 (aarch64_simd_imm_zero_p): Delete.
5150 (aarch64_check_zero_based_sve_index_immediate): Declare.
5151 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5152 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5153 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5154 (aarch64_sve_float_mul_immediate_p): Likewise.
5155 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5157 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5158 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5159 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5160 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5161 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5162 (aarch64_regmode_natural_size): Likewise.
5163 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5164 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5166 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5167 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5168 for VG and the SVE predicate registers.
5169 (V_ALIASES): Add a "z"-prefixed alias.
5170 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5171 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5172 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5173 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5174 (REG_CLASS_NAMES): Add entries for them.
5175 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5176 and the predicate registers.
5177 (aarch64_sve_vg): Declare.
5178 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5179 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5180 (REGMODE_NATURAL_SIZE): Define.
5181 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5183 * config/aarch64/aarch64.c: Include cfgrtl.h.
5184 (simd_immediate_info): Add a constructor for series vectors,
5185 and an associated step field.
5186 (aarch64_sve_vg): New variable.
5187 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5188 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5189 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5190 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5191 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5192 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5193 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5194 (aarch64_get_mask_mode): New functions.
5195 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5196 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5197 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5198 predicate modes and predicate registers. Explicitly restrict
5199 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5200 to store a vector mode if it is recognized by
5201 aarch64_classify_vector_mode.
5202 (aarch64_regmode_natural_size): New function.
5203 (aarch64_hard_regno_caller_save_mode): Return the original mode
5205 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5206 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5207 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5208 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5210 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5211 does not overlap dest if the function is frame-related. Handle
5213 (aarch64_split_add_offset): New function.
5214 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5215 them aarch64_add_offset.
5216 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5217 and update call to aarch64_sub_sp.
5218 (aarch64_add_cfa_expression): New function.
5219 (aarch64_expand_prologue): Pass extra temporary registers to the
5220 functions above. Handle the case in which we need to emit new
5221 DW_CFA_expressions for registers that were originally saved
5222 relative to the stack pointer, but now have to be expressed
5223 relative to the frame pointer.
5224 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5226 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5227 IP0 and IP1 values for SVE frames.
5228 (aarch64_expand_vec_series): New function.
5229 (aarch64_expand_sve_widened_duplicate): Likewise.
5230 (aarch64_expand_sve_const_vector): Likewise.
5231 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5232 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5233 into the register, rather than emitting a SET directly.
5234 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5235 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5236 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5237 (offset_9bit_signed_scaled_p): New functions.
5238 (aarch64_replicate_bitmask_imm): New function.
5239 (aarch64_bitmask_imm): Use it.
5240 (aarch64_cannot_force_const_mem): Reject expressions involving
5241 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5242 (aarch64_classify_index): Handle SVE indices, by requiring
5243 a plain register index with a scale that matches the element size.
5244 (aarch64_classify_address): Handle SVE addresses. Assert that
5245 the mode of the address is VOIDmode or an integer mode.
5246 Update call to aarch64_classify_symbol.
5247 (aarch64_classify_symbolic_expression): Update call to
5248 aarch64_classify_symbol.
5249 (aarch64_const_vec_all_in_range_p): New function.
5250 (aarch64_print_vector_float_operand): Likewise.
5251 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5252 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5253 and the FP immediates 1.0 and 0.5.
5254 (aarch64_print_address_internal): Handle SVE addresses.
5255 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5256 (aarch64_regno_regclass): Handle predicate registers.
5257 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5259 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5260 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5261 (aarch64_convert_sve_vector_bits): New function.
5262 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5263 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5265 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5266 Handle SVE vector and predicate modes. Accept VL-based constants
5267 that need only one temporary register, and VL offsets that require
5268 no temporary registers.
5269 (aarch64_conditional_register_usage): Mark the predicate registers
5270 as fixed if SVE isn't available.
5271 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5272 Return true for SVE vector and predicate modes.
5273 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5274 rather than an unsigned int. Handle SVE modes.
5275 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5277 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5279 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5280 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5281 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5282 (aarch64_sve_float_mul_immediate_p): New functions.
5283 (aarch64_sve_valid_immediate): New function.
5284 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5285 Explicitly reject structure modes. Check for INDEX constants.
5286 Handle PTRUE and PFALSE constants.
5287 (aarch64_check_zero_based_sve_index_immediate): New function.
5288 (aarch64_simd_imm_zero_p): Delete.
5289 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5290 vector modes. Accept constants in the range of CNT[BHWD].
5291 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5292 ask for an Advanced SIMD mode.
5293 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5294 (aarch64_simd_vector_alignment): Handle SVE predicates.
5295 (aarch64_vectorize_preferred_vector_alignment): New function.
5296 (aarch64_simd_vector_alignment_reachable): Use it instead of
5298 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5299 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5301 (MAX_VECT_LEN): Delete.
5302 (expand_vec_perm_d): Add a vec_flags field.
5303 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5304 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5305 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5307 (aarch64_evpc_rev): Rename to...
5308 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5309 (aarch64_evpc_rev_global): New function.
5310 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5311 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5313 (aarch64_evpc_sve_tbl): New function.
5314 (aarch64_expand_vec_perm_const_1): Update after rename of
5315 aarch64_evpc_rev. Handle SVE permutes too, trying
5316 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5317 than aarch64_evpc_tbl.
5318 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5319 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5320 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5321 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5322 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5323 (aarch64_expand_sve_vcond): New functions.
5324 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5325 of aarch64_vector_mode_p.
5326 (aarch64_dwarf_poly_indeterminate_value): New function.
5327 (aarch64_compute_pressure_classes): Likewise.
5328 (aarch64_can_change_mode_class): Likewise.
5329 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5330 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5331 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5332 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5333 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5334 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5335 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5336 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5338 (Dn, Dl, Dr): Accept const as well as const_vector.
5339 (Dz): Likewise. Compare against CONST0_RTX.
5340 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5341 of "vector" where appropriate.
5342 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5343 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5344 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5345 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5346 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5347 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5348 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5349 (v_int_equiv): Extend to SVE modes.
5350 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5352 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5353 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5354 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5355 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5356 (SVE_COND_FP_CMP): New int iterators.
5357 (perm_hilo): Handle the new unpack unspecs.
5358 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5360 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5361 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5362 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5363 (aarch64_equality_operator, aarch64_constant_vector_operand)
5364 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5365 (aarch64_sve_nonimmediate_operand): Likewise.
5366 (aarch64_sve_general_operand): Likewise.
5367 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5368 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5369 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5370 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5371 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5372 (aarch64_sve_float_arith_immediate): Likewise.
5373 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5374 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5375 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5376 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5377 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5378 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5379 (aarch64_sve_float_arith_operand): Likewise.
5380 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5381 (aarch64_sve_float_mul_operand): Likewise.
5382 (aarch64_sve_vec_perm_operand): Likewise.
5383 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5384 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5385 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5386 as well as const_vector.
5387 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5388 in file. Use CONST0_RTX and CONSTM1_RTX.
5389 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5390 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5391 Use aarch64_simd_imm_zero.
5392 * config/aarch64/aarch64-sve.md: New file.
5393 * config/aarch64/aarch64.md: Include it.
5394 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5395 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5396 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5397 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5398 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5399 (sve): New attribute.
5400 (enabled): Disable instructions with the sve attribute unless
5402 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5403 aarch64_expand_mov_immediate.
5404 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5405 CNT[BHSD] immediates.
5406 (movti): Split CONST_POLY_INT moves into two halves.
5407 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5408 Split additions that need a temporary here if the destination
5409 is the stack pointer.
5410 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5411 (*add<mode>3_poly_1): New instruction.
5412 (set_clobber_cc): New expander.
5414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5416 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5417 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5418 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5419 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5420 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5421 Change innermode from fixed_mode_size to machine_mode.
5422 (simplify_subreg): Update call accordingly. Handle a constant-sized
5423 subreg of a variable-length CONST_VECTOR.
5425 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5426 Alan Hayward <alan.hayward@arm.com>
5427 David Sherwood <david.sherwood@arm.com>
5429 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5430 (add_offset_to_base): New function, split out from...
5431 (create_mem_ref): ...here. When handling a scale other than 1,
5432 check first whether the address is valid without the offset.
5433 Add it into the base if so, leaving the index and scale as-is.
5435 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5438 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5439 fold_for_warn before checking if arg2 is INTEGER_CST.
5441 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5443 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5444 (store_multiple_operation): Delete.
5445 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5446 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5447 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5448 guarded by TARGET_STRING.
5449 (rs6000_output_load_multiple): Delete.
5450 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5451 OPTION_MASK_STRING / TARGET_STRING handling.
5452 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5453 (const rs6000_opt_masks) <"string">: Change mask to 0.
5454 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5455 (MASK_STRING): Delete.
5456 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5458 (load_multiple): Delete.
5465 (store_multiple): Delete.
5472 (movmemsi_8reg): Delete.
5473 (corresponding unnamed define_insn): Delete.
5474 (movmemsi_6reg): Delete.
5475 (corresponding unnamed define_insn): Delete.
5476 (movmemsi_4reg): Delete.
5477 (corresponding unnamed define_insn): Delete.
5478 (movmemsi_2reg): Delete.
5479 (corresponding unnamed define_insn): Delete.
5480 (movmemsi_1reg): Delete.
5481 (corresponding unnamed define_insn): Delete.
5482 * config/rs6000/rs6000.opt (mno-string): New.
5483 (mstring): Replace by deprecation warning stub.
5484 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5486 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5488 * regrename.c (regrename_do_replace): If replacing the same
5489 reg multiple times, try to reuse last created gen_raw_REG.
5492 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5493 main to workaround a bug in GDB.
5495 2018-01-12 Tom de Vries <tom@codesourcery.com>
5498 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5500 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5502 PR rtl-optimization/80481
5503 * ira-color.c (get_cap_member): New function.
5504 (allocnos_conflict_by_live_ranges_p): Use it.
5505 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5506 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5508 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5511 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5512 (*saddl_se_1): Ditto.
5514 (*ssubl_se_1): Ditto.
5516 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5518 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5519 rather than wi::to_widest for DR_INITs.
5520 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5521 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5522 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5524 (vect_analyze_group_access_1): Note that here.
5526 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5528 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5529 polynomial type sizes.
5531 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5533 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5534 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5535 (gimple_add_tmp_var): Likewise.
5537 2018-01-12 Martin Liska <mliska@suse.cz>
5539 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5540 (gimple_alloc_sizes): Likewise.
5541 (dump_gimple_statistics): Use PRIu64 in printf format.
5542 * gimple.h: Change uint64_t to int.
5544 2018-01-12 Martin Liska <mliska@suse.cz>
5546 * tree-core.h: Use uint64_t instead of int.
5547 * tree.c (tree_node_counts): Likewise.
5548 (tree_node_sizes): Likewise.
5549 (dump_tree_statistics): Use PRIu64 in printf format.
5551 2018-01-12 Martin Liska <mliska@suse.cz>
5553 * Makefile.in: As qsort_chk is implemented in vec.c, add
5554 vec.o to linkage of gencfn-macros.
5555 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5556 passing the info to record_node_allocation_statistics.
5557 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5559 * ggc-common.c (struct ggc_usage): Add operator== and use
5560 it in operator< and compare function.
5561 * mem-stats.h (struct mem_usage): Likewise.
5562 * vec.c (struct vec_usage): Remove operator< and compare
5563 function. Can be simply inherited.
5565 2018-01-12 Martin Jambor <mjambor@suse.cz>
5568 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5569 * tree-ssa-math-opts.c: Include domwalk.h.
5570 (convert_mult_to_fma_1): New function.
5571 (fma_transformation_info): New type.
5572 (fma_deferring_state): Likewise.
5573 (cancel_fma_deferring): New function.
5574 (result_of_phi): Likewise.
5575 (last_fma_candidate_feeds_initial_phi): Likewise.
5576 (convert_mult_to_fma): Added deferring logic, split actual
5577 transformation to convert_mult_to_fma_1.
5578 (math_opts_dom_walker): New type.
5579 (math_opts_dom_walker::after_dom_children): New method, body moved
5580 here from pass_optimize_widening_mul::execute, added deferring logic
5582 (pass_optimize_widening_mul::execute): Moved most of code to
5583 math_opts_dom_walker::after_dom_children.
