1 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4 * lra-assigns.c (assign_by_spills): Return a flag of reload
5 assignment failure. Do not process the reload assignment
6 failures. Do not spill other reload pseudos if they has the same
8 (lra_assign): Add a return arg. Set up from the result of
10 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12 usage_insns if it is not NULL.
13 (spill_hard_reg_in_range): New function.
14 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17 (lra_assign): Change prototype.
18 * lra.c (lra): Add code to deal with fails by splitting hard reg
21 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
25 Accept complain bool parameter. Only emit errors if it is true.
26 (arm_parse_cpu_option_name): Likewise.
27 (arm_target_thumb_only): Adjust callers of the above.
28 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
29 prototype to take a default true bool parameter.
30 (arm_parse_arch_option_name): Likewise.
32 2018-03-09 David Malcolm <dmalcolm@redhat.com>
33 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
37 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
38 * configure: Regenerate.
39 * configure.ac ("linker --version-script option"): New.
40 ("linker soname option"): New.
42 2018-03-09 Richard Biener <rguenther@suse.de>
44 PR tree-optimization/84775
45 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
46 immediate uses of predicate stmts and mark them modified.
49 PR tree-optimization/84178
50 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
52 (version_loop_for_if_conversion): Delay update_ssa call.
53 (tree_if_conversion): Delay update_ssa until after predicate
56 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
59 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
60 when the function accesses prior frames.
62 2018-03-08 Jakub Jelinek <jakub@redhat.com>
65 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
66 gen_llsym, otherwise call maybe_gen_llsym.
69 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
70 has ',' character inside of it.
72 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
75 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
76 as clobbering CC_REGNUM.
78 2018-03-08 Richard Biener <rguenther@suse.de>
81 * tree-scalar-evolution.c: Include tree-into-ssa.h.
82 (follow_copies_to_constant): Do not follow SSA names registered
85 2018-03-08 Richard Biener <rguenther@suse.de>
87 PR tree-optimization/84178
88 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
90 (version_loop_for_if_conversion): Delay update_ssa call.
91 (tree_if_conversion): Delay update_ssa until after predicate
94 2018-03-08 David Malcolm <dmalcolm@redhat.com>
96 PR tree-optimization/84178
97 * tree-if-conv.c (release_bb_predicate): Remove the
98 the assertion that the stmts have NULL use_ops.
99 Discard the statements, asserting that they haven't
100 yet been added to a BB.
102 2018-03-08 Richard Biener <rguenther@suse.de>
104 PR tree-optimization/84746
105 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
106 (phi_translate): Pass in destination ANTIC_OUT set.
107 (phi_translate_1): Likewise. For a simplified result lookup
108 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
109 (phi_translate_set): Adjust.
110 (do_pre_regular_insertion): Likewise.
111 (do_pre_partial_partial_insertion): Likewise.
113 2018-03-08 Martin Liska <mliska@suse.cz>
115 PR gcov-profile/84735
116 * doc/gcov.texi: Document usage of profile files.
117 * gcov-io.h: Document changes in the format.
119 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
123 * dwarf2out.c (struct dw_line_info_table): Update comments for
125 (FORCE_RESET_NEXT_VIEW): New.
126 (FORCE_RESETTING_VIEW_P): New.
127 (RESETTING_VIEW_P): Check for -1 too.
128 (ZERO_VIEW_P): Likewise.
129 (new_line_info_table): Force-reset next view.
130 (dwarf2out_begin_function): Likewise.
131 (dwarf2out_source_line): Simplify zero_view_p initialization.
132 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
133 view directly. Omit view when omitting .loc at line 0.
135 2018-03-08 Jakub Jelinek <jakub@redhat.com>
137 PR tree-optimization/84740
138 * tree-switch-conversion.c (process_switch): Call build_constructors
139 only if info.phi_count is non-zero.
141 PR tree-optimization/84739
142 * tree-tailcall.c (find_tail_calls): Check call arguments against
143 DECL_ARGUMENTS (current_function_decl) rather than
144 DECL_ARGUMENTS (func) when checking for tail recursion.
146 2018-03-07 Jakub Jelinek <jakub@redhat.com>
148 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
149 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
150 Volker Reichelt's entry and add entries for people that perform
151 GCC fuzzy testing and report numerous bugs.
153 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
156 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
157 readonly data in sdata, if that is disabled.
158 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
159 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
160 -mreadonly-in-sdata option.
162 2018-03-07 Martin Sebor <msebor@redhat.com>
164 PR tree-optimization/84468
165 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
166 basic block when looking for nul assignment.
168 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
171 * except.h (output_function_exception_table): Adjust prototype.
172 * except.c (output_function_exception_table): Remove FNNAME parameter
173 and add SECTION parameter. Ouput one part of the table at a time.
174 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
175 the first part of the exception table and emit unwind directives.
176 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
177 (i386_pe_seh_cold_init): Likewise.
178 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
179 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
180 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
181 (ix86_output_call_insn): Emit a nop in one more case for SEH.
182 * config/i386/winnt.c: Include except.h.
183 (struct seh_frame_state): Add reg_offset, after_prologue and
184 in_cold_section fields.
185 (i386_pe_seh_end_prologue): Set seh->after_prologue.
186 (i386_pe_seh_cold_init): New function.
187 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
188 to seh->in_cold_section.
189 (seh_emit_push): Record the offset of the push.
190 (seh_emit_save): Record the offet of the save.
191 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
192 Test seh->after_prologue to disregard the epilogue.
193 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
194 (i386_pe_end_cold_function): New function.
196 2018-03-07 Jakub Jelinek <jakub@redhat.com>
199 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
200 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
203 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
205 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
206 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
209 * multiple_target.c: Include tree-inline.h and intl.h.
210 (expand_target_clones): Diagnose and fail if node->definition and
211 !tree_versionable_function_p (node->decl).
213 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
215 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
217 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
218 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
219 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
221 2018-03-06 Jakub Jelinek <jakub@redhat.com>
224 * combine.c (try_combine): Use reg_or_subregno instead of handling
225 just paradoxical SUBREGs and REGs.
227 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
229 * config/arc/arc.c (arc_finalize_pic): Remove function.
230 (arc_must_save_register): We use single base PIC register, remove
231 checks to save/restore the PIC register.
232 (arc_expand_prologue): Likewise.
233 * config/arc/arc-protos.h (arc_set_default_type_attributes):
235 (arc_verify_short): Likewise.
236 (arc_attr_type): Likewise.
237 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
238 (walk_stores): Likewise.
239 (arc_address_cost): Make it static.
240 (arc_verify_short): Likewise.
241 (branch_dest): Likewise.
242 (arc_attr_type): Likewise.
243 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
244 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
245 (arc_final_prescan_insn): Remove inserting the nops due to
246 hardware hazards. It is done in reorg step.
247 (insn_length_variant_t): Remove.
248 (insn_length_parameters_t): Likewise.
249 (arc_insn_length_parameters): Likewise.
250 (arc_get_insn_variants): Likewise.
251 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
253 2018-03-06 Jakub Jelinek <jakub@redhat.com>
256 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
259 PR tree-optimization/84687
260 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
262 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
264 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
266 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
267 Rename to ppc_speculation_barrier.
268 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
269 __builtin_ppc_speculation_barrier.
271 2018-03-05 Jakub Jelinek <jakub@redhat.com>
274 * combine.c (combine_simplify_rtx): Don't try to simplify if
275 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
278 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
280 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
281 to 32 bytes when compiling for POWER9.
283 2018-03-05 Jakub Jelinek <jakub@redhat.com>
286 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
287 regparm >= 3 with no arg reg available also for calls with
288 flag_force_indirect_call. Pass decl to ix86_function_regparm.
291 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
293 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
295 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
298 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
300 2018-03-05 Richard Biener <rguenther@suse.de>
302 PR tree-optimization/84486
303 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
304 When inserting a __builtin_assume_aligned call set the LHS
305 SSA name alignment info accordingly.
307 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
309 PR tree-optimization/84114
310 * config/aarch64/aarch64.c (aarch64_reassociation_width)
311 Avoid reassociation of FLOAT_MODE addition.
313 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
315 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
316 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
317 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
318 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
319 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
320 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
321 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
323 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
324 __builtin_ia32_wbinvd): New builtins.