5584 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5585 * config/i386/i386.c (ix86_option_override_internal): Added
5586 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5588 2018-01-12 Richard Biener <rguenther@suse.de>
5591 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5592 inline instance vars.
5594 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5597 * config/rx/rx.c (rx_is_restricted_memory_address):
5600 2018-01-12 Richard Biener <rguenther@suse.de>
5602 PR tree-optimization/80846
5603 * target.def (split_reduction): New target hook.
5604 * targhooks.c (default_split_reduction): New function.
5605 * targhooks.h (default_split_reduction): Declare.
5606 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5607 target requests first reduce vectors by combining low and high
5609 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5610 (get_vectype_for_scalar_type_and_size): Export.
5611 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5612 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5613 * doc/tm.texi: Regenerate.
5614 * config/i386/i386.c (ix86_split_reduction): Implement
5615 TARGET_VECTORIZE_SPLIT_REDUCTION.
5617 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5620 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5621 in PIC mode except for TARGET_VXWORKS_RTP.
5622 * config/sparc/sparc.c: Include cfgrtl.h.
5623 (TARGET_INIT_PIC_REG): Define.
5624 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5625 (sparc_pic_register_p): New predicate.
5626 (sparc_legitimate_address_p): Use it.
5627 (sparc_legitimize_pic_address): Likewise.
5628 (sparc_delegitimize_address): Likewise.
5629 (sparc_mode_dependent_address_p): Likewise.
5630 (gen_load_pcrel_sym): Remove 4th parameter.
5631 (load_got_register): Adjust call to above. Remove obsolete stuff.
5632 (sparc_expand_prologue): Do not call load_got_register here.
5633 (sparc_flat_expand_prologue): Likewise.
5634 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5635 (sparc_use_pseudo_pic_reg): New function.
5636 (sparc_init_pic_reg): Likewise.
5637 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5638 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5640 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5642 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5643 Add item for branch_cost.
5645 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5647 PR rtl-optimization/83565
5648 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5649 not extend the result to a larger mode for rotate operations.
5650 (num_sign_bit_copies1): Likewise.
5652 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5655 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5657 Use values-Xc.o for -pedantic.
5658 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5660 2018-01-12 Martin Liska <mliska@suse.cz>
5663 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5665 (possible_polymorphic_call_targets): Use it.
5666 (ipa_devirt): Likewise.
5668 2018-01-12 Martin Liska <mliska@suse.cz>
5670 * profile-count.h (enum profile_quality): Use 0 as invalid
5671 enum value of profile_quality.
5673 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5675 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5676 -mext-string options.
5678 2018-01-12 Richard Biener <rguenther@suse.de>
5680 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5681 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5682 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5684 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5686 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5688 * configure.ac (--with-long-double-format): Add support for the
5689 configuration option to change the default long double format on
5691 * config.gcc (powerpc*-linux*-*): Likewise.
5692 * configure: Regenerate.
5693 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5694 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5695 used without modification.
5697 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5699 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5700 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5701 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5702 MISC_BUILTIN_SPEC_BARRIER.
5703 (rs6000_init_builtins): Likewise.
5704 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5706 (speculation_barrier): New define_insn.
5707 * doc/extend.texi: Document __builtin_speculation_barrier.
5709 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5712 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5713 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5714 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5716 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5717 integral modes instead of "ss" and "sd".
5718 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5719 vectors with 32-bit and 64-bit elements.
5720 (vecdupssescalarmodesuffix): New mode attribute.
5721 (vec_dup<mode>): Use it.
5723 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5726 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5727 frame if argument is passed on stack.
5729 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5732 * ree.c (combine_reaching_defs): Optimize also
5733 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5734 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5736 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5739 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5741 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5744 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5745 after they are computed.
5747 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5749 PR tree-optimization/83695
5750 * gimple-loop-linterchange.cc
5751 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5752 reset cached scev information after interchange.
5753 (pass_linterchange::execute): Remove call to scev_reset_htab.
5755 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5757 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5758 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5759 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5760 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5761 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5762 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5763 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5764 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5765 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5766 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5767 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5768 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5769 (V_lane_reg): Likewise.
5770 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5772 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5773 (vfmal_lane_low<mode>_intrinsic,
5774 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5775 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5776 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5777 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5778 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5779 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5781 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5783 * config/arm/arm-cpus.in (fp16fml): New feature.
5784 (ALL_SIMD): Add fp16fml.
5785 (armv8.2-a): Add fp16fml as an option.
5786 (armv8.3-a): Likewise.
5787 (armv8.4-a): Add fp16fml as part of fp16.
5788 * config/arm/arm.h (TARGET_FP16FML): Define.
5789 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5791 * config/arm/arm-modes.def (V2HF): Define.
5792 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5793 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5794 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5795 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5796 vfmsl_low, vfmsl_high): New set of builtins.
5797 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5798 (vfml_op): New code attribute.
5799 (VFMLHALVES): New int iterator.
5800 (VFML, VFMLSEL): New mode attributes.
5801 (V_reg): Define mapping for V2HF.
5802 (V_hi, V_lo): New mode attributes.
5803 (VF_constraint): Likewise.
5804 (vfml_half, vfml_half_selector): New int attributes.
5805 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5807 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5808 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5810 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5811 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5812 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5813 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5815 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5816 Document new effective target and option set.
5818 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5820 * config/arm/arm-cpus.in (armv8_4): New feature.
5821 (ARMv8_4a): New fgroup.
5822 (armv8.4-a): New arch.
5823 * config/arm/arm-tables.opt: Regenerate.
5824 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5825 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5826 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5827 Add matching rules for -march=armv8.4-a and extensions.
5828 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5830 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5833 * config/rx/rx.md (BW): New mode attribute.
5834 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5836 2018-01-11 Richard Biener <rguenther@suse.de>
5838 PR tree-optimization/83435
5839 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5840 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5841 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5843 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5844 Alan Hayward <alan.hayward@arm.com>
5845 David Sherwood <david.sherwood@arm.com>
5847 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5849 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5850 (aarch64_print_address_internal): Use it to check for a zero offset.
5852 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5853 Alan Hayward <alan.hayward@arm.com>
5854 David Sherwood <david.sherwood@arm.com>
5856 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5857 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5858 Return a poly_int64 rather than a HOST_WIDE_INT.
5859 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5860 rather than a HOST_WIDE_INT.
5861 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5862 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5863 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5864 final_offset from HOST_WIDE_INT to poly_int64.
5865 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5866 to_constant when getting the number of units in an Advanced SIMD
5868 (aarch64_builtin_vectorized_function): Check for a constant number
5870 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5872 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5873 attribute instead of GET_MODE_NUNITS.
5874 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5875 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5876 GET_MODE_SIZE for fixed-size registers.
5877 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5878 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5879 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5880 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5881 (aarch64_print_operand, aarch64_print_address_internal)
5882 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5883 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5884 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5885 Handle polynomial GET_MODE_SIZE.
5886 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5887 wider than SImode without modification.
5888 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5889 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5890 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5891 passing and returning SVE modes.
5892 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5893 rather than GEN_INT.
5894 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5895 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5896 (aarch64_allocate_and_probe_stack_space): Likewise.
5897 (aarch64_layout_frame): Cope with polynomial offsets.
5898 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5899 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5901 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5902 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5903 poly_int64 rather than a HOST_WIDE_INT.
5904 (aarch64_get_separate_components, aarch64_process_components)
5905 (aarch64_expand_prologue, aarch64_expand_epilogue)
5906 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5907 (aarch64_anchor_offset): New function, split out from...
5908 (aarch64_legitimize_address): ...here.
5909 (aarch64_builtin_vectorization_cost): Handle polynomial
5910 TYPE_VECTOR_SUBPARTS.
5911 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5913 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5914 number of elements from the PARALLEL rather than the mode.
5915 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5916 rather than GET_MODE_BITSIZE.
5917 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5918 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5919 (aarch64_expand_vec_perm_const_1): Handle polynomial
5920 d->perm.length () and d->perm elements.
5921 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5922 Apply to_constant to d->perm elements.
5923 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5924 polynomial CONST_VECTOR_NUNITS.
5925 (aarch64_move_pointer): Take amount as a poly_int64 rather
5927 (aarch64_progress_pointer): Avoid temporary variable.
5928 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5929 the mode attribute instead of GET_MODE.
5931 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5932 Alan Hayward <alan.hayward@arm.com>
5933 David Sherwood <david.sherwood@arm.com>
5935 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5936 x exists before using it.
5937 (aarch64_add_constant_internal): Rename to...
5938 (aarch64_add_offset_1): ...this. Replace regnum with separate
5939 src and dest rtxes. Handle the case in which they're different,
5940 including when the offset is zero. Replace scratchreg with an rtx.
5941 Use 2 additions if there is no spare register into which we can
5942 move a 16-bit constant.
5943 (aarch64_add_constant): Delete.
5944 (aarch64_add_offset): Replace reg with separate src and dest
5945 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5946 Use aarch64_add_offset_1.
5947 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5948 an rtx rather than an int. Take the delta as a poly_int64
5949 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5950 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5951 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5952 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5953 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5955 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5956 aarch64_add_constant.
5958 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5960 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5961 Use scalar_float_mode.
5963 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5965 * config/aarch64/aarch64-simd.md
5966 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5967 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5968 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5969 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5970 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5971 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5972 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5973 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5974 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5975 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5977 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5980 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5981 targ_options->x_arm_arch_string is non NULL.
5983 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5985 * config/aarch64/aarch64.h
5986 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5988 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5991 * expmed.c (emit_store_flag_force): Swap if const op0
5992 and change VOIDmode to mode of op0.
5994 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5996 PR rtl-optimization/83761
5997 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5998 than bytes to mode_for_size.
6000 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6003 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6004 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6007 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6010 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6011 when in layout mode.
6012 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6013 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6016 2018-01-10 Michael Collison <michael.collison@arm.com>
6018 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6019 * config/aarch64/aarch64-option-extension.def: Add
6020 AARCH64_OPT_EXTENSION of 'fp16fml'.
6021 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6022 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6023 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6024 * config/aarch64/constraints.md (Ui7): New constraint.
6025 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6026 (VFMLA_SEL_W): Ditto.
6029 (VFMLA16_LOW): New int iterator.
6030 (VFMLA16_HIGH): Ditto.
6031 (UNSPEC_FMLAL): New unspec.
6032 (UNSPEC_FMLSL): Ditto.
6033 (UNSPEC_FMLAL2): Ditto.
6034 (UNSPEC_FMLSL2): Ditto.
6035 (f16mac): New code attribute.
6036 * config/aarch64/aarch64-simd-builtins.def
6037 (aarch64_fmlal_lowv2sf): Ditto.
6038 (aarch64_fmlsl_lowv2sf): Ditto.
6039 (aarch64_fmlalq_lowv4sf): Ditto.
6040 (aarch64_fmlslq_lowv4sf): Ditto.
6041 (aarch64_fmlal_highv2sf): Ditto.
6042 (aarch64_fmlsl_highv2sf): Ditto.
6043 (aarch64_fmlalq_highv4sf): Ditto.
6044 (aarch64_fmlslq_highv4sf): Ditto.
6045 (aarch64_fmlal_lane_lowv2sf): Ditto.
6046 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6047 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6048 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6049 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6050 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6051 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6052 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6053 (aarch64_fmlal_lane_highv2sf): Ditto.
6054 (aarch64_fmlsl_lane_highv2sf): Ditto.
6055 (aarch64_fmlal_laneq_highv2sf): Ditto.
6056 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6057 (aarch64_fmlalq_lane_highv4sf): Ditto.
6058 (aarch64_fmlsl_lane_highv4sf): Ditto.
6059 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6060 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6061 * config/aarch64/aarch64-simd.md:
6062 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6063 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6064 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6065 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6066 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6067 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6068 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6069 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6070 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6071 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6072 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6073 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6074 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6075 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6076 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6077 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6078 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6079 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6080 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6081 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6082 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6083 (vfmlsl_low_u32): Ditto.
6084 (vfmlalq_low_u32): Ditto.
6085 (vfmlslq_low_u32): Ditto.
6086 (vfmlal_high_u32): Ditto.
6087 (vfmlsl_high_u32): Ditto.
6088 (vfmlalq_high_u32): Ditto.
6089 (vfmlslq_high_u32): Ditto.
6090 (vfmlal_lane_low_u32): Ditto.
6091 (vfmlsl_lane_low_u32): Ditto.
6092 (vfmlal_laneq_low_u32): Ditto.
6093 (vfmlsl_laneq_low_u32): Ditto.
6094 (vfmlalq_lane_low_u32): Ditto.
6095 (vfmlslq_lane_low_u32): Ditto.
6096 (vfmlalq_laneq_low_u32): Ditto.
6097 (vfmlslq_laneq_low_u32): Ditto.
6098 (vfmlal_lane_high_u32): Ditto.
6099 (vfmlsl_lane_high_u32): Ditto.
6100 (vfmlal_laneq_high_u32): Ditto.
6101 (vfmlsl_laneq_high_u32): Ditto.
6102 (vfmlalq_lane_high_u32): Ditto.
6103 (vfmlslq_lane_high_u32): Ditto.
6104 (vfmlalq_laneq_high_u32): Ditto.
6105 (vfmlslq_laneq_high_u32): Ditto.
6106 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6107 (AARCH64_FL_FOR_ARCH8_4): New.
6108 (AARCH64_ISA_F16FML): New ISA flag.
6109 (TARGET_F16FML): New feature flag for fp16fml.
6110 (doc/invoke.texi): Document new fp16fml option.
6112 2018-01-10 Michael Collison <michael.collison@arm.com>
6114 * config/aarch64/aarch64-builtins.c:
6115 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6116 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6117 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6118 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6119 (AARCH64_ISA_SHA3): New ISA flag.
6120 (TARGET_SHA3): New feature flag for sha3.
6121 * config/aarch64/iterators.md (sha512_op): New int attribute.
6122 (CRYPTO_SHA512): New int iterator.
6123 (UNSPEC_SHA512H): New unspec.
6124 (UNSPEC_SHA512H2): Ditto.
6125 (UNSPEC_SHA512SU0): Ditto.
6126 (UNSPEC_SHA512SU1): Ditto.
6127 * config/aarch64/aarch64-simd-builtins.def
6128 (aarch64_crypto_sha512hqv2di): New builtin.
6129 (aarch64_crypto_sha512h2qv2di): Ditto.
6130 (aarch64_crypto_sha512su0qv2di): Ditto.
6131 (aarch64_crypto_sha512su1qv2di): Ditto.
6132 (aarch64_eor3qv8hi): Ditto.
6133 (aarch64_rax1qv2di): Ditto.
6134 (aarch64_xarqv2di): Ditto.
6135 (aarch64_bcaxqv8hi): Ditto.
6136 * config/aarch64/aarch64-simd.md:
6137 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6138 (aarch64_crypto_sha512su0qv2di): Ditto.
6139 (aarch64_crypto_sha512su1qv2di): Ditto.
6140 (aarch64_eor3qv8hi): Ditto.
6141 (aarch64_rax1qv2di): Ditto.
6142 (aarch64_xarqv2di): Ditto.
6143 (aarch64_bcaxqv8hi): Ditto.
6144 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6145 (vsha512h2q_u64): Ditto.
6146 (vsha512su0q_u64): Ditto.
6147 (vsha512su1q_u64): Ditto.
6148 (veor3q_u16): Ditto.
6149 (vrax1q_u64): Ditto.
6151 (vbcaxq_u16): Ditto.
6152 * config/arm/types.md (crypto_sha512): New type attribute.
6153 (crypto_sha3): Ditto.
6154 (doc/invoke.texi): Document new sha3 option.
6156 2018-01-10 Michael Collison <michael.collison@arm.com>
6158 * config/aarch64/aarch64-builtins.c:
6159 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6160 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6161 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6162 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6163 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6164 (AARCH64_ISA_SM4): New ISA flag.
6165 (TARGET_SM4): New feature flag for sm4.
6166 * config/aarch64/aarch64-simd-builtins.def
6167 (aarch64_sm3ss1qv4si): Ditto.
6168 (aarch64_sm3tt1aq4si): Ditto.
6169 (aarch64_sm3tt1bq4si): Ditto.
6170 (aarch64_sm3tt2aq4si): Ditto.
6171 (aarch64_sm3tt2bq4si): Ditto.
6172 (aarch64_sm3partw1qv4si): Ditto.
6173 (aarch64_sm3partw2qv4si): Ditto.
6174 (aarch64_sm4eqv4si): Ditto.
6175 (aarch64_sm4ekeyqv4si): Ditto.
6176 * config/aarch64/aarch64-simd.md:
6177 (aarch64_sm3ss1qv4si): Ditto.
6178 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6179 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6180 (aarch64_sm4eqv4si): Ditto.
6181 (aarch64_sm4ekeyqv4si): Ditto.
6182 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6183 (sm3part_op): Ditto.
6184 (CRYPTO_SM3TT): Ditto.
6185 (CRYPTO_SM3PART): Ditto.
6186 (UNSPEC_SM3SS1): New unspec.
6187 (UNSPEC_SM3TT1A): Ditto.
6188 (UNSPEC_SM3TT1B): Ditto.
6189 (UNSPEC_SM3TT2A): Ditto.
6190 (UNSPEC_SM3TT2B): Ditto.
6191 (UNSPEC_SM3PARTW1): Ditto.
6192 (UNSPEC_SM3PARTW2): Ditto.
6193 (UNSPEC_SM4E): Ditto.
6194 (UNSPEC_SM4EKEY): Ditto.
6195 * config/aarch64/constraints.md (Ui2): New constraint.
6196 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6197 * config/arm/types.md (crypto_sm3): New type attribute.
6198 (crypto_sm4): Ditto.
6199 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6200 (vsm3tt1aq_u32): Ditto.
6201 (vsm3tt1bq_u32): Ditto.
6202 (vsm3tt2aq_u32): Ditto.
6203 (vsm3tt2bq_u32): Ditto.
6204 (vsm3partw1q_u32): Ditto.
6205 (vsm3partw2q_u32): Ditto.
6206 (vsm4eq_u32): Ditto.
6207 (vsm4ekeyq_u32): Ditto.
6208 (doc/invoke.texi): Document new sm4 option.
6210 2018-01-10 Michael Collison <michael.collison@arm.com>
6212 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6213 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6214 (AARCH64_FL_FOR_ARCH8_4): New.
6215 (AARCH64_FL_V8_4): New flag.
6216 (doc/invoke.texi): Document new armv8.4-a option.
6218 2018-01-10 Michael Collison <michael.collison@arm.com>
6220 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6221 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6222 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6223 * config/aarch64/aarch64-option-extension.def: Add
6224 AARCH64_OPT_EXTENSION of 'sha2'.
6225 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6226 (crypto): Disable sha2 and aes if crypto disabled.
6227 (crypto): Enable aes and sha2 if enabled.
6228 (simd): Disable sha2 and aes if simd disabled.
6229 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6231 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6232 (TARGET_SHA2): New feature flag for sha2.
6233 (TARGET_AES): New feature flag for aes.
6234 * config/aarch64/aarch64-simd.md:
6235 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6236 conditional on TARGET_AES.
6237 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6238 (aarch64_crypto_sha1hsi): Make pattern conditional
6240 (aarch64_crypto_sha1hv4si): Ditto.
6241 (aarch64_be_crypto_sha1hv4si): Ditto.
6242 (aarch64_crypto_sha1su1v4si): Ditto.
6243 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6244 (aarch64_crypto_sha1su0v4si): Ditto.
6245 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6246 (aarch64_crypto_sha256su0v4si): Ditto.
6247 (aarch64_crypto_sha256su1v4si): Ditto.
6248 (doc/invoke.texi): Document new aes and sha2 options.
6250 2018-01-10 Martin Sebor <msebor@redhat.com>
6252 PR tree-optimization/83781
6253 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6256 2018-01-11 Martin Sebor <msebor@gmail.com>
6257 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6259 PR tree-optimization/83501
6260 PR tree-optimization/81703
6262 * tree-ssa-strlen.c (get_string_cst): Rename...
6263 (get_string_len): ...to this. Handle global constants.
6264 (handle_char_store): Adjust.
6266 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6267 Jim Wilson <jimw@sifive.com>
6269 * config/riscv/riscv-protos.h (riscv_output_return): New.
6270 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6271 (riscv_attribute_table, riscv_output_return),
6272 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6273 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6274 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6276 (riscv_expand_prologue): Add early return for naked function.
6277 (riscv_expand_epilogue): Likewise.
6278 (riscv_function_ok_for_sibcall): Return false for naked function.
6279 (riscv_set_current_function): New.
6280 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6281 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6282 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6283 * doc/extend.texi (RISC-V Function Attributes): New.
6285 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6287 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6288 check for 128-bit long double before checking TCmode.
6289 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6290 128-bit long doubles before checking TFmode or TCmode.
6291 (FLOAT128_IBM_P): Likewise.
6293 2018-01-10 Martin Sebor <msebor@redhat.com>
6295 PR tree-optimization/83671
6296 * builtins.c (c_strlen): Unconditionally return zero for the empty
6298 Use -Warray-bounds for warnings.
6299 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6300 for non-constant array indices with COMPONENT_REF, arrays of
6301 arrays, and pointers to arrays.
6302 (gimple_fold_builtin_strlen): Determine and set length range for
6303 non-constant character arrays.
6305 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6308 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6311 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6313 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6315 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6318 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6319 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6320 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6321 indexed_or_indirect_operand predicate.
6322 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6323 (*vsx_le_perm_load_v8hi): Likewise.
6324 (*vsx_le_perm_load_v16qi): Likewise.
6325 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6326 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6327 (*vsx_le_perm_store_v8hi): Likewise.
6328 (*vsx_le_perm_store_v16qi): Likewise.
6329 (eight unnamed splitters): Likewise.
6331 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6333 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6334 * config/rs6000/emmintrin.h: Likewise.
6335 * config/rs6000/mmintrin.h: Likewise.
6336 * config/rs6000/xmmintrin.h: Likewise.
6338 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6341 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6343 * tree.c (tree_nop_conversion): Return true for location wrapper
6345 (maybe_wrap_with_location): New function.
6346 (selftest::check_strip_nops): New function.
6347 (selftest::test_location_wrappers): New function.
6348 (selftest::tree_c_tests): Call it.
6349 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6350 (maybe_wrap_with_location): New decl.
6351 (EXPR_LOCATION_WRAPPER_P): New macro.
6352 (location_wrapper_p): New inline function.
6353 (tree_strip_any_location_wrapper): New inline function.
6355 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6358 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6359 stack_realign_offset for the largest alignment of stack slot
6361 (ix86_find_max_used_stack_alignment): New function.
6362 (ix86_finalize_stack_frame_flags): Use it. Set
6363 max_used_stack_alignment if we don't realign stack.
6364 * config/i386/i386.h (machine_function): Add
6365 max_used_stack_alignment.
6367 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6369 * config/arm/arm.opt (-mbranch-cost): New option.
6370 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6373 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6376 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6377 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6379 2018-01-10 Richard Biener <rguenther@suse.de>
6382 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6383 early out so it also covers the case where we have a non-NULL
6386 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6388 PR tree-optimization/83753
6389 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6390 for non-strided grouped accesses if the number of elements is 1.
6392 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6395 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6396 * i386.h (TARGET_USE_GATHER): Define.
6397 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6399 2018-01-10 Martin Liska <mliska@suse.cz>
6402 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6403 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6405 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6406 CLEANUP_NO_PARTITIONING is not set.
6408 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6410 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6411 for vectors, as a partial revert of r254296.
6412 * rtl.h (const_vec_p): Delete.
6413 (const_vec_duplicate_p): Don't test for vector CONSTs.
6414 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6415 * expmed.c (make_tree): Likewise.
6418 * common.md (E, F): Use CONSTANT_P instead of checking for
6420 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6421 checking for CONST_VECTOR.
6423 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6426 * predict.c (force_edge_cold): Handle in more sane way edges
6429 2018-01-09 Carl Love <cel@us.ibm.com>
6431 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6433 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6434 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6435 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6436 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6437 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6438 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6439 * config/rs6000/rs6000-protos.h: Add extern defition for
6440 rs6000_generate_float2_double_code.