325 (SPECIAL_ARGS2): New.
326 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
327 (SPECIAL_ARGS2): New.
328 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
329 (ix86_valid_target_attribute_inner_p): Ditto.
330 (ix86_init_mmx_sse_builtins): Add special_args2.
331 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
332 TARGET_WBNOINVD_P): New.
333 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
334 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
335 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
336 * config/i386/immintrin.h (_wbinvd): New intrinsic.
337 * config/i386/pconfigintrin.h: New file.
338 * config/i386/wbnoinvdintrin.h: Ditto.
339 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
340 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
342 2018-03-05 Richard Biener <rguenther@suse.de>
344 PR tree-optimization/84670
345 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
347 (BB_VISITED_WITH_VISITED_SUCCS): New define.
348 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
349 (compute_antic_aux): Only assert the number of values in ANTIC_IN
350 doesn't grow if all successors (recursively) were visited at least
353 2018-03-05 Richard Biener <rguenther@suse.de>
355 PR tree-optimization/84650
356 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
357 if executed in the loop pipeline.
359 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
361 * doc/configfiles.texi (Configuration Files): Move info about
362 conditionalizing $target-protos.h to...
363 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
364 differs from $target-protos.h.
366 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
367 Chung-Ju Wu <jasonwucj@gmail.com>
369 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
370 * config/nds32/nds32-multiple.md (setmemsi): Define.
371 * config/nds32/nds32-memory-manipulation.c
372 (nds32_gen_dup_4_byte_to_word_value): New.
373 (emit_setmem_word_loop): New.
374 (emit_setmem_byte_loop): New.
375 (nds32_expand_setmem_loop): New.
376 (nds32_expand_setmem_loop_v3m): New.
377 (nds32_expand_setmem_unroll): New.
378 (nds32_expand_setmem): New.
380 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
381 Chung-Ju Wu <jasonwucj@gmail.com>
383 * config/nds32/nds32-memory-manipulation.c
384 (nds32_emit_load_store): New.
385 (nds32_emit_post_inc_load_store): New.
386 (nds32_emit_mem_move): New.
387 (nds32_emit_mem_move_block): New.
388 (nds32_expand_movmemsi_loop_unknown_size): New.
389 (nds32_expand_movmemsi_loop_known_size): New.
390 (nds32_expand_movmemsi_loop): New.
391 (nds32_expand_movmemsi_unroll): New.
392 (nds32_expand_movmemqi): Rename ...
393 (nds32_expand_movmemsi): ... to this.
394 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
395 (movmemsi): ... to this.
396 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
397 (nds32_expand_movmemsi): ... to this.
399 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
400 Monk Chiang <sh.chiang04@gmail.com>
401 Chung-Ju Wu <jasonwucj@gmail.com>
403 * config/nds32/nds32-protos.h
404 (nds32_expand_load_multiple): New arguments.
405 (nds32_expand_store_multiple): Ditto.
406 (nds32_valid_multiple_load_store): Rename ...
407 (nds32_valid_multiple_load_store_p): ... to this.
408 * config/nds32/nds32-memory-manipulation.c
409 (nds32_expand_load_multiple): Refine implementation.
410 (nds32_expand_store_multiple): Ditto.
411 * config/nds32/nds32-multiple.md
412 (load_multiple): Update nds32_expand_load_multiple interface.
413 (store_multiple): Update nds32_expand_store_multiple interface.
414 * config/nds32/nds32-predicates.c
415 (nds32_valid_multiple_load_store): Rename ...
416 (nds32_valid_multiple_load_store_p): ... to this and refine
418 * config/nds32/predicates.md
419 (nds32_load_multiple_and_update_address_operation): New predicate.
420 (nds32_store_multiple_and_update_address_operation): New predicate.
422 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
423 Chung-Ju Wu <jasonwucj@gmail.com>
425 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
426 (combo): New attribute.
427 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
429 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
431 * config/nds32/nds32.opt: Change -mcmodel= default value.
433 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
434 Monk Chiang <sh.chiang04@gmail.com>
435 Chung-Ju Wu <jasonwucj@gmail.com>
437 * config/nds32/constants.md (unspec_element): New enum.
438 * config/nds32/constraints.md (Umw): New constraint.
439 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
440 * config/nds32/nds32-intrinsic.md: Likewise.
441 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
442 (nds32_valid_smw_lwm_base_p): New.
443 (nds32_output_smw_single_word): New.
444 (nds32_output_lmw_single_word): New.
445 (nds32_expand_unaligned_load): New.
446 (nds32_expand_unaligned_store): New.
447 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
448 (nds32_output_smw_single_word): Declare.
449 (nds32_output_lmw_single_word): Declare.
450 (nds32_expand_unaligned_load): Declare.
451 (nds32_expand_unaligned_store): Declare.
452 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
453 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
454 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
455 NDS32_BUILTIN_UASTORE_DW.
456 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
459 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
460 Kito Cheng <kito.cheng@gmail.com>
461 Chung-Ju Wu <jasonwucj@gmail.com>
463 * config/nds32/nds32-intrinsic.c
464 (nds32_expand_builtin_null_ftype_reg): Delete.
465 (nds32_expand_builtin_reg_ftype_imm): Ditto.
466 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
467 (nds32_read_argument): New.
468 (nds32_legitimize_target): Ditto.
469 (nds32_legitimize_argument): Ditto.
470 (nds32_check_constant_argument): Ditto.
471 (nds32_expand_unop_builtin): Ditto.
472 (nds32_expand_unopimm_builtin): Ditto.
473 (nds32_expand_binop_builtin): Ditto.
474 (nds32_builtin_decl_impl): Ditto.
475 (builtin_description): Ditto.
476 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
477 (nds32_init_builtins_impl): Ditto.
478 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
479 (nds32_builtin_decl): New.
480 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
481 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
483 2018-03-02 Jeff Law <law@redhat.com>
485 * reorg.c (stop_search_p): Handle DEBUG_INSN.
486 (redundant_insn, fill_simple_delay_slots): Likewise.
487 (fill_slots_from_thread): Likewise.
488 * resource.c (mark_referenced_resources): Likewise.
489 (mark_set_resources, find_dead_or_set_registers): Likewise.
491 2018-03-02 Jakub Jelinek <jakub@redhat.com>
493 * substring-locations.h (format_warning_va): Formatting fix for
495 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
497 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
498 * substring-locations.c: Include intl.h.
499 (format_warning_va): Turned into small wrapper around
500 format_warning_n_va, renamed to ...
501 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
502 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
504 (format_warning_at_substring_n): New function.
505 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
506 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
507 format_warning_at_substring with just a shorter name instead of
508 const function pointer.
509 (fmtwarn_n): New function.
510 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
511 appropriate, get rid of all the fmtstr temporaries, move conditionals
512 with G_() wrapped string literals directly into fmtwarn arguments,
513 cast dir.len to (int), formatting fixes.
515 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
517 * doc/invoke.texi: Remove "Cilk Plus" references.
519 2018-03-02 Jakub Jelinek <jakub@redhat.com>
520 Richard Biener <rguenther@suse.de>
523 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
524 for error or warning attributes if CALL_FROM_THUNK_P is set.
527 2018-03-02 Jakub Jelinek <jakub@redhat.com>
530 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
531 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
534 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
535 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
537 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
538 instead of -1U in last predictors element's probability member.
540 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
543 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
544 arguments if they are comparable.
546 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
548 PR tree-optimization/84634
549 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
550 masks and masked_loop_p with a single loop_masks, making sure it's
551 null for bb vectorization.
553 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
555 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
556 (vect_analyze_data_ref_access): Use loop->safe_len rather than
557 loop->force_vectorize to check whether there is no alias.