6441 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6443 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6444 (float2_v2df): Add define_expand.
6446 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6449 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6450 op_mode in the force_to_mode call.
6452 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6454 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6455 instead of checking each element individually.
6456 (aarch64_evpc_uzp): Likewise.
6457 (aarch64_evpc_zip): Likewise.
6458 (aarch64_evpc_ext): Likewise.
6459 (aarch64_evpc_rev): Likewise.
6460 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6461 instead of checking each element individually. Return true without
6463 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6464 whether all selected elements come from the same input, instead of
6465 checking each element individually. Remove calls to gen_rtx_REG,
6466 start_sequence and end_sequence and instead assert that no rtl is
6469 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6471 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6472 order of HIGH and CONST checks.
6474 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6476 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6477 if the destination isn't an SSA_NAME.
6479 2018-01-09 Richard Biener <rguenther@suse.de>
6481 PR tree-optimization/83668
6482 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6484 (canonicalize_loop_form): ... here, renamed from ...
6485 (canonicalize_loop_closed_ssa_form): ... this and amended to
6486 swap successor edges for loop exit blocks to make us use
6487 the RPO order we need for initial schedule generation.
6489 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6491 PR tree-optimization/64811
6492 * match.pd: When optimizing comparisons with Inf, avoid
6493 introducing or losing exceptions from comparisons with NaN.
6495 2018-01-09 Martin Liska <mliska@suse.cz>
6498 * asan.c (shadow_mem_size): Add gcc_assert.
6500 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6502 Don't save registers in main().
6505 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6506 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6507 * config/avr/avr.c (avr_set_current_function): Don't error if
6508 naked, OS_task or OS_main are specified at the same time.
6509 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6511 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6513 * common/config/avr/avr-common.c (avr_option_optimization_table):
6514 Switch on -mmain-is-OS_task for optimizing compilations.
6516 2018-01-09 Richard Biener <rguenther@suse.de>
6518 PR tree-optimization/83572
6519 * graphite.c: Include cfganal.h.
6520 (graphite_transform_loops): Connect infinite loops to exit
6521 and remove fake edges at the end.
6523 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6525 * ipa-inline.c (edge_badness): Revert accidental checkin.
6527 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6530 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6531 symbols; not inline clones.
6533 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6536 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6537 hard registers. Formatting fixes.
6539 PR preprocessor/83722
6540 * gcc.c (try_generate_repro): Pass
6541 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6542 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6545 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6546 Kito Cheng <kito.cheng@gmail.com>
6548 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6549 (riscv_leaf_function_p): Delete.
6550 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6552 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6554 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6556 (do_ifelse): New function.
6557 (do_isel): New function.
6558 (do_sub3): New function.
6559 (do_add3): New function.
6560 (do_load_mask_compare): New function.
6561 (do_overlap_load_compare): New function.
6562 (expand_compare_loop): New function.
6563 (expand_block_compare): Call expand_compare_loop() when appropriate.
6564 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6566 (-mblock-compare-inline-loop-limit): New option.
6568 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6571 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6572 Reverse order of second and third operands in first alternative.
6573 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6574 of first and second elements in UNSPEC_VPERMR vector.
6575 (altivec_expand_vec_perm_le): Likewise.
6577 2018-01-08 Jeff Law <law@redhat.com>
6579 PR rtl-optimizatin/81308
6580 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6581 (process_switch): If group_case_labels makes a change, then set
6583 (pass_convert_switch::execute): If a switch is converted, then
6584 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6586 PR rtl-optimization/81308
6587 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6590 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6592 PR target/83663 - Revert r255946
6593 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6594 generation for cases where splatting a value is not useful.
6595 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6596 across a vec_duplicate and a paradoxical subreg forming a vector
6597 mode to a vec_concat.
6599 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6601 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6602 -march=armv8.3-a variants.
6603 * config/arm/t-multilib: Likewise.
6604 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6606 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6608 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6610 (cceq_ior_compare_complement): Give it a name so I can use it, and
6611 change boolean_or_operator predicate to boolean_operator so it can
6612 be used to generate a crand.
6613 (eqne): New code iterator.
6614 (bd/bd_neg): New code_attrs.
6615 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6616 a single define_insn.
6617 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6618 decrement (bdnzt/bdnzf/bdzt/bdzf).
6619 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6620 with the new names of the branch decrement patterns, and added the
6621 names of the branch decrement conditional patterns.
6623 2018-01-08 Richard Biener <rguenther@suse.de>
6625 PR tree-optimization/83563
6626 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6629 2018-01-08 Richard Biener <rguenther@suse.de>
6632 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6634 2018-01-08 Richard Biener <rguenther@suse.de>
6636 PR tree-optimization/83685
6637 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6638 references to abnormals.
6640 2018-01-08 Richard Biener <rguenther@suse.de>
6643 * dwarf2out.c (output_indirect_strings): Handle empty
6644 skeleton_debug_str_hash.
6645 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6647 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6649 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6650 (emit_store_direct): Likewise.
6651 (arc_trampoline_adjust_address): Likewise.
6652 (arc_asm_trampoline_template): New function.
6653 (arc_initialize_trampoline): Use asm_trampoline_template.
6654 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6655 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6656 * config/arc/arc.md (flush_icache): Delete pattern.
6658 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6660 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6661 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6664 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6667 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6668 by not USED_FOR_TARGET.
6669 (make_pass_resolve_sw_modes): Likewise.
6671 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6673 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6676 2018-01-08 Richard Biener <rguenther@suse.de>
6679 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6681 2018-01-08 Richard Biener <rguenther@suse.de>
6684 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6686 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6689 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6690 basic blocks with a small number of successors.
6691 (convert_control_dep_chain_into_preds): Improve handling of
6693 (dump_predicates): Split apart into...
6694 (dump_pred_chain): ...here...
6695 (dump_pred_info): ...and here.
6696 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6697 (can_chain_union_be_invalidated_p): Improve check for invalidation
6699 (uninit_uses_cannot_happen): Avoid unnecessary if
6700 convert_control_dep_chain_into_preds yielded nothing.
6702 2018-01-06 Martin Sebor <msebor@redhat.com>
6704 PR tree-optimization/83640
6705 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6706 subtracting negative offset from size.
6707 (builtin_access::overlap): Adjust offset bounds of the access to fall
6708 within the size of the object if possible.
6710 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6712 PR rtl-optimization/83699
6713 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6714 extract_bit_field_as_subreg to cases in which the extracted
6715 value is also a vector.
6717 * lra-constraints.c (process_alt_operands): Test for the equivalence
6718 substitutions when detecting a possible reload cycle.
6720 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6723 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6724 by default if flag_selective_schedling{,2}. Formatting fixes.
6726 PR rtl-optimization/83682
6727 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6728 if it has non-VECTOR_MODE element mode.
6729 (vec_duplicate_p): Likewise.
6732 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6733 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6735 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6738 * config/i386/i386-builtin.def
6739 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6740 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6741 Require also OPTION_MASK_ISA_AVX512F in addition to
6742 OPTION_MASK_ISA_GFNI.
6743 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6744 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6745 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6746 to OPTION_MASK_ISA_GFNI.
6747 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6748 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6749 OPTION_MASK_ISA_AVX512BW.
6750 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6751 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6752 addition to OPTION_MASK_ISA_GFNI.
6753 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6754 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6755 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6756 to OPTION_MASK_ISA_GFNI.
6757 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6758 a requirement for all ISAs rather than any of them with a few
6760 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6762 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6763 bitmasks to be enabled with 3 exceptions, instead of requiring any
6764 enabled ISA with lots of exceptions.
6765 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6766 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6767 Change avx512bw in isa attribute to avx512f.
6768 * config/i386/sgxintrin.h: Add license boilerplate.
6769 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6770 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6771 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6772 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6774 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6775 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6776 temporarily sse2 rather than sse if not enabled already.
6779 * config/i386/sse.md (VI248_VLBW): Rename to ...
6780 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6781 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6782 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6783 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6784 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6785 mode iterator instead of VI248_VLBW.
6787 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6789 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6790 (record_modified): Skip clobbers; add debug output.
6791 (param_change_prob): Use sreal frequencies.
6793 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6795 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6796 punt for user-aligned variables.
6798 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6800 * tree-chrec.c (chrec_contains_symbols): Return true for
6803 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6806 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6807 of (x|y) == x for BICS pattern.
6809 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6811 PR tree-optimization/83605
6812 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6813 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6816 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6818 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6819 * config/epiphany/rtems.h: New file.
6821 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6822 Uros Bizjak <ubizjak@gmail.com>
6825 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6826 QIreg_operand instead of register_operand predicate.
6827 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6828 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6829 comments instead of -fmitigate[-_]rop.
6831 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6834 * cgraphunit.c (symbol_table::compile): Switch to text_section
6835 before calling assembly_start debug hook.
6836 * run-rtl-passes.c (run_rtl_passes): Likewise.
6839 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6841 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6842 range_int_cst_p rather than !symbolic_range_p before calling
6843 extract_range_from_multiplicative_op_1.
6845 2018-01-04 Jeff Law <law@redhat.com>
6847 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6848 redundant test in assertion.
6850 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6852 * doc/rtl.texi: Document machine_mode wrapper classes.
6854 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6856 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6859 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6861 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6862 the VEC_PERM_EXPR fold to fail.
6864 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6867 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6868 to switched_sections.
6870 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6873 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6876 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6879 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6880 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6882 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6885 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6886 is BLKmode and bitpos not zero or mode change is needed.
6888 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6891 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6894 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6897 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6898 instead of MULT rtx. Update all corresponding splitters.
6900 (*ssub<modesuffix>): Ditto.
6902 (*cmp_sadd_di): Update split patterns.
6903 (*cmp_sadd_si): Ditto.
6904 (*cmp_sadd_sidi): Ditto.
6905 (*cmp_ssub_di): Ditto.
6906 (*cmp_ssub_si): Ditto.
6907 (*cmp_ssub_sidi): Ditto.
6908 * config/alpha/predicates.md (const23_operand): New predicate.
6909 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6910 Look for ASHIFT, not MULT inner operand.
6911 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6913 2018-01-04 Martin Liska <mliska@suse.cz>
6915 PR gcov-profile/83669
6916 * gcov.c (output_intermediate_file): Add version to intermediate
6918 * doc/gcov.texi: Document new field 'version' in intermediate
6919 file format. Fix location of '-k' option of gcov command.
6921 2018-01-04 Martin Liska <mliska@suse.cz>
6924 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6926 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6928 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6930 2018-01-03 Martin Sebor <msebor@redhat.com>
6932 PR tree-optimization/83655
6933 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6934 checking calls with invalid arguments.
6936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6938 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6939 (vectorizable_mask_load_store): Delete.
6940 (vectorizable_call): Return false for masked loads and stores.
6941 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6942 instead of gimple_assign_rhs1.
6943 (vectorizable_load): Handle IFN_MASK_LOAD.
6944 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6948 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6950 (vectorizable_mask_load_store): ...here.
6951 (vectorizable_load): ...and here.
6953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6955 * tree-vect-stmts.c (vect_build_all_ones_mask)
6956 (vect_build_zero_merge_argument): New functions, split out from...
6957 (vectorizable_load): ...here.
6959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6961 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6963 (vectorizable_mask_load_store): ...here.
6964 (vectorizable_store): ...and here.
6966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6968 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6970 (vectorizable_mask_load_store): ...here.
6972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6974 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6975 (vect_model_store_cost): Take a vec_load_store_type instead of a
6977 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6978 (vect_model_store_cost): Take a vec_load_store_type instead of a
6980 (vectorizable_mask_load_store): Update accordingly.
6981 (vectorizable_store): Likewise.
6982 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6986 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6987 IFN_MASK_LOAD calls here rather than...
6988 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6991 Alan Hayward <alan.hayward@arm.com>
6992 David Sherwood <david.sherwood@arm.com>
6994 * expmed.c (extract_bit_field_1): For vector extracts,
6995 fall back to extract_bit_field_as_subreg if vec_extract
6998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6999 Alan Hayward <alan.hayward@arm.com>
7000 David Sherwood <david.sherwood@arm.com>
7002 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7003 they are variable or constant sized.
7004 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7005 slots for constant-sized data.