559 2018-03-02 Jakub Jelinek <jakub@redhat.com>
562 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
564 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
566 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
567 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
568 instead of a loop around prev_real_insn.
569 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
573 * config/i386/i386.c (ix86_print_operand): Use conditional
574 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
577 2018-03-02 Richard Biener <rguenther@suse.de>
579 PR tree-optimization/84427
580 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
581 (bitmap_set_subtract_values): Rewrite to handle multiple
584 (prune_clobbered_mems): Likewise.
585 (phi_translate): Take edge instead of pred/phiblock.
586 (phi_translate_1): Likewise.
587 (phi_translate_set): Likewise. Insert all translated
588 exprs for a value into the set, keeping possibly multiple
589 expressions per value.
590 (compute_antic_aux): Adjust for phi_translate changes.
591 When intersecting union the expressions and prune those
592 not in the final value set, keeping possibly multiple
593 expressions per value. Do not use value-insertion
594 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
595 all expressions. Add verification that the value-sets
596 only shrink during iteration.
597 (compute_partial_antic_aux): Adjust for the phi_translate changes.
598 (do_pre_regular_insertion): Likewise.
599 (do_pre_partial_partial_insertion): Likewise.
601 2018-03-02 Richard Biener <rguenther@suse.de>
604 * config/darwin.c (saved_debug_info_level): New static global.
605 (darwin_asm_lto_start): Disable debug info generation for LTO out.
606 (darwin_asm_lto_end): Restore debug info generation settings.
608 2018-03-01 Martin Liska <mliska@suse.cz>
611 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
614 2018-03-01 Richard Biener <rguenther@suse.de>
617 * dwarf2out.c (gen_variable_die): Properly handle late VLA
618 type annotation with LTO when debug was disabled at compile-time.
620 2018-03-01 Matthew Fortune <mfortune@gmail.com>
622 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
624 (mips_final_postscan_insn): Likewise.
626 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
628 PR rtl-optimization/84528
629 * alias.c (init_alias_target): Add commentary.
630 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
631 a unique base value if the frame pointer is not eliminated
632 to the stack pointer.
634 2018-03-01 Tom de Vries <tom@codesourcery.com>
636 PR rtl-optimization/83327
637 * lra-int.h (hard_regs_spilled_into): Declare.
638 * lra.c (hard_regs_spilled_into): Define.
639 (init_reg_info): Init hard_regs_spilled_into.
640 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
641 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
642 (process_bb_lives): Handle hard_regs_spilled_into.
643 (lra_create_live_ranges_1): Before doing liveness propagation, clear
644 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
646 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
648 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
649 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
650 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
651 * config/rs6000/aix72.h: New file.
653 2018-02-28 Jakub Jelinek <jakub@redhat.com>
655 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
656 instead of warning_at with conditional singular and plural messages
660 * stor-layout.c (update_alignment_for_field): For
661 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
662 && !DECL_PACKED (field), do the alignment update, just use
663 only desired_align instead of MAX (type_align, desired_align)
665 (place_field): Don't do known_align < desired_align handling
666 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
667 is non-NULL, instead do it after rli->prev_field handling and
668 only if not within a bitfield word. For DECL_PACKED (field)
669 use type_align of BITS_PER_UNIT.
671 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
673 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
674 superfluous parentheses and trailing spaces.
676 2018-02-28 Richard Biener <rguenther@suse.de>
678 PR tree-optimization/84584
679 * graphite-scop-detection.c (scop_detection::add_scop): Discard
680 SCoPs with fake exit edge.
682 2018-02-28 Martin Liska <mliska@suse.cz>
685 * timevar.c (timer::print): Fix format to properly print 100%
688 2018-02-28 Richard Biener <rguenther@suse.de>
691 * genmatch.c (capture_info::walk_match): Do not mark
692 captured expressions without operands as expr_p given
693 they act more like predicates and should be subject to
694 "lost tail" side-effect preserving.
696 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
698 PR rtl-optimization/81611
699 * auto-inc-dec.c (attempt_change): Move dead note from
700 mem_insn if it's the next use of regno
701 (find_address): Take address use of reg holding
702 non-incremented value. Add parm to limit search to the named
704 (merge_in_block): Attempt to use a mem insn that is the next
705 use of the original regno.
707 2018-02-27 Martin Sebor <msebor@redhat.com>
710 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
711 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
713 2018-02-27 Martin Sebor <msebor@redhat.com>
716 * diagnostic-core.h (warning_n, error_n, inform_n): Change
717 n argument to unsigned HOST_WIDE_INT.
718 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
719 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
720 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
721 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
723 2018-02-27 Richard Biener <rguenther@suse.de>
725 PR tree-optimization/84512
726 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
727 Do not use the estimate returned from record_stmt_cost for
728 the scalar iteration cost but sum properly using add_stmt_cost.
730 2018-02-27 Richard Biener <rguenther@suse.de>
732 PR tree-optimization/84466
733 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
734 Adjust last change to less strictly validate use operands.
736 2018-02-27 Martin Liska <mliska@suse.cz>
738 PR gcov-profile/84548
739 * gcov.c (process_file): Allow partial overlap and consider it
740 also as group functions.
741 (output_lines): Properly calculate range of lines for a group.
743 2018-02-27 Martin Liska <mliska@suse.cz>
745 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
746 'ggc' suffixes. Change first column width.
747 (timer::print): Fix formatting of the column.
749 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
751 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
752 preserve inline entry blocks for the sake of debug inline
753 entry point markers alone.
754 (remove_unused_locals): Suggest in comments a better place to
755 force the preservation of inline entry blocks that are
756 otherwise unused, but do not preserve them.
758 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
760 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
762 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
765 * config/i386/constraints.md (Bs): Replace
766 ix86_indirect_branch_register with
767 TARGET_INDIRECT_BRANCH_REGISTER.
769 * config/i386/i386.md (indirect_jump): Likewise.
770 (tablejump): Likewise.
771 (*sibcall_memory): Likewise.
772 (*sibcall_value_memory): Likewise.
773 Peepholes of indirect call and jump via memory: Likewise.
774 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
775 (*sibcall_value_GOT_32): Likewise.
776 * config/i386/predicates.md (indirect_branch_operand): Likewise.
777 (GOT_memory_operand): Likewise.
778 (call_insn_operand): Likewise.
779 (sibcall_insn_operand): Likewise.
780 (GOT32_symbol_operand): Likewise.
781 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
783 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
785 PR rtl-optimization/83496
786 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
787 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
788 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
789 redundant insn, if any.
790 (relax_delay_slots): Likewise.
791 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
793 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
795 PR tree-optimization/83965
796 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
797 that grouped statements are part of a reduction chain. Return
798 true if the statement is not marked as a reduction itself but
800 (vect_recog_dot_prod_pattern): Don't check whether the statement
801 is part of a group here.
802 (vect_recog_sad_pattern): Likewise.
803 (vect_recog_widen_sum_pattern): Likewise.
805 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
808 * final.c (rest_of_clean_state): Also look for calls inside sequences.
810 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
813 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
815 (ix86_output_indirect_function_return): New prototype.
816 (ix86_split_simple_return_pop_internal): Likewise.
817 * config/i386/i386.c (indirect_return_via_cx): New.
818 (indirect_return_via_cx_bnd): Likewise.
819 (indirect_thunk_name): Handle return va CX_REG.
820 (output_indirect_thunk_function): Create alias for
821 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
822 (ix86_output_indirect_jmp): Remove the bool argument.
823 (ix86_output_indirect_function_return): New function.
824 (ix86_split_simple_return_pop_internal): Likewise.
825 * config/i386/i386.md (*indirect_jump): Don't pass false
826 to ix86_output_indirect_jmp.
827 (*tablejump_1): Likewise.
828 (simple_return_pop_internal): Change it to define_insn_and_split.
829 Call ix86_split_simple_return_pop_internal to split it for
831 (simple_return_indirect_internal): Call
832 ix86_output_indirect_function_return instead of
833 ix86_output_indirect_jmp.
835 2018-02-26 Jakub Jelinek <jakub@redhat.com>
838 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
839 memset and value initialization afterwards.
841 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
843 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
845 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
848 * common/config/aarch64/aarch64-common.c
849 (aarch_option_optimization_table[]): Switch
850 off fomit-frame-pointer
852 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
853 Chung-Ju Wu <jasonwucj@gmail.com>
855 * config/nds32/nds32-multiple.md (load_multiple): Disallow
857 (store_multiple): Ditto.
859 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
861 * config.gcc: Add --with-cpu support for nds32 target.
862 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
863 * config/nds32/nds32.opt: Add -mcpu= option.
865 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
867 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
868 isel=yes): Warn for these deprecated options.
870 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
872 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
873 ISA_2_5_MASKS_EMBEDDED.
875 2018-02-23 Jakub Jelinek <jakub@redhat.com>
877 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
878 p->max as pointers rather than using iterative_hash_expr.
880 2018-02-23 Carl Love <cel@us.ibm.com>
882 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
883 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
885 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
886 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
887 P8V_BUILTIN_VEC_VUNSIGNED2.
889 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
892 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
893 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
895 * lra-constraints.c (curr_insn_transform): Set up
896 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
898 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
899 * lra-eliminations.c (spill_pseudos): Ditto.
900 (process_insn_for_elimination): Ditto.
901 * lra-lives.c (reg_early_clobber_p): Use the new macros.
902 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
903 LRA_NON_CLOBBERED_ALT.
905 2018-02-22 Martin Sebor <msebor@redhat.com>
907 PR tree-optimization/84480
908 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
909 to maybe_diag_stxncpy_trunc. Call it.
910 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
911 from gimple_fold_builtin_strcpy. Print inlining stack.
912 (handle_builtin_stxncpy): Print inlining stack.
913 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
915 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
918 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
919 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
920 and -fcheck-pointer-bounds are used together.
921 (indirect_thunk_prefix): New enum.
922 (indirect_thunk_need_prefix): New function.
923 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
924 "_nt" instead of "_bnd" for NOTRACK prefix.
925 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
926 (output_indirect_thunk_function): Likewise.
928 (ix86_code_end): Update output_indirect_thunk_function calls.
929 (ix86_output_indirect_branch_via_reg): Replace
930 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
931 (ix86_output_indirect_branch_via_push): Likewise.
932 (ix86_output_function_return): Likewise.
933 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
934 incompatible with -fcf-protection=branch and
935 -fcheck-pointer-bounds.
937 2018-02-22 Steve Ellcey <sellcey@cavium.com>
940 * config/aarch64/aarch64.c (aarch64_print_address_internal):
941 Change gcc_assert call to output_operand_lossage.
943 2018-02-22 Steve Ellcey <sellcey@cavium.com>
945 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
947 2018-02-22 DJ Delorie <dj@redhat.com>
948 Sebastian Perta <sebastian.perta@renesas.com>
949 Oleg Endo <olegendo@gcc.gnu.org>
951 * config/rx/rx.c (rx_rtx_costs): New function.
952 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
954 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
956 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
958 2018-02-22 Martin Liska <mliska@suse.cz>
961 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
962 Add "native" as a possible value.
964 2018-02-22 Martin Liska <mliska@suse.cz>
967 * config/i386/i386.c (ix86_option_override_internal):
968 Add "native" as a possible value for -march and -mtune.
970 2018-02-22 Jakub Jelinek <jakub@redhat.com>
973 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
974 to all type variants.
976 PR tree-optimization/84503
977 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
978 width as info->bitpos + info->bitsize - start.
979 (merged_store_group::merge_overlapping): Simplify width computation.
980 (check_no_overlap): New function.
981 (imm_store_chain_info::try_coalesce_bswap): Compute expected
982 start + width and last_order of the group, fail if check_no_overlap
984 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
985 to group if check_no_overlap fails.
987 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
989 * config/rs6000/altivec.md: Delete contraint arguments to
990 define_expand, define_split, and define_peephole2, and in
991 define_insn_and_split if always unused.
992 * config/rs6000/darwin.md: Ditto.
993 * config/rs6000/dfp.md: Ditto.
994 * config/rs6000/rs6000.md: Ditto.
995 * config/rs6000/sync.md: Ditto.
996 * config/rs6000/vector.md: Ditto.
997 * config/rs6000/vsx.md: Ditto.
999 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1001 * config/rs6000/altivec.md: Write output control strings as braced
1002 blocks instead of double-quoted strings.
1003 * config/rs6000/darwin.md: Ditto.
1004 * config/rs6000/rs6000.md: Ditto.
1005 * config/rs6000/vector.md: Ditto.
1006 * config/rs6000/vsx.md: Ditto.
1008 2018-02-21 Jason Merrill <jason@redhat.com>
1010 PR c++/84314 - ICE with templates and fastcall attribute.
1011 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1013 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1015 * ipa-cp.c (determine_versionability): Fix comment typos.
1017 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1020 * ipa-cp.c (determine_versionability): Do not version functions caling
1023 2018-02-21 Martin Liska <mliska@suse.cz>
1026 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1027 Add "native" as a possible value.
1028 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1029 the macro when native cpu detection is available.
1031 2018-02-21 Martin Liska <mliska@suse.cz>
1034 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1035 Add "native" as a possible value.
1036 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1037 when native cpu detection is available.
1039 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1040 Martin Sebor <msebor@redhat.com>
1042 PR tree-optimization/84478
1043 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1045 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1046 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1047 support which is conservatively correct, for 2 only stay conservative
1048 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1049 argument to the 2 argument get_range_strlen, adjust 6 arg
1050 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1052 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1053 (gimple_fold_builtin_strlen): Pass true as last argument to
1056 2018-02-20 Martin Sebor <msebor@redhat.com>
1059 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1060 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1061 (builtin_memref::builtin_memref): Factor out parts into
1062 set_base_and_offset and call it.
1064 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1067 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1068 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1069 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1070 search at the associated MODE_INT.
1072 2018-02-20 Jeff Law <law@redhat.com>
1075 PR tree-optimization/81592
1077 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1078 for range data rather than using global data.
1079 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1080 range data rather than using global data.
1081 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1082 pass it to children as needed.
1083 (struct directive::fmtresult): Similarly.
1084 (struct directive::set_width): Similarly.
1085 (struct directive::set_precision): Similarly.
1086 (format_integer, format_directive, parse_directive): Similarly.
1087 (format_none): Accept unnamed vr_values parameter.
1088 (format_percent, format_floating, format_character): Similarly.
1089 (format_string, format_plain): Similarly.
1090 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1091 the EVRP range analyzer for range data rather than using global data.
1092 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1093 gimple-ssa-evrp-analyze.h
1094 (class sprintf_dom_walker): Add after_dom_children member function.
1095 Add evrp_range_analyzer member.
1096 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1097 range analyzer as needed.
1098 (sprintf_dom_walker::after_dom_children): New member function.
1099 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1101 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1102 (evrp_range_analyzer::pop_to_marker): Likewise.
1104 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1106 PR tree-optimization/84419
1107 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1108 with the required type if its current type is compatible but
1111 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1114 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1115 after vectorization.
1117 2018-02-20 Martin Liska <mliska@suse.cz>
1120 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1121 possible values if we don't have a hint.
1123 2018-02-20 Martin Liska <mliska@suse.cz>
1127 * final.c (shorten_branches): Build align_tab array with one
1129 * opts.c (finish_options): Add alignment option limit check.
1130 (MAX_CODE_ALIGN): Likewise.
1131 (MAX_CODE_ALIGN_VALUE): Likewise.
1132 * doc/invoke.texi: Document maximum allowed option value for
1133 all -falign-* options.
1135 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1138 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1139 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1140 * var-tracking.c (emit_note_insn_var_location): Remove all references
1141 to NOTE_INSN_CALL_ARG_LOCATION.
1142 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1143 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1144 Use copy_rtx_if_shared.