7007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7008 Alan Hayward <alan.hayward@arm.com>
7009 David Sherwood <david.sherwood@arm.com>
7011 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7012 handling COND_EXPRs with boolean comparisons, try to find a better
7013 basis for the mask type than the boolean itself.
7015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7017 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7018 is calculated and how it can be overridden.
7019 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7020 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7022 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7026 Alan Hayward <alan.hayward@arm.com>
7027 David Sherwood <david.sherwood@arm.com>
7029 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7030 Remove the mode argument.
7031 (aarch64_simd_valid_immediate): Remove the mode and inverse
7033 * config/aarch64/iterators.md (bitsize): New iterator.
7034 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7035 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7036 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7037 aarch64_simd_valid_immediate.
7038 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7039 (aarch64_reg_or_bic_imm): Likewise.
7040 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7041 with an insn_type enum and msl with a modifier_type enum.
7042 Replace element_width with a scalar_mode. Change the shift
7043 to unsigned int. Add constructors for scalar_float_mode and
7044 scalar_int_mode elements.
7045 (aarch64_vect_float_const_representable_p): Delete.
7046 (aarch64_can_const_movi_rtx_p)
7047 (aarch64_simd_scalar_immediate_valid_for_move)
7048 (aarch64_simd_make_constant): Update call to
7049 aarch64_simd_valid_immediate.
7050 (aarch64_advsimd_valid_immediate_hs): New function.
7051 (aarch64_advsimd_valid_immediate): Likewise.
7052 (aarch64_simd_valid_immediate): Remove mode and inverse
7053 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7054 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7055 and aarch64_float_const_representable_p on the result.
7056 (aarch64_output_simd_mov_immediate): Remove mode argument.
7057 Update call to aarch64_simd_valid_immediate and use of
7058 simd_immediate_info.
7059 (aarch64_output_scalar_simd_mov_immediate): Update call
7062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7063 Alan Hayward <alan.hayward@arm.com>
7064 David Sherwood <david.sherwood@arm.com>
7066 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7067 (mode_nunits): Likewise CONST_MODE_NUNITS.
7068 * machmode.def (ADJUST_NUNITS): Document.
7069 * genmodes.c (mode_data::need_nunits_adj): New field.
7070 (blank_mode): Update accordingly.
7071 (adj_nunits): New variable.
7072 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7074 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7075 listed in adj_nunits.
7076 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7077 listed in adj_nunits. Don't emit case statements for such modes.
7078 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7079 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7080 nothing if adj_nunits is nonnull.
7081 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7082 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7083 (emit_mode_fbit): Update use of print_maybe_const_decl.
7084 (emit_move_size): Likewise. Treat the array as non-const
7086 (emit_mode_adjustments): Handle adj_nunits.
7088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7090 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7091 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7092 (VECTOR_MODES): Use it.
7093 (make_vector_modes): Take the prefix as an argument.
7095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7096 Alan Hayward <alan.hayward@arm.com>
7097 David Sherwood <david.sherwood@arm.com>
7099 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7100 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7101 for MODE_VECTOR_BOOL.
7102 * machmode.def (VECTOR_BOOL_MODE): Document.
7103 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7104 (make_vector_bool_mode): New function.
7105 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7107 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7108 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7110 * stor-layout.c (int_mode_for_mode): Likewise.
7111 * tree.c (build_vector_type_for_mode): Likewise.
7112 * varasm.c (output_constant_pool_2): Likewise.
7113 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7114 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7115 for MODE_VECTOR_BOOL.
7116 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7117 of mode class checks.
7118 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7119 instead of a list of mode class checks.
7120 (expand_vector_scalar_condition): Likewise.
7121 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7124 Alan Hayward <alan.hayward@arm.com>
7125 David Sherwood <david.sherwood@arm.com>
7127 * machmode.h (mode_size): Change from unsigned short to
7129 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7130 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7131 or if measurement_type is not polynomial.
7132 (fixed_size_mode::includes_p): Check for constant-sized modes.
7133 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7134 return a poly_uint16 rather than an unsigned short.
7135 (emit_mode_size): Change the type of mode_size from unsigned short
7136 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7137 (emit_mode_adjustments): Cope with polynomial vector sizes.
7138 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7140 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7142 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7143 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7144 * caller-save.c (setup_save_areas): Likewise.
7145 (replace_reg_with_saved_mem): Likewise.
7146 * calls.c (emit_library_call_value_1): Likewise.
7147 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7148 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7149 (gen_lowpart_for_combine): Likewise.
7150 * convert.c (convert_to_integer_1): Likewise.
7151 * cse.c (equiv_constant, cse_insn): Likewise.
7152 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7153 (cselib_subst_to_values): Likewise.
7154 * dce.c (word_dce_process_block): Likewise.
7155 * df-problems.c (df_word_lr_mark_ref): Likewise.
7156 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7157 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7158 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7159 (rtl_for_decl_location): Likewise.
7160 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7161 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7162 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7163 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7164 (expand_expr_real_1): Likewise.
7165 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7166 (pad_below): Likewise.
7167 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7168 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7169 * ira.c (get_subreg_tracking_sizes): Likewise.
7170 * ira-build.c (ira_create_allocno_objects): Likewise.
7171 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7172 (ira_sort_regnos_for_alter_reg): Likewise.
7173 * ira-costs.c (record_operand_costs): Likewise.
7174 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7175 (resolve_simple_move): Likewise.
7176 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7177 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7178 (lra_constraints): Likewise.
7179 (CONST_POOL_OK_P): Reject variable-sized modes.
7180 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7181 (add_pseudo_to_slot, lra_spill): Likewise.
7182 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7183 * optabs-query.c (get_best_extraction_insn): Likewise.
7184 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7185 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7186 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7187 * recog.c (offsettable_address_addr_space_p): Likewise.
7188 * regcprop.c (maybe_mode_change): Likewise.
7189 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7190 * regrename.c (build_def_use): Likewise.
7191 * regstat.c (dump_reg_info): Likewise.
7192 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7193 (find_reloads, find_reloads_subreg_address): Likewise.
7194 * reload1.c (eliminate_regs_1): Likewise.
7195 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7196 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7197 (simplify_binary_operation_1, simplify_subreg): Likewise.
7198 * targhooks.c (default_function_arg_padding): Likewise.
7199 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7200 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7201 (verify_gimple_assign_ternary): Likewise.
7202 * tree-inline.c (estimate_move_cost): Likewise.
7203 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7204 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7205 (get_address_cost_ainc): Likewise.
7206 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7207 (vect_supportable_dr_alignment): Likewise.
7208 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7209 (vectorizable_reduction): Likewise.
7210 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7211 (vectorizable_operation, vectorizable_load): Likewise.
7212 * tree.c (build_same_sized_truth_vector_type): Likewise.
7213 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7214 * var-tracking.c (emit_note_insn_var_location): Likewise.
7215 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7216 (ADDR_VEC_ALIGN): Likewise.
7218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7219 Alan Hayward <alan.hayward@arm.com>
7220 David Sherwood <david.sherwood@arm.com>
7222 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7224 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7225 or if measurement_type is polynomial.
7226 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7227 * combine.c (make_extraction): Likewise.
7228 * dse.c (find_shift_sequence): Likewise.
7229 * dwarf2out.c (mem_loc_descriptor): Likewise.
7230 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7231 (extract_bit_field, extract_low_bits): Likewise.
7232 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7233 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7234 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7235 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7236 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7237 * reload.c (find_reloads): Likewise.
7238 * reload1.c (alter_reg): Likewise.
7239 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7240 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7241 * tree-if-conv.c (predicate_mem_writes): Likewise.
7242 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7243 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7244 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7245 * valtrack.c (dead_debug_insert_temp): Likewise.
7246 * varasm.c (mergeable_constant_section): Likewise.
7247 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7250 Alan Hayward <alan.hayward@arm.com>
7251 David Sherwood <david.sherwood@arm.com>
7253 * expr.c (expand_assignment): Cope with polynomial mode sizes
7254 when assigning to a CONCAT.
7256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7257 Alan Hayward <alan.hayward@arm.com>
7258 David Sherwood <david.sherwood@arm.com>
7260 * machmode.h (mode_precision): Change from unsigned short to
7262 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7264 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7265 or if measurement_type is not polynomial.
7266 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7267 in which the mode is already known to be a scalar_int_mode.
7268 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7269 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7271 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7272 for GET_MODE_PRECISION.
7273 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7274 for GET_MODE_PRECISION.
7275 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7277 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7278 (expand_field_assignment, make_extraction): Likewise.
7279 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7280 (get_last_value): Likewise.
7281 * convert.c (convert_to_integer_1): Likewise.
7282 * cse.c (cse_insn): Likewise.
7283 * expr.c (expand_expr_real_1): Likewise.
7284 * lra-constraints.c (simplify_operand_subreg): Likewise.
7285 * optabs-query.c (can_atomic_load_p): Likewise.
7286 * optabs.c (expand_atomic_load): Likewise.
7287 (expand_atomic_store): Likewise.
7288 * ree.c (combine_reaching_defs): Likewise.
7289 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7290 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7291 * tree.h (type_has_mode_precision_p): Likewise.
7292 * ubsan.c (instrument_si_overflow): Likewise.
7294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7295 Alan Hayward <alan.hayward@arm.com>
7296 David Sherwood <david.sherwood@arm.com>
7298 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7299 polynomial numbers of units.
7300 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7301 (valid_vector_subparts_p): New function.
7302 (build_vector_type): Remove temporary shim and take the number
7303 of units as a poly_uint64 rather than an int.
7304 (build_opaque_vector_type): Take the number of units as a
7305 poly_uint64 rather than an int.
7306 * tree.c (build_vector_from_ctor): Handle polynomial
7307 TYPE_VECTOR_SUBPARTS.
7308 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7309 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7310 (build_vector_from_val): If the number of units is variable,
7311 use build_vec_duplicate_cst for constant operands and
7312 VEC_DUPLICATE_EXPR otherwise.
7313 (make_vector_type): Remove temporary is_constant ().
7314 (build_vector_type, build_opaque_vector_type): Take the number of
7315 units as a poly_uint64 rather than an int.
7316 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7318 * cfgexpand.c (expand_debug_expr): Likewise.
7319 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7320 (store_constructor, expand_expr_real_1): Likewise.
7321 (const_scalar_mask_from_tree): Likewise.
7322 * fold-const-call.c (fold_const_reduction): Likewise.
7323 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7324 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7325 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7326 (fold_relational_const): Likewise.
7327 (native_interpret_vector): Likewise. Change the size from an
7328 int to an unsigned int.
7329 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7330 TYPE_VECTOR_SUBPARTS.
7331 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7332 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7333 duplicating a non-constant operand into a variable-length vector.
7334 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7335 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7336 * ipa-icf.c (sem_variable::equals): Likewise.
7337 * match.pd: Likewise.
7338 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7339 * print-tree.c (print_node): Likewise.
7340 * stor-layout.c (layout_type): Likewise.
7341 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7342 * tree-cfg.c (verify_gimple_comparison): Likewise.
7343 (verify_gimple_assign_binary): Likewise.
7344 (verify_gimple_assign_ternary): Likewise.
7345 (verify_gimple_assign_single): Likewise.
7346 * tree-pretty-print.c (dump_generic_node): Likewise.
7347 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7348 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7349 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7350 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7351 (vect_shift_permute_load_chain): Likewise.
7352 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7353 (expand_vector_condition, optimize_vector_constructor): Likewise.
7354 (lower_vec_perm, get_compute_type): Likewise.
7355 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7356 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7357 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7358 (vect_recog_mask_conversion_pattern): Likewise.
7359 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7360 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7361 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7362 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7363 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7364 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7365 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7366 (supportable_widening_operation): Likewise.
7367 (supportable_narrowing_operation): Likewise.
7368 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7370 * varasm.c (output_constant): Likewise.
7372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7373 Alan Hayward <alan.hayward@arm.com>
7374 David Sherwood <david.sherwood@arm.com>
7376 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7377 so that both the length == 3 and length != 3 cases set up their
7378 own permute vectors. Add comments explaining why we know the
7379 number of elements is constant.
7380 (vect_permute_load_chain): Likewise.