1145 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1146 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1147 (dwarf2out_var_location): Remove handling of
1148 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1150 * final.c (final_scan_insn): Remove all references to
1151 NOTE_INSN_CALL_ARG_LOCATION.
1152 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1153 before dumping final insns.
1154 * except.c (emit_note_eh_region_end): Remove all references to
1155 NOTE_INSN_CALL_ARG_LOCATION.
1156 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1157 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1158 * config/arc/arc.c (hwloop_optimize): Likewise.
1159 * config/arm/arm.c (create_fix_barrier): Likewise.
1160 * config/s390/s390.c (s390_chunkify_start): Likewise.
1161 * config/sh/sh.c (find_barrier): Likewise.
1162 * config/i386/i386.c (rest_of_insert_endbranch,
1163 ix86_seh_fixup_eh_fallthru): Likewise.
1164 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1165 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1166 * config/frv/frv.c (frv_function_prologue): Likewise.
1167 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1169 (note_outside_basic_block_p): Remove all references to
1170 NOTE_INSN_CALL_ARG_LOCATION.
1171 * gengtype.c (adjust_field_rtx_def): Likewise.
1172 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1174 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1175 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1178 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1181 PR tree-optimization/84452
1182 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1183 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1186 2018-02-19 Martin Liska <mliska@suse.cz>
1189 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1191 2018-02-19 Martin Liska <mliska@suse.cz>
1192 Richard Sandiford <richard.sandiford@linaro.org>
1194 PR tree-optimization/82491
1195 * gimple-fold.c (get_base_constructor): Make earlier bail out
1198 2018-02-19 Carl Love <cel@us.ibm.com>
1200 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1201 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1203 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1204 P8V_BUILTIN_VEC_NEG.
1206 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1208 * config/rl78/rl78.md (movdf): New define expand.
1210 2018-02-19 Martin Liska <mliska@suse.cz>
1213 * doc/invoke.texi: Fix typo.
1214 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1216 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1218 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1219 handle rs6000_single_float and rs6000_double_float specially for
1222 2018-02-16 Jeff Law <law@redhat.com>
1224 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1227 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1230 * ipa-inline.c (inline_small_functions): Fix a typo.
1232 2018-02-16 Nathan Sidwell <nathan@acm.org>
1234 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1236 2018-02-16 Carl Love <cel@us.ibm.com>
1238 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1239 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1240 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1241 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1242 expansion to P8V_BUILTIN_VEC_FLOAT2.
1244 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1246 PR rtl-optimization/70023
1247 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1248 src_regno into account.
1250 2018-02-16 Carl Love <cel@us.ibm.com>
1252 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1253 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1254 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1255 * config/rs6000/rs6000.c: Remove case statements for
1256 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1257 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1258 and P9V_BUILTIN_VEC_VINSERT4B.
1259 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1260 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1261 * config/rs6000/vsx.md:
1262 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1265 2018-02-16 Carl Love <cel@us.ibm.com>
1267 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1269 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1271 * config/rs6000/rs6000-c.c: Add the definitions for
1272 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1273 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1274 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1275 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1276 definition for insert4b and define insn *insert3b_internal.
1277 * doc/extend.texi: Add documentation for vec_extract4b.
1279 2018-02-16 Nathan Sidwell <nathan@acm.org>
1281 * doc/extend.texi (Backwards Compatibility): Mention friend
1282 injection. Note for-scope is deprecated.
1283 * doc/invoke.texi (-ffriend-injection): Deprecate.
1285 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1287 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1288 that moved to I2, also allow destinations that are a paradoxical
1289 subreg (instead of a normal reg).
1291 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1294 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1297 2018-02-16 Richard Biener <rguenther@suse.de>
1299 PR tree-optimization/84037
1300 PR tree-optimization/84016
1302 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1303 Adjust vec_construct for the fact we need additional higher latency
1304 128bit inserts for AVX256 and AVX512 vector builds.
1305 (ix86_add_stmt_cost): Scale vector construction cost for
1308 2018-02-16 Richard Biener <rguenther@suse.de>
1310 PR tree-optimization/84417
1311 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1312 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1313 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1315 2018-02-16 Martin Liska <mliska@suse.cz>
1318 * internal-fn.def (ASAN_CHECK): Set proper flags.
1319 (ASAN_MARK): Likewise.
1321 2018-02-16 Julia Koval <julia.koval@intel.com>
1323 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1324 from PTA_CANNONLAKE.
1326 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1329 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1330 Use ++iter rather than iter++ for std::list iterators.
1331 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1332 defer deleting them until all nodes in the forest are processed. Do
1333 free even leaf nodes. Change to_process into auto_vec.
1336 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1337 * vec.h (vec_default_construct): Use memset instead of placement new
1338 if BROKEN_VALUE_INITIALIZATION is defined.
1339 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1340 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1343 PR rtl-optimization/83723
1344 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1345 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1346 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1348 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1350 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1352 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1354 PR rtl-optimization/81443
1355 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1356 from inner REGs to paradoxical SUBREGs.
1358 2018-02-16 Richard Biener <rguenther@suse.de>
1360 PR tree-optimization/84399
1361 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1362 For operands we can analyze at their definition make sure we can
1363 analyze them at each use as well.
1365 2018-02-16 Richard Biener <rguenther@suse.de>
1367 PR tree-optimization/84190
1368 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1369 volatile accesses if the decl isn't volatile.
1371 2018-02-15 Jason Merrill <jason@redhat.com>
1373 PR c++/84314 - ICE with templates and fastcall attribute.
1374 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1375 TYPE_CANONICAL on an existing type.
1377 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1379 PR tree-optimization/84383
1380 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1381 dstoff nor call operand_equal_p if dstbase is NULL.
1383 PR tree-optimization/84334
1384 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1385 also a CONSTANT_CLASS_P, punt.
1387 2018-02-14 Jim Wilson <jimw@sifive.com>
1389 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1390 first SMALL_OPERAND check. New local min_second_step. Move assert
1391 to where locals are set. Add TARGET_RVC support.
1392 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1394 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1396 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1398 2018-02-14 Martin Sebor <msebor@redhat.com>
1400 PR tree-optimization/83698
1401 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1402 arrays constrain the offset range to their bounds.
1403 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1404 (builtin_access::overlap): Avoid setting the size of overlap if it's
1406 (maybe_diag_overlap): Also consider arrays when deciding what values
1407 of offsets to include in diagnostics.
1409 2018-02-14 Martin Sebor <msebor@redhat.com>
1412 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1413 that correspond to the kind of a declaration.
1415 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1418 * config/pa/pa.md: Load address of PIC label using the linkage table
1419 if the label is nonlocal.
1421 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1423 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1424 warning message if user requests -maltivec=be.
1425 * doc/invoke.texi: Document deprecation of -maltivec=be.
1427 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1430 * config/rs6000/rs6000-c.c: Update definitions for
1431 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1432 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1434 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1437 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1438 add _get_ssp intrinsics. Remove argument from
1439 __builtin_ia32_rdssp[d|q].
1440 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1441 * config/i386/i386-builtin.def: Remove argument from
1442 __builtin_ia32_rdssp[d|q].
1443 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1444 ix86_expand_special_args_builtin for _rdssp[d|q].
1445 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1446 Clear register before usage.
1447 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1448 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1450 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1452 PR tree-optimization/84357
1453 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1454 operand 1 of an ARRAY_REF too.
1456 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1459 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1460 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1462 (set_of_reg): New struct.
1463 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1464 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1465 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1467 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1468 Split into bitclr, bitset, bitinvert patterns if appropriate.
1469 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1470 use rx_fuse_in_memory_bitop.
1471 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1472 to named insn, correct maximum insn length.
1474 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1477 * machmode.def: Define a complex mode for PARTIAL_INT.
1478 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1480 * doc/rtl.texi: Document CSPImode.
1481 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1483 (msp430_hard_regno_nregs_with_padding): Likewise.
1485 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1488 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1490 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1492 PR rtl-optimization/84169
1493 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1494 we generated a parallel as new i3 and we split that to new i2 and i3
1495 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1496 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1497 those to i2, not i1. Partially rewrite this scan code.
1499 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1502 * stor-layout.c (place_field): For variable length fields, adjust
1503 offset_align afterwards not just based on the field's alignment,
1504 but also on the size.
1507 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1508 of exps and logs in the use_exp2 case.
1510 2018-02-13 Jeff Law <law@redhat.com>
1512 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1515 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1518 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1522 * common.opt (gas-loc-support, gas-locview-support): New.
1523 (ginline-points, ginternal-reset-location-views): New.
1524 * doc/invoke.texi: Document them. Use @itemx where intended.
1525 (gvariable-location-views): Adjust.
1526 * target.def (reset_location_view): New.
1527 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1528 (TARGET_RESET_LOCATION_VIEW): New.
1529 * doc/tm.texi: Rebuilt.
1530 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1531 (dwarf2out_default_as_locview_support): New.
1532 (output_asm_line_debug_info): Use option variables.
1533 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1534 (output_loc_list): Likewise.
1535 (add_high_low_attributes): Check option variables.
1536 Don't output entry view attribute in strict mode.
1537 (gen_inlined_subroutine_die): Check option variables.
1538 (dwarf2out_inline_entry): Likewise.
1539 (init_sections_and_labels): Likewise.
1540 (dwarf2out_early_finish): Likewise.
1541 (maybe_reset_location_view): New, from...
1542 (dwarf2out_var_location): ... here. Call it.
1543 * debug.h (dwarf2out_default_as_loc_support): Declare.
1544 (dwarf2out_default_as_locview_support): Declare.
1545 * hooks.c (hook_int_rtx_insn_0): New.
1546 * hooks.h (hook_int_rtx_insn_0): Declare.
1547 * toplev.c (process_options): Take -gas-loc-support and
1548 -gas-locview-support from dwarf2out. Enable
1549 -gvariable-location-views by default only with locview
1550 assembler support. Enable -ginternal-reset-location-views by
1551 default only if the target defines the corresponding hook.
1552 Enable -ginline-points by default if location views are
1553 enabled; force it disabled if statement frontiers are
1555 * tree-inline.c (expand_call_inline): Check option variables.
1556 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1558 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1560 PR tree-optimization/84321
1561 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1562 handling. Also check whether the anti-range contains any values
1563 that satisfy the mask; switch to a VR_RANGE if not.
1565 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1568 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1570 2018-02-13 Martin Jambor <mjambor@suse.cz>
1573 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1574 of call statements, also set location of a load to a temporary.
1576 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1578 * config/rl78/rl78.c (add_vector_labels): New function.
1579 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1580 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1581 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1582 which checks that no arguments are passed.
1583 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1584 * doc/extend.texi: Documentation for the new attribute.
1586 2018-02-13 Andreas Schwab <schwab@suse.de>
1588 * config/riscv/linux.h (CPP_SPEC): Define.
1590 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1593 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1594 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1595 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1596 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1597 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1598 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1599 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1600 temporarily for AES and PCLMUL builtins.
1602 PR tree-optimization/84339
1603 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1604 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1608 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1609 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1610 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1612 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1614 * omp-simd-clone.h: New file.
1615 * omp-simd-clone.c: Include omp-simd-clone.h.
1616 (expand_simd_clones): No longer static.
1617 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1618 cgraph.h and omp-simd-clone.h.
1619 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1620 (vect_recog_widen_shift_pattern): Formatting fix.
1621 (vect_pattern_recog_1): Don't check optab for calls.
1624 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1625 operands[2] into a REG before using gen_lowpart on it.
1627 2018-02-12 Jeff Law <law@redhat.com>
1630 * config/sh/sh.c (find_barrier): Consider a sibling call
1633 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1634 successfully back substituting a reg.
1636 2018-02-12 Richard Biener <rguenther@suse.de>
1638 PR tree-optimization/84037
1639 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1640 parameter, move visited init to caller.
1641 (vect_slp_analyze_operations): Separate cost from validity
1642 check, initialize visited once for all instances.
1643 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1645 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1647 (vect_model_promotion_demotion_cost): Likewise.
1648 (vectorizable_bswap): Guard cost modeling with !slp_node
1649 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1651 (vectorizable_call): Likewise.
1652 (vectorizable_conversion): Likewise.
1653 (vectorizable_assignment): Likewise.
1654 (vectorizable_shift): Likewise.
1655 (vectorizable_operation): Likewise.
1656 (vectorizable_store): Likewise.
1657 (vectorizable_load): Likewise.
1658 (vectorizable_condition): Likewise.
1659 (vectorizable_comparison): Likewise.
1661 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1664 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1665 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1666 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1668 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1671 * params.def (inline-min-speedup): Increase from 8 to 15.
1672 (max-inline-insns-auto): Decrease from 40 to 30.
1673 * ipa-split.c (consider_split): Add some buffer for function to
1674 be considered inlining candidate.
1675 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1678 2018-02-12 Richard Biener <rguenther@suse.de>
1680 PR tree-optimization/84037
1681 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1682 matched stmts if we cannot swap the non-matched ones.
1684 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1686 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1687 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1688 _mm_maskz_scalef_round_ss): New intrinsics.
1689 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1690 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1691 __builtin_ia32_scalefss_round): Remove.
1692 (__builtin_ia32_scalefsd_mask_round,
1693 __builtin_ia32_scalefss_mask_round): New intrinsics.
1694 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1695 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1696 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1697 "<round_constraint>")): Changed to ...
1698 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1699 "<round_scalar_constraint>")): ... this.
1700 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1701 %0, %1, %2<round_op3>}"): Changed to ...
1702 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1703 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1704 %2<round_scalar_mask_op3>}"): ... this.
1705 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1707 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1709 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1710 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1711 (_mm_maskz_sqrt_round_ss): New intrinsics.
1712 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1713 (__builtin_ia32_sqrtsd_mask_round)
1714 (__builtin_ia32_sqrtss_mask_round): New builtins.
1715 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1716 (__builtin_ia32_sqrtss_round): Remove.
1717 (__builtin_ia32_sqrtsd_mask_round)
1718 (__builtin_ia32_sqrtss_mask_round): New builtins.
1719 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1720 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1721 ((match_operand:VF_128 1 "vector_operand"
1722 "xBm,<round_constraint>")): Changed to ...
1723 ((match_operand:VF_128 1 "vector_operand"
1724 "xBm,<round_scalar_constraint>")): ... this.
1725 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1726 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1727 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1728 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1729 %<iptr>1<round_scalar_mask_op3>}): ... this.
1730 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1731 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1733 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1736 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1737 Cast vec_cmpeq result to correct type.
1738 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1739 Cast vec_cmpgt result to correct type.
1741 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1743 * final.c (final_scan_insn_1): Renamed from...
1744 (final_scan_insn): ... this. New wrapper, to recover
1745 seen from the outermost call in recursive ones.
1746 * config/sparc/sparc.c (output_return): Drop seen from call.
1747 (output_sibcall): Likewise.
1748 * config/visium/visium.c (output_branch): Likewise.
1750 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1752 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1755 2018-02-10 Alan Modra <amodra@gmail.com>
1758 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1759 Specify LR as an input.
1761 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1764 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1765 remove_member_access_dummy_vars): New functions.
1766 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1767 lower_omp_1, execute_lower_omp): Use them.
1769 PR rtl-optimization/84308
1770 * shrink-wrap.c (spread_components): Release todo vector.
1772 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1774 PR rtl-optimization/57193
1775 * ira-color.c (struct allocno_color_data): Add member
1776 conflict_allocno_hard_prefs.
1777 (update_conflict_allocno_hard_prefs): New.
1778 (bucket_allocno_compare_func): Add a preference based on
1779 conflict_allocno_hard_prefs.
1780 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1781 (color_allocnos): Remove a dead code. Initiate
1782 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1784 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1787 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1788 constraint from =wa to wa. Avoid a subreg on the output operand,
1789 instead use a pseudo and subreg it in a move.
1790 (p9_xxbrd_<mode>): Changed to ...