7382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7383 Alan Hayward <alan.hayward@arm.com>
7384 David Sherwood <david.sherwood@arm.com>
7386 * machmode.h (mode_nunits): Change from unsigned char to
7388 (ONLY_FIXED_SIZE_MODES): New macro.
7389 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7390 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7391 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7393 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7394 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7395 or if measurement_type is not polynomial.
7396 * genmodes.c (ZERO_COEFFS): New macro.
7397 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7399 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7400 Use ZERO_COEFFS when emitting initializers.
7401 * data-streamer.h (bp_pack_poly_value): New function.
7402 (bp_unpack_poly_value): Likewise.
7403 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7404 for GET_MODE_NUNITS.
7405 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7406 for GET_MODE_NUNITS.
7407 * tree.c (make_vector_type): Remove temporary shim and make
7408 the real function take the number of units as a poly_uint64
7410 (build_vector_type_for_mode): Handle polynomial nunits.
7411 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7412 * emit-rtl.c (const_vec_series_p_1): Likewise.
7413 (gen_rtx_CONST_VECTOR): Likewise.
7414 * fold-const.c (test_vec_duplicate_folding): Likewise.
7415 * genrecog.c (validate_pattern): Likewise.
7416 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7417 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7418 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7419 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7420 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7421 * rtlanal.c (subreg_get_info): Likewise.
7422 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7423 (vect_grouped_load_supported): Likewise.
7424 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7425 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7426 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7427 (simplify_const_unary_operation, simplify_binary_operation_1)
7428 (simplify_const_binary_operation, simplify_ternary_operation)
7429 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7430 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7431 instead of CONST_VECTOR_NUNITS.
7432 * varasm.c (output_constant_pool_2): Likewise.
7433 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7434 explicit-encoded elements in the XVEC for variable-length vectors.
7436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7438 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7441 Alan Hayward <alan.hayward@arm.com>
7442 David Sherwood <david.sherwood@arm.com>
7444 * coretypes.h (fixed_size_mode): Declare.
7445 (fixed_size_mode_pod): New typedef.
7446 * builtins.h (target_builtins::x_apply_args_mode)
7447 (target_builtins::x_apply_result_mode): Change type to
7448 fixed_size_mode_pod.
7449 * builtins.c (apply_args_size, apply_result_size, result_vector)
7450 (expand_builtin_apply_args_1, expand_builtin_apply)
7451 (expand_builtin_return): Update accordingly.
7453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7455 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7456 * cselib.c (cselib_hash_rtx): Likewise.
7457 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7458 CONST_VECTOR encoding.
7460 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7461 Jeff Law <law@redhat.com>
7464 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7465 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7466 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7467 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7470 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7471 explicitly probe *sp in a noreturn function if there were any callee
7472 register saves or frame pointer is needed.
7474 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7477 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7478 BLKmode for ternary, binary or unary expressions.
7481 * var-tracking.c (delete_vta_debug_insn): New inline function.
7482 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7483 insns from get_insns () to NULL instead of each bb separately.
7484 Use delete_vta_debug_insn. No longer static.
7485 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7486 delete_vta_debug_insns callers.
7487 * rtl.h (delete_vta_debug_insns): Declare.
7488 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7489 instead of variable_tracking_main.
7491 2018-01-03 Martin Sebor <msebor@redhat.com>
7493 PR tree-optimization/83603
7494 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7495 arguments past the endof the argument list in functions declared
7496 without a prototype.
7497 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7498 Avoid checking when arguments are null.
7500 2018-01-03 Martin Sebor <msebor@redhat.com>
7503 * doc/extend.texi (attribute const): Fix a typo.
7504 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7505 issuing -Wsuggest-attribute for void functions.
7507 2018-01-03 Martin Sebor <msebor@redhat.com>
7509 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7510 offset_int::from instead of wide_int::to_shwi.
7511 (maybe_diag_overlap): Remove assertion.
7512 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7513 * gimple-ssa-sprintf.c (format_directive): Same.
7514 (parse_directive): Same.
7515 (sprintf_dom_walker::compute_format_length): Same.
7516 (try_substitute_return_value): Same.
7518 2018-01-03 Jeff Law <law@redhat.com>
7521 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7522 non-constant residual for zero at runtime and avoid probing in
7523 that case. Reorganize code for trailing problem to mirror handling
7526 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7528 PR tree-optimization/83501
7529 * tree-ssa-strlen.c (get_string_cst): New.
7530 (handle_char_store): Call get_string_cst.
7532 2018-01-03 Martin Liska <mliska@suse.cz>
7534 PR tree-optimization/83593
7535 * tree-ssa-strlen.c: Include tree-cfg.h.
7536 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7537 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7538 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7540 (strlen_dom_walker::before_dom_children): Call
7541 gimple_purge_dead_eh_edges. Dump tranformation with details
7543 (strlen_dom_walker::before_dom_children): Update call by adding
7544 new argument cleanup_eh.
7545 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7547 2018-01-03 Martin Liska <mliska@suse.cz>
7550 * cif-code.def (VARIADIC_THUNK): New enum value.
7551 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7554 2018-01-03 Jan Beulich <jbeulich@suse.com>
7556 * sse.md (mov<mode>_internal): Tighten condition for when to use
7557 vmovdqu<ssescalarsize> for TI and OI modes.
7559 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7561 Update copyright years.
7563 2018-01-03 Martin Liska <mliska@suse.cz>
7566 * ipa-visibility.c (function_and_variable_visibility): Skip
7567 functions with noipa attribure.
7569 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7571 * gcc.c (process_command): Update copyright notice dates.
7572 * gcov-dump.c (print_version): Ditto.
7573 * gcov.c (print_version): Ditto.
7574 * gcov-tool.c (print_version): Ditto.
7575 * gengtype.c (create_file): Ditto.
7576 * doc/cpp.texi: Bump @copying's copyright year.
7577 * doc/cppinternals.texi: Ditto.
7578 * doc/gcc.texi: Ditto.
7579 * doc/gccint.texi: Ditto.
7580 * doc/gcov.texi: Ditto.
7581 * doc/install.texi: Ditto.
7582 * doc/invoke.texi: Ditto.
7584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7586 * vector-builder.h (vector_builder::m_full_nelts): Change from
7587 unsigned int to poly_uint64.
7588 (vector_builder::full_nelts): Update prototype accordingly.
7589 (vector_builder::new_vector): Likewise.
7590 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7591 (vector_builder::operator ==): Likewise.
7592 (vector_builder::finalize): Likewise.
7593 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7594 Take the number of elements as a poly_uint64 rather than an
7596 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7597 from unsigned int to poly_uint64.
7598 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7599 (vec_perm_indices::new_vector): Likewise.
7600 (vec_perm_indices::length): Likewise.
7601 (vec_perm_indices::nelts_per_input): Likewise.
7602 (vec_perm_indices::input_nelts): Likewise.
7603 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7604 number of elements per input as a poly_uint64 rather than an
7605 unsigned int. Use the original encoding for variable-length
7606 vectors, rather than clamping each individual element.
7607 For the second and subsequent elements in each pattern,
7608 clamp the step and base before clamping their sum.
7609 (vec_perm_indices::series_p): Handle polynomial element counts.
7610 (vec_perm_indices::all_in_range_p): Likewise.
7611 (vec_perm_indices_to_tree): Likewise.
7612 (vec_perm_indices_to_rtx): Likewise.
7613 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7614 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7615 (tree_vector_builder::new_binary_operation): Handle polynomial
7616 element counts. Return false if we need to know the number
7617 of elements at compile time.
7618 * fold-const.c (fold_vec_perm): Punt if the number of elements
7619 isn't known at compile time.
7621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7623 * vec-perm-indices.h (vec_perm_builder): Change element type
7624 from HOST_WIDE_INT to poly_int64.
7625 (vec_perm_indices::element_type): Update accordingly.
7626 (vec_perm_indices::clamp): Handle polynomial element_types.
7627 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7628 (vec_perm_indices::all_in_range_p): Likewise.
7629 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7631 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7632 polynomial vec_perm_indices element types.
7633 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7634 * fold-const.c (fold_vec_perm): Likewise.
7635 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7636 * tree-vect-generic.c (lower_vec_perm): Likewise.
7637 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7638 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7639 element type to HOST_WIDE_INT.
7641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7642 Alan Hayward <alan.hayward@arm.com>
7643 David Sherwood <david.sherwood@arm.com>
7645 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7646 rather than an int. Use plus_constant.
7647 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7648 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7651 Alan Hayward <alan.hayward@arm.com>
7652 David Sherwood <david.sherwood@arm.com>
7654 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7655 a HOST_WIDE_INT to a poly_int64.
7657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7658 Alan Hayward <alan.hayward@arm.com>
7659 David Sherwood <david.sherwood@arm.com>
7661 * calls.c (load_register_parameters): Cope with polynomial
7662 mode sizes. Require a constant size for BLKmode parameters
7663 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7664 forces a parameter to be padded at the lsb end in order to
7665 fill a complete number of words, require the parameter size
7666 to be ordered wrt UNITS_PER_WORD.
7668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7669 Alan Hayward <alan.hayward@arm.com>
7670 David Sherwood <david.sherwood@arm.com>
7672 * reload1.c (spill_stack_slot_width): Change element type
7673 from unsigned int to poly_uint64_pod.
7674 (alter_reg): Treat mode sizes as polynomial.
7676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7677 Alan Hayward <alan.hayward@arm.com>
7678 David Sherwood <david.sherwood@arm.com>
7680 * reload.c (complex_word_subreg_p): New function.
7681 (reload_inner_reg_of_subreg, push_reload): Use it.
7683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7684 Alan Hayward <alan.hayward@arm.com>
7685 David Sherwood <david.sherwood@arm.com>
7687 * lra-constraints.c (process_alt_operands): Reject matched
7688 operands whose sizes aren't ordered.
7689 (match_reload): Refer to this check here.
7691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7692 Alan Hayward <alan.hayward@arm.com>
7693 David Sherwood <david.sherwood@arm.com>
7695 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7696 that the mode size is in the set {1, 2, 4, 8, 16}.
7698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7699 Alan Hayward <alan.hayward@arm.com>
7700 David Sherwood <david.sherwood@arm.com>
7702 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7703 Use plus_constant instead of gen_rtx_PLUS.
7705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7706 Alan Hayward <alan.hayward@arm.com>
7707 David Sherwood <david.sherwood@arm.com>
7709 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7710 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7711 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7712 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7713 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7714 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7715 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7716 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7717 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7718 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7720 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7721 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7722 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7723 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7724 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7725 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7726 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7727 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7728 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7729 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7731 * expr.c (emit_move_resolve_push): Treat the input and result
7732 of PUSH_ROUNDING as a poly_int64.
7733 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7734 (emit_push_insn): Likewise.
7735 * lra-eliminations.c (mark_not_eliminable): Likewise.
7736 * recog.c (push_operand): Likewise.
7737 * reload1.c (elimination_effects): Likewise.
7738 * rtlanal.c (nonzero_bits1): Likewise.
7739 * calls.c (store_one_arg): Likewise. Require the padding to be
7740 known at compile time.
7742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7743 Alan Hayward <alan.hayward@arm.com>
7744 David Sherwood <david.sherwood@arm.com>
7746 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7747 Use plus_constant instead of gen_rtx_PLUS.
7749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7750 Alan Hayward <alan.hayward@arm.com>
7751 David Sherwood <david.sherwood@arm.com>
7753 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7757 Alan Hayward <alan.hayward@arm.com>
7758 David Sherwood <david.sherwood@arm.com>
7760 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7761 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7762 via stack temporaries. Treat the mode size as polynomial too.
7764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7765 Alan Hayward <alan.hayward@arm.com>
7766 David Sherwood <david.sherwood@arm.com>
7768 * expr.c (expand_expr_real_2): When handling conversions involving
7769 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7770 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7771 as a poly_uint64 too.
7773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7774 Alan Hayward <alan.hayward@arm.com>
7775 David Sherwood <david.sherwood@arm.com>
7777 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7780 Alan Hayward <alan.hayward@arm.com>
7781 David Sherwood <david.sherwood@arm.com>
7783 * combine.c (can_change_dest_mode): Handle polynomial
7784 REGMODE_NATURAL_SIZE.
7785 * expmed.c (store_bit_field_1): Likewise.
7786 * expr.c (store_constructor): Likewise.
7787 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7788 and polynomial REGMODE_NATURAL_SIZE.
7789 (gen_lowpart_common): Likewise.
7790 * reginfo.c (record_subregs_of_mode): Likewise.
7791 * rtlanal.c (read_modify_subreg_p): Likewise.
7793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7794 Alan Hayward <alan.hayward@arm.com>
7795 David Sherwood <david.sherwood@arm.com>
7797 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7798 numbers of elements.
7800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7801 Alan Hayward <alan.hayward@arm.com>
7802 David Sherwood <david.sherwood@arm.com>
7804 * match.pd: Cope with polynomial numbers of vector elements.
7806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7807 Alan Hayward <alan.hayward@arm.com>
7808 David Sherwood <david.sherwood@arm.com>
7810 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7811 in a POINTER_PLUS_EXPR.
7813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7814 Alan Hayward <alan.hayward@arm.com>
7815 David Sherwood <david.sherwood@arm.com>
7817 * omp-simd-clone.c (simd_clone_subparts): New function.
7818 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7819 (ipa_simd_modify_function_body): Likewise.
7821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7822 Alan Hayward <alan.hayward@arm.com>
7823 David Sherwood <david.sherwood@arm.com>
7825 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7826 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7827 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7828 (expand_vector_condition, vector_element): Likewise.
7829 (subparts_gt): New function.
7830 (get_compute_type): Use subparts_gt.
7831 (count_type_subparts): Delete.
7832 (expand_vector_operations_1): Use subparts_gt instead of
7833 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7840 (vect_compile_time_alias): ...this new function. Do the calculation
7841 on poly_ints rather than trees.
7842 (vect_prune_runtime_alias_test_list): Update call accordingly.
7844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7845 Alan Hayward <alan.hayward@arm.com>
7846 David Sherwood <david.sherwood@arm.com>
7848 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7850 (vect_schedule_slp_instance): Likewise.
7852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7853 Alan Hayward <alan.hayward@arm.com>
7854 David Sherwood <david.sherwood@arm.com>
7856 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7857 constant and extern definitions for variable-length vectors.
7858 (vect_get_constant_vectors): Note that the number of units
7859 is known to be constant.
7861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7862 Alan Hayward <alan.hayward@arm.com>
7863 David Sherwood <david.sherwood@arm.com>
7865 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7866 of units as polynomial. Choose between WIDE and NARROW based
7869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7870 Alan Hayward <alan.hayward@arm.com>
7871 David Sherwood <david.sherwood@arm.com>
7873 * tree-vect-stmts.c (simd_clone_subparts): New function.
7874 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7877 Alan Hayward <alan.hayward@arm.com>
7878 David Sherwood <david.sherwood@arm.com>
7880 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7881 vectors as polynomial. Use build_index_vector for
7884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7885 Alan Hayward <alan.hayward@arm.com>
7886 David Sherwood <david.sherwood@arm.com>
7888 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7889 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7890 for variable-length vectors.
7891 (vectorizable_mask_load_store): Treat the number of units as
7892 polynomial, asserting that it is constant if the condition has
7893 already been enforced.
7894 (vectorizable_store, vectorizable_load): Likewise.
7896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7897 Alan Hayward <alan.hayward@arm.com>
7898 David Sherwood <david.sherwood@arm.com>
7900 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7901 of units as polynomial. Punt if we can't tell at compile time
7902 which vector contains the final result.
7904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7905 Alan Hayward <alan.hayward@arm.com>
7906 David Sherwood <david.sherwood@arm.com>
7908 * tree-vect-loop.c (vectorizable_induction): Treat the number
7909 of units as polynomial. Punt on SLP inductions. Use an integer
7910 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7911 cast of such a series for variable-length floating-point
7914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7915 Alan Hayward <alan.hayward@arm.com>
7916 David Sherwood <david.sherwood@arm.com>
7918 * tree.h (build_index_vector): Declare.
7919 * tree.c (build_index_vector): New function.
7920 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7921 of units as polynomial, forcibly converting it to a constant if
7922 vectorizable_reduction has already enforced the condition.
7923 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7924 to create a {1,2,3,...} vector.
7925 (vectorizable_reduction): Treat the number of units as polynomial.
7926 Choose vectype_in based on the largest scalar element size rather
7927 than the smallest number of units. Enforce the restrictions
7930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7931 Alan Hayward <alan.hayward@arm.com>
7932 David Sherwood <david.sherwood@arm.com>
7934 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7935 number of units as polynomial.
7937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7938 Alan Hayward <alan.hayward@arm.com>
7939 David Sherwood <david.sherwood@arm.com>
7941 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7942 * target.def (autovectorize_vector_sizes): Return the vector sizes
7943 by pointer, using vector_sizes rather than a bitmask.
7944 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7945 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7946 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7948 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7949 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7950 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7951 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7952 * omp-general.c (omp_max_vf): Likewise.
7953 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7954 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7955 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7956 * tree-vect-slp.c (vect_slp_bb): Likewise.
7957 * doc/tm.texi: Regenerate.
7958 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7960 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7961 the vector size as a poly_uint64 rather than an unsigned int.
7962 (current_vector_size): Change from an unsigned int to a poly_uint64.
7963 (get_vectype_for_scalar_type): Update accordingly.
7964 * tree.h (build_truth_vector_type): Take the size and number of
7965 units as a poly_uint64 rather than an unsigned int.
7966 (build_vector_type): Add a temporary overload that takes
7967 the number of units as a poly_uint64 rather than an unsigned int.
7968 * tree.c (make_vector_type): Likewise.
7969 (build_truth_vector_type): Take the number of units as a poly_uint64
7970 rather than an unsigned int.
7972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7973 Alan Hayward <alan.hayward@arm.com>
7974 David Sherwood <david.sherwood@arm.com>
7976 * target.def (get_mask_mode): Take the number of units and length
7977 as poly_uint64s rather than unsigned ints.
7978 * targhooks.h (default_get_mask_mode): Update accordingly.
7979 * targhooks.c (default_get_mask_mode): Likewise.
7980 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7981 * doc/tm.texi: Regenerate.
7983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7984 Alan Hayward <alan.hayward@arm.com>
7985 David Sherwood <david.sherwood@arm.com>
7987 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7988 * omp-general.c (omp_max_vf): Likewise.
7989 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7990 (expand_omp_simd): Handle polynomial safelen.
7991 * omp-low.c (omplow_simd_context): Add a default constructor.
7992 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7993 (lower_rec_simd_input_clauses): Update accordingly.
7994 (lower_rec_input_clauses): Likewise.
7996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7997 Alan Hayward <alan.hayward@arm.com>
7998 David Sherwood <david.sherwood@arm.com>
8000 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8001 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8002 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8003 (vect_analyze_slp_cost): Likewise.
8004 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8005 (vect_model_load_cost): Likewise.
8007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8008 Alan Hayward <alan.hayward@arm.com>
8009 David Sherwood <david.sherwood@arm.com>
8011 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8012 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8013 from an unsigned int * to a poly_uint64_pod *.
8014 (calculate_unrolling_factor): New function.
8015 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8018 Alan Hayward <alan.hayward@arm.com>
8019 David Sherwood <david.sherwood@arm.com>
8021 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8022 from an unsigned int to a poly_uint64.
8023 (_loop_vec_info::slp_unrolling_factor): Likewise.
8024 (_loop_vec_info::vectorization_factor): Change from an int
8026 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8027 (vect_get_num_vectors): New function.
8028 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8029 (vect_get_num_copies): Use vect_get_num_vectors.
8030 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8031 to an unsigned int *.
8032 (vect_analyze_data_refs): Change min_vf from an int * to a
8034 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8035 than an unsigned HOST_WIDE_INT.
8036 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8037 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8038 to an unsigned int *.
8039 (vect_analyze_data_ref_dependences): Likewise.
8040 (vect_compute_data_ref_alignment): Handle polynomial vf.
8041 (vect_enhance_data_refs_alignment): Likewise.
8042 (vect_prune_runtime_alias_test_list): Likewise.
8043 (vect_shift_permute_load_chain): Likewise.
8044 (vect_supportable_dr_alignment): Likewise.
8045 (dependence_distance_ge_vf): Take the vectorization factor as a
8046 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8047 (vect_analyze_data_refs): Change min_vf from an int * to a
8049 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8050 vfm1 as a poly_uint64 rather than an int. Make the same change
8051 for the returned bound_scalar.
8052 (vect_gen_vector_loop_niters): Handle polynomial vf.
8053 (vect_do_peeling): Likewise. Update call to
8054 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8055 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8057 * tree-vect-loop.c (vect_determine_vectorization_factor)
8058 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8059 (vect_get_known_peeling_cost): Likewise.
8060 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8061 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8062 (vect_transform_loop): Likewise. Use the lowest possible VF when
8063 updating the upper bounds of the loop.
8064 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8066 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8067 polynomial unroll factors.
8068 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8069 (vect_make_slp_decision): Likewise.
8070 (vect_supported_load_permutation_p): Likewise, and polynomial
8072 (vect_analyze_slp_cost): Handle polynomial vf.
8073 (vect_slp_analyze_node_operations): Likewise.
8074 (vect_slp_analyze_bb_1): Likewise.
8075 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8076 than an unsigned HOST_WIDE_INT.
8077 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8078 (vectorizable_load): Handle polynomial vf.
8079 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8081 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8084 Alan Hayward <alan.hayward@arm.com>
8085 David Sherwood <david.sherwood@arm.com>
8087 * match.pd: Handle bit operations involving three constants
8088 and try to fold one pair.
8090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8092 * tree-vect-loop-manip.c: Include gimple-fold.h.
8093 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8094 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8095 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8096 Add a path that uses a step of VF instead of 1, but disable it
8098 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8099 and niters_no_overflow parameters. Update calls to
8100 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8101 Create a new SSA name if the latter choses to use a ste other
8102 than zero, and return it via niters_vector_mult_vf_var.
8103 * tree-vect-loop.c (vect_transform_loop): Update calls to
8104 vect_do_peeling, vect_gen_vector_loop_niters and
8105 slpeel_make_loop_iterate_ntimes.
8106 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8107 (vect_gen_vector_loop_niters): Update declarations after above changes.
8109 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8111 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8112 128-bit round to integer instructions.
8113 (ceil<mode>2): Likewise.
8114 (btrunc<mode>2): Likewise.
8115 (round<mode>2): Likewise.
8117 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8119 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8120 unaligned VSX load/store on P8/P9.
8121 (expand_block_clear): Allow the use of unaligned VSX
8122 load/store on P8/P9.
8124 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8126 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8128 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8129 swap associated with both a load and a store.
8131 2018-01-02 Andrew Waterman <andrew@sifive.com>
8133 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8134 * config/riscv/riscv.md (clear_cache): Use it.
8136 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8138 * web.c: Remove out-of-date comment.
8140 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8142 * expr.c (fixup_args_size_notes): Check that any existing
8143 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8144 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8145 (emit_single_push_insn): ...here.
8147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8149 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8150 (const_vector_encoded_nelts): New function.
8151 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8152 (const_vector_int_elt, const_vector_elt): Declare.
8153 * emit-rtl.c (const_vector_int_elt_1): New function.
8154 (const_vector_elt): Likewise.
8155 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8156 of CONST_VECTOR_ELT.
8158 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8160 * expr.c: Include rtx-vector-builder.h.
8161 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8162 directly on the tree encoding.
8163 (const_vector_from_tree): Likewise.
8164 * optabs.c: Include rtx-vector-builder.h.
8165 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8166 sequence of "u" values.
8167 * vec-perm-indices.c: Include rtx-vector-builder.h.
8168 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8169 directly on the vec_perm_indices encoding.
8171 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8173 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8174 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8175 * rtx-vector-builder.h: New file.
8176 * rtx-vector-builder.c: Likewise.
8177 * rtl.h (rtx_def::u2): Add a const_vector field.
8178 (CONST_VECTOR_NPATTERNS): New macro.
8179 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8180 (CONST_VECTOR_DUPLICATE_P): Likewise.
8181 (CONST_VECTOR_STEPPED_P): Likewise.
8182 (CONST_VECTOR_ENCODED_ELT): Likewise.