1791 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1792 (p9_xxbrd_v2df): New expander.
1793 (p9_xxbrw_<mode>): Changed to ...
1794 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1795 (p9_xxbrw_v4sf): New expander.
1797 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1799 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1801 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1804 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1805 multiply in 32-bit mode.
1806 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1807 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1810 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1812 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1813 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1814 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1815 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1817 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1820 * dwarf2out.c (is_trivial_indirect_ref): New function.
1821 (dwarf2out_late_global_decl): Do not generate a location
1822 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1823 and that are not defined in the current unit.
1825 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1827 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1828 instead of a libcall for UNORDERED.
1830 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1833 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1834 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1836 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1839 * config/s390/s390.c (s390_set_current_function): Invoke
1840 s390_indirect_branch_settings also if fndecl didn't change.
1842 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1844 * config/rs6000/rs6000.md (blockage): Set length to zero.
1846 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1848 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1850 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1853 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1854 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1858 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1859 PARALLEL incoming that failed vt_get_decl_and_offset check.
1862 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1863 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1865 (get_variable_section): For decls in named .bss* sections pass true as
1866 second argument to bss_initializer_p.
1868 2018-02-09 Marek Polacek <polacek@redhat.com>
1869 Jakub Jelinek <jakub@redhat.com>
1872 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1873 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1874 Sync some changes from cxx_fold_indirect_ref.
1876 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1878 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1880 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1881 (BLOCK_INLINE_ENTRY_LABEL): New.
1882 (dwarf2out_var_location): Disregard inline entry markers.
1883 (inline_entry_data): New struct.
1884 (inline_entry_data_hasher): New hashtable type.
1885 (inline_entry_data_hasher::hash): New.
1886 (inline_entry_data_hasher::equal): New.
1887 (inline_entry_data_table): New variable.
1888 (add_high_low_attributes): Add DW_AT_entry_pc and
1889 DW_AT_GNU_entry_view attributes if a pending entry is found
1890 in inline_entry_data_table. Add old entry_pc attribute only
1891 if debug nonbinding markers are disabled.
1892 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1893 markers are enabled.
1894 (block_within_block_p, dwarf2out_inline_entry): New.
1895 (dwarf2out_finish): Check that no entries remained in
1896 inline_entry_data_table.
1897 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1898 (final_scan_insn, notice_source_line): Likewise.
1899 (rest_of_clean_state): Skip inline entry markers.
1900 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1902 * gimple.c (gimple_build_debug_inline_entry): New.
1903 * gimple.h (enum gimple_debug_subcode): Add
1904 GIMPLE_DEBUG_INLINE_ENTRY.
1905 (gimple_build_debug_inline_entry): Declare.
1906 (gimple_debug_inline_entry_p): New.
1907 (gimple_debug_nonbind_marker_p): Adjust.
1908 * insn-notes.def (INLINE_ENTRY): New.
1909 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1910 inline entry marker notes.
1911 (print_insn): Likewise.
1912 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1913 (INSN_DEBUG_MARKER_KIND): Likewise.
1914 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1915 * tree-inline.c (expand_call_inline): Build and insert
1916 debug_inline_entry stmt.
1917 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1918 inline entry blocks early, if nonbind markers are enabled.
1919 (dump_scope_block): Dump fragment info.
1920 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1921 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1922 (gimple_build_debug_inline_entry): New.
1923 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1924 Enable/disable inline entry points too.
1925 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1926 (DEBUG_INSN): Describe inline entry markers.
1928 * common.opt (gvariable-location-views): New.
1929 (gvariable-location-views=incompat5): New.
1930 * config.in: Rebuilt.
1931 * configure: Rebuilt.
1932 * configure.ac: Test assembler for view support.
1933 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1934 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1935 * dwarf2out.c (var_loc_view): New typedef.
1936 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1937 (dwarf2out_locviews_in_attribute): New.
1938 (dwarf2out_locviews_in_loclist): New.
1939 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1940 (enum dw_line_info_opcode): Add LI_adv_address.
1941 (struct dw_line_info_table): Add view.
1942 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1943 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1944 (zero_view_p): New variable.
1945 (ZERO_VIEW_P): New macro.
1946 (output_asm_line_debug_info): New.
1947 (struct var_loc_node): Add view.
1948 (add_AT_view_list, AT_loc_list): New.
1949 (add_var_loc_to_decl): Add view param. Test it against last.
1950 (new_loc_list): Add view params. Record them.
1951 (AT_loc_list_ptr): Handle loc and view lists.
1952 (view_list_to_loc_list_val_node): New.
1953 (print_dw_val): Handle dw_val_class_view_list.
1954 (size_of_die): Likewise.
1955 (value_format): Likewise.
1956 (loc_list_has_views): New.
1957 (gen_llsym): Set vl_symbol too.
1958 (maybe_gen_llsym, skip_loc_list_entry): New.
1959 (dwarf2out_maybe_output_loclist_view_pair): New.
1960 (output_loc_list): Output view list or entries too.
1961 (output_view_list_offset): New.
1962 (output_die): Handle dw_val_class_view_list.
1963 (output_dwarf_version): New.
1964 (output_compilation_unit_header): Use it.
1965 (output_skeleton_debug_sections): Likewise.
1966 (output_rnglists, output_line_info): Likewise.
1967 (output_pubnames, output_aranges): Update version comments.
1968 (output_one_line_info_table): Output view numbers in asm comments.
1969 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1970 Call maybe_gen_llsym.
1971 (loc_list_from_tree_1): Adjust.
1972 (add_AT_location_description): Create view list attribute if
1973 needed, check it's absent otherwise.
1974 (convert_cfa_to_fb_loc_list): Adjust.
1975 (maybe_emit_file): Call output_asm_line_debug_info for test.
1976 (dwarf2out_var_location): Reset views as needed. Precompute
1977 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1978 attribute. Set view.
1979 (new_line_info_table): Reset next view.
1980 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1981 (dwarf2out_source_line): Likewise. Output view resets and labels to
1982 the assembler, or select appropriate line info opcodes.
1983 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1984 (optimize_string_length): Catch it. Adjust.
1985 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1986 dw_val_class_view_list, and remove it if no longer needed.
1987 (hash_loc_list): Hash view numbers.
1988 (loc_list_hasher::equal): Compare them.
1989 (optimize_location_lists): Check whether a view list symbol is
1990 needed, and whether the locview attribute is present, and
1991 whether they match. Remove the locview attribute if no longer
1993 (index_location_lists): Call skip_loc_list_entry for test.
1994 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1995 Use output_dwarf_version.
1996 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1997 (struct dw_val_node): Add val_view_list.
1998 * final.c (SEEN_NEXT_VIEW): New.
1999 (set_next_view_needed): New.
2000 (clear_next_view_needed): New.
2001 (maybe_output_next_view): New.
2002 (final_start_function): Rename to...
2003 (final_start_function_1): ... this. Take pointer to FIRST,
2004 add SEEN parameter. Emit param bindings in the initial view.
2005 (final_start_function): Reintroduce SEEN-less interface.
2006 (final): Rename to...
2007 (final_1): ... this. Take SEEN parameter. Output final pending
2008 next view at the end.
2009 (final): Reintroduce seen-less interface.
2010 (final_scan_insn): Output pending next view before switching
2011 sections or ending a block. Mark the next view as needed when
2012 outputting variable locations. Notify debug backend of section
2013 changes, and of location view changes.
2014 (rest_of_handle_final): Adjust.
2015 * toplev.c (process_options): Autodetect value for debug variable
2016 location views option. Warn on incompat5 without -gdwarf-5.
2017 * doc/invoke.texi (gvariable-location-views): New.
2018 (gvariable-location-views=incompat5): New.
2019 (gno-variable-location-views): New.
2021 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2023 PR tree-optimization/84136
2024 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2025 that the result of find_edge is non-NULL.
2027 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2030 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2031 storing integer register in SImode. Fix cost of 256 and 512
2032 byte aligned SSE register store.