8183 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8184 (unwrap_const_vec_duplicate): Likewise.
8185 (const_vec_series_p): Check for a non-duplicated vector encoding.
8186 Say that the function only returns true for integer vectors.
8187 * emit-rtl.c: Include rtx-vector-builder.h.
8188 (gen_const_vec_duplicate_1): Delete.
8189 (gen_const_vector): Call gen_const_vec_duplicate instead of
8190 gen_const_vec_duplicate_1.
8191 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8192 (gen_const_vec_duplicate): Use rtx_vector_builder.
8193 (gen_const_vec_series): Likewise.
8194 (gen_rtx_CONST_VECTOR): Likewise.
8195 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8196 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8197 Build a new vector rather than modifying a CONST_VECTOR in-place.
8198 (handle_special_swappables): Update call accordingly.
8199 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8200 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8201 Build a new vector rather than modifying a CONST_VECTOR in-place.
8202 (handle_special_swappables): Update call accordingly.
8204 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8206 * simplify-rtx.c (simplify_const_binary_operation): Use
8207 CONST_VECTOR_ELT instead of XVECEXP.
8209 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8211 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8212 the selector elements to be different from the data elements
8213 if the selector is a VECTOR_CST.
8214 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8215 ssizetype for the selector.
8217 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8219 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8220 before testing each element individually.
8221 * tree-vect-generic.c (lower_vec_perm): Likewise.
8223 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8225 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8226 * selftest-run-tests.c (selftest::run_tests): Call it.
8227 * vector-builder.h (vector_builder::operator ==): New function.
8228 (vector_builder::operator !=): Likewise.
8229 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8230 (vec_perm_indices::all_from_input_p): New function.
8231 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8232 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8233 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8234 instead of reading the VECTOR_CST directly. Detect whether both
8235 vector inputs are the same before constructing the vec_perm_indices,
8236 and update the number of inputs argument accordingly. Use the
8237 utility functions added above. Only construct sel2 if we need to.
8239 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8241 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8242 the broadcast of the low byte.
8243 (expand_mult_highpart): Use an explicit encoding for the permutes.
8244 * optabs-query.c (can_mult_highpart_p): Likewise.
8245 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8246 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8247 (vectorizable_bswap): Likewise.
8248 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8249 explicit encoding for the power-of-2 permutes.
8250 (vect_permute_store_chain): Likewise.
8251 (vect_grouped_load_supported): Likewise.
8252 (vect_permute_load_chain): Likewise.
8254 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8256 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8257 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8258 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8259 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8260 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8261 (vect_gen_perm_mask_any): Likewise.
8263 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8265 * int-vector-builder.h: New file.
8266 * vec-perm-indices.h: Include int-vector-builder.h.
8267 (vec_perm_indices): Redefine as an int_vector_builder.
8268 (auto_vec_perm_indices): Delete.
8269 (vec_perm_builder): Redefine as a stand-alone class.
8270 (vec_perm_indices::vec_perm_indices): New function.
8271 (vec_perm_indices::clamp): Likewise.
8272 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8273 (vec_perm_indices::new_vector): New function.
8274 (vec_perm_indices::new_expanded_vector): Update for new
8275 vec_perm_indices class.
8276 (vec_perm_indices::rotate_inputs): New function.
8277 (vec_perm_indices::all_in_range_p): Operate directly on the
8278 encoded form, without computing elided elements.
8279 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8280 encoding. Update for new vec_perm_indices class.
8281 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8282 the given vec_perm_builder.
8283 (expand_vec_perm_var): Update vec_perm_builder constructor.
8284 (expand_mult_highpart): Use vec_perm_builder instead of
8285 auto_vec_perm_indices.
8286 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8287 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8288 or double series encoding as appropriate.
8289 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8290 vec_perm_indices instead of auto_vec_perm_indices.
8291 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8292 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8293 (vect_permute_store_chain): Likewise.
8294 (vect_grouped_load_supported): Likewise.
8295 (vect_permute_load_chain): Likewise.
8296 (vect_shift_permute_load_chain): Likewise.
8297 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8298 (vect_transform_slp_perm_load): Likewise.
8299 (vect_schedule_slp_instance): Likewise.
8300 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8301 (vectorizable_mask_load_store): Likewise.
8302 (vectorizable_bswap): Likewise.
8303 (vectorizable_store): Likewise.
8304 (vectorizable_load): Likewise.
8305 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8306 vec_perm_indices instead of auto_vec_perm_indices. Use
8307 tree_to_vec_perm_builder to read the vector from a tree.
8308 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8309 vec_perm_builder instead of a vec_perm_indices.
8310 (have_whole_vector_shift): Use vec_perm_builder and
8311 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8312 truncation to calc_vec_perm_mask_for_shift.
8313 (vect_create_epilog_for_reduction): Likewise.
8314 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8315 from auto_vec_perm_indices to vec_perm_indices.
8316 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8317 instead of changing individual elements.
8318 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8319 the vector in d.perm.
8320 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8321 from auto_vec_perm_indices to vec_perm_indices.
8322 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8323 instead of changing individual elements.
8324 (arm_vectorize_vec_perm_const): Use new_vector to install
8325 the vector in d.perm.
8326 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8327 Update vec_perm_builder constructor.
8328 (rs6000_expand_interleave): Likewise.
8329 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8330 (rs6000_expand_interleave): Likewise.
8332 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8334 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8335 to qimode could truncate the indices.
8336 * optabs.c (expand_vec_perm_var): Likewise.
8338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8340 * Makefile.in (OBJS): Add vec-perm-indices.o.
8341 * vec-perm-indices.h: New file.
8342 * vec-perm-indices.c: Likewise.
8343 * target.h (vec_perm_indices): Replace with a forward class
8345 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8346 * optabs.h: Include vec-perm-indices.h.
8347 (expand_vec_perm): Delete.
8348 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8349 (expand_vec_perm_const): Declare.
8350 * target.def (vec_perm_const_ok): Replace with...
8351 (vec_perm_const): ...this new hook.
8352 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8353 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8354 * doc/tm.texi: Regenerate.
8355 * optabs.def (vec_perm_const): Delete.
8356 * doc/md.texi (vec_perm_const): Likewise.
8357 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8358 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8359 expand_vec_perm for constant permutation vectors. Assert that
8360 the mode of variable permutation vectors is the integer equivalent
8361 of the mode that is being permuted.
8362 * optabs-query.h (selector_fits_mode_p): Declare.
8363 * optabs-query.c: Include vec-perm-indices.h.
8364 (selector_fits_mode_p): New function.
8365 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8366 is defined, instead of checking whether the vec_perm_const_optab
8367 exists. Use targetm.vectorize.vec_perm_const instead of
8368 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8369 fit in the vector mode before using a variable permute.
8370 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8371 vec_perm_indices instead of an rtx.
8372 (expand_vec_perm): Replace with...
8373 (expand_vec_perm_const): ...this new function. Take the selector
8374 as a vec_perm_indices rather than an rtx. Also take the mode of
8375 the selector. Update call to shift_amt_for_vec_perm_mask.
8376 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8377 Use vec_perm_indices::new_expanded_vector to expand the original
8378 selector into bytes. Check whether the indices fit in the vector
8379 mode before using a variable permute.
8380 (expand_vec_perm_var): Make global.
8381 (expand_mult_highpart): Use expand_vec_perm_const.
8382 * fold-const.c: Includes vec-perm-indices.h.
8383 * tree-ssa-forwprop.c: Likewise.
8384 * tree-vect-data-refs.c: Likewise.
8385 * tree-vect-generic.c: Likewise.
8386 * tree-vect-loop.c: Likewise.
8387 * tree-vect-slp.c: Likewise.
8388 * tree-vect-stmts.c: Likewise.
8389 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8391 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8392 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8393 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8394 (aarch64_vectorize_vec_perm_const): ...this new function.
8395 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8396 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8397 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8398 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8399 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8400 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8401 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8403 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8404 check for NEON modes.
8405 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8406 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8407 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8408 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8410 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8411 the old VEC_PERM_CONST conditions.
8412 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8413 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8414 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8415 (ia64_vectorize_vec_perm_const_ok): Merge into...
8416 (ia64_vectorize_vec_perm_const): ...this new function.
8417 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8418 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8419 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8420 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8421 * config/mips/mips.c (mips_expand_vec_perm_const)
8422 (mips_vectorize_vec_perm_const_ok): Merge into...
8423 (mips_vectorize_vec_perm_const): ...this new function.
8424 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8425 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8426 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8427 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8428 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8429 (rs6000_expand_vec_perm_const): Delete.
8430 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8432 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8433 (altivec_expand_vec_perm_const_le): Take each operand individually.
8434 Operate on constant selectors rather than rtxes.
8435 (altivec_expand_vec_perm_const): Likewise. Update call to
8436 altivec_expand_vec_perm_const_le.
8437 (rs6000_expand_vec_perm_const): Delete.
8438 (rs6000_vectorize_vec_perm_const_ok): Delete.
8439 (rs6000_vectorize_vec_perm_const): New function.
8440 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8441 an element count and rtx array.
8442 (rs6000_expand_extract_even): Update call accordingly.
8443 (rs6000_expand_interleave): Likewise.
8444 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8445 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8446 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8447 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8448 (rs6000_expand_vec_perm_const): Delete.
8449 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8450 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8451 (altivec_expand_vec_perm_const_le): Take each operand individually.
8452 Operate on constant selectors rather than rtxes.
8453 (altivec_expand_vec_perm_const): Likewise. Update call to
8454 altivec_expand_vec_perm_const_le.
8455 (rs6000_expand_vec_perm_const): Delete.
8456 (rs6000_vectorize_vec_perm_const_ok): Delete.
8457 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8458 reference to the SPE evmerge intructions.
8459 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8460 an element count and rtx array.
8461 (rs6000_expand_extract_even): Update call accordingly.
8462 (rs6000_expand_interleave): Likewise.
8463 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8464 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8466 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8468 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8470 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8471 vector mode and that that mode matches the mode of the data
8473 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8474 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8475 directly using expand_vec_perm_1 when forcing selectors into
8477 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8479 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8481 * optabs-query.h (can_vec_perm_p): Delete.
8482 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8483 * optabs-query.c (can_vec_perm_p): Split into...
8484 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8485 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8486 particular selector is valid.
8487 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8488 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8489 (vect_grouped_load_supported): Likewise.
8490 (vect_shift_permute_load_chain): Likewise.
8491 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8492 (vect_transform_slp_perm_load): Likewise.
8493 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8494 (vectorizable_bswap): Likewise.
8495 (vect_gen_perm_mask_checked): Likewise.
8496 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8497 implementations of variable permutation vectors into account
8498 when deciding which selector to use.
8499 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8500 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8501 with a false third argument.
8502 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8503 to test whether the constant selector is valid and can_vec_perm_var_p
8504 to test whether a variable selector is valid.
8506 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8508 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8509 * optabs-query.c (can_vec_perm_p): Likewise.
8510 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8511 instead of vec_perm_indices.
8512 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8513 (vect_gen_perm_mask_checked): Likewise,
8514 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8515 (vect_gen_perm_mask_checked): Likewise,
8517 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8519 * optabs-query.h (qimode_for_vec_perm): Declare.
8520 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8521 (qimode_for_vec_perm): ...this new function.
8522 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8524 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8526 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8527 does not have a conditional at the top.
8529 2018-01-02 Richard Biener <rguenther@suse.de>
8531 * ipa-inline.c (big_speedup_p): Fix expression.
8533 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8536 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8539 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8543 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8544 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8545 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8546 cond_taken_branch_cost 3->4.
8548 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8550 PR tree-optimization/83581
8551 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8552 TODO_cleanup_cfg if any changes have been made.
8555 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8556 convert_modes if target mode has the right side, but different mode
8560 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8561 last argument when extracting from CONCAT. If either from_real or
8562 from_imag is NULL, use expansion through memory. If result is not
8563 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8564 the parts directly to inner mode, if even that fails, use expansion
8568 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8569 check for bswap in mode rather than HImode and use that in expand_unop
8572 Copyright (C) 2018 Free Software Foundation, Inc.
8574 Copying and distribution of this file, with or without modification,
8575 are permitted in any medium without royalty provided the copyright
8576 notice and this notice are preserved.