2034 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2036 * config/i386/i386.c (ix86_multiplication_cost): Fix
2037 multiplication cost for TARGET_AVX512DQ.
2039 2018-02-08 Marek Polacek <polacek@redhat.com>
2041 PR tree-optimization/84238
2042 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2045 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2047 PR tree-optimization/84265
2048 * tree-vect-stmts.c (vectorizable_store): Don't treat
2049 VMAT_CONTIGUOUS accesses as grouped.
2050 (vectorizable_load): Likewise.
2052 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2054 PR tree-optimization/81635
2055 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2056 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2057 (test_round_for_mask): New functions.
2058 (wide_int_cc_tests): Call test_round_for_mask.
2059 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2060 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2061 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2062 range returned by get_range_info.
2064 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2067 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2068 * symtab.c: Include builtins.h
2069 (symtab_node::output_to_lto_symbol_table_p): Move here
2070 from lto-streamer-out.c:output_symbol_p.
2071 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2072 (output_symbol_p): Move all logic to symtab.c
2073 (produce_symtab): Update.
2075 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2077 * config/s390/s390-opts.h (enum indirect_branch): Define.
2078 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2079 (s390_indirect_branch_via_thunk)
2080 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2081 (enum s390_indirect_branch_type): Define.
2082 * config/s390/s390.c (struct s390_frame_layout, struct
2083 machine_function): Remove.
2084 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2085 (indirect_branch_table_label_no, indirect_branch_table_name):
2087 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2088 (enum s390_indirect_branch_option): Define.
2089 (s390_return_addr_from_memory): New function.
2090 (s390_handle_string_attribute): New function.
2091 (s390_attribute_table): Add new attribute handler.
2092 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2093 (s390_indirect_branch_via_thunk): New function.
2094 (s390_indirect_branch_via_inline_thunk): New function.
2095 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2096 sibling call optimization for non z10 compiles.
2097 (s390_emit_call): Force indirect branch target to be a single
2098 register. Add r1 clobber for non-z10 compiles.
2099 (s390_emit_epilogue): Emit return jump via return_use expander.
2100 (s390_reorg): Handle JUMP_INSNs as execute targets.
2101 (s390_option_override_internal): Perform validity checks for the
2102 new command line options.
2103 (s390_indirect_branch_attrvalue): New function.
2104 (s390_indirect_branch_settings): New function.
2105 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2106 (s390_output_indirect_thunk_function): New function.
2107 (s390_code_end): Implement target hook.
2108 (s390_case_values_threshold): Implement target hook.
2109 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2111 * config/s390/s390.h (struct s390_frame_layout)
2112 (struct machine_function): Move here from s390.c.
2113 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2114 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2115 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2116 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2117 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2118 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2119 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2120 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2121 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2122 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2123 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2124 (mnemonic attribute): Add values which aren't recognized
2126 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2127 pattern for branch conversion. Fix mnemonic attribute.
2128 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2129 indirect branch via thunk if requested.
2130 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2131 ("*indirect_jump"): Disable for branch conversion using out of
2133 ("indirect_jump_via_thunk<mode>_z10")
2134 ("indirect_jump_via_thunk<mode>")
2135 ("indirect_jump_via_inlinethunk<mode>_z10")
2136 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2137 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2138 ("casesi_jump_via_inlinethunk<mode>_z10")
2139 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2140 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2141 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2142 ("*indirect2_jump"): Disable for branch conversion.
2143 ("casesi_jump"): Turn into expander and expand patterns for branch
2145 ("return_use"): New expander.
2146 ("*return"): Emit return via thunk and rename it to ...
2147 ("*return<mode>"): ... this one.
2148 * config/s390/s390.opt: Add new options and and enum for the
2151 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2153 * lra-constraints.c (match_reload): Unconditionally use
2154 gen_lowpart_SUBREG, rather than selecting between that
2155 and equivalent gen_rtx_SUBREG code.
2157 2018-02-08 Richard Biener <rguenther@suse.de>
2159 PR tree-optimization/84233
2160 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2161 changed flag instead of boguously re-using phi_inserted.
2163 2018-02-08 Martin Jambor <mjambor@suse.cz>
2165 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2166 static local variables.
2168 2018-02-08 Richard Biener <rguenther@suse.de>
2170 PR tree-optimization/84278
2171 * tree-vect-stmts.c (vectorizable_store): When looking for
2172 smaller vector types to perform grouped strided loads/stores
2173 make sure the mode is supported by the target.
2174 (vectorizable_load): Likewise.
2176 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2178 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2179 Increase LDP/STP opportunities by adding adjacent callee-saves.
2181 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2183 PR rtl-optimization/84068
2184 PR rtl-optimization/83459
2185 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2187 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2189 PR tree-optimization/84224
2190 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2191 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2194 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2197 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2198 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2199 count, remove one USE.
2201 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2203 * doc/install.texi (Configuration): Document the
2204 --with-long-double-format={ibm,ieee} PowerPC configuration
2208 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2209 Convert from define_expand to be define_insn_and_split. Rework
2210 float/double/_Float128 conversions to QI/HI/SImode to work with
2211 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2212 conversions to QI/HImode types did a store and then a load to
2213 truncate the value. For conversions to VSX registers, don't split
2214 the insn, instead emit the code directly. Use the code iterator
2215 any_fix to combine signed and unsigned conversions.
2216 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2217 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2218 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2219 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2220 (fix_<mode>di2_hw): Likewise.
2221 (fixuns_<mode>di2_hw): Likewise.
2222 (fix_<mode>si2_hw): Likewise.
2223 (fixuns_<mode>si2_hw): Likewise.
2224 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2225 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2226 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2227 fix<uns>_trunc<SFDF:mode>si2_p8.
2228 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2230 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2231 (fix<uns>_<mode>_mem): Likewise.
2232 (fctiw<u>z_<mode>_mem): Likewise.
2233 (fix<uns>_<mode>_mem): Likewise.
2234 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2235 the register allocator from doing a direct move to the GPRs to do
2236 a store, and instead use the ISA 3.0 store byte/half-word from
2237 vector register instruction. For IEEE 128-bit floating point,
2238 also optimize stores of 32-bit ints.
2239 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2241 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2243 * genextract.c (push_pathstr_operand): New function to support
2245 (walk_rtx): Call push_pathstr_operand.
2246 (print_path): Support [a-zA-Z].
2248 2018-02-07 Richard Biener <rguenther@suse.de>
2250 PR tree-optimization/84037
2251 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2252 (cse_and_gimplify_to_preheader): Declare.
2253 (vect_get_place_in_interleaving_chain): Likewise.
2254 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2256 (_loop_vec_info::~_loop_vec_info): Delete it.
2257 (cse_and_gimplify_to_preheader): New function.
2258 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2259 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2260 (vectorizable_load): Likewise. For grouped stores always base
2261 the IV on the first element.
2262 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2263 condition before gimplifying.
2265 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2267 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2268 *DIV_EXPR and *MOD_EXPR.
2270 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2273 * config/i386/i386.c (ix86_option_override_internal): Mask out
2274 the CF_SET bit when checking -fcf-protection.
2276 2018-02-07 Tom de Vries <tom@codesourcery.com>
2279 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2282 2018-02-07 Richard Biener <rguenther@suse.de>
2284 PR tree-optimization/84204
2285 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2288 PR tree-optimization/84205
2289 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2290 special-case isl_ast_op_zdiv_r.
2292 PR tree-optimization/84223
2293 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2294 Only add conditions from within the region.
2295 (gather_bbs::after_dom_children): Adjust.
2297 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2300 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2301 * config/avr/avr.md: Only post-reload split REG-REG moves if
2302 either register is GENERAL_REG_P.
2304 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2306 PR tree-optimization/84235
2307 * tree-ssa-scopedtables.c
2308 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2309 if the subtraction is performed in floating point type where NaNs are
2310 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2311 build 1. Formatting fix.
2313 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2316 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2317 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2318 and skip it regardless of bb boundaries. Use CALL_P macro,
2319 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2320 unnecessarily, formatting fix.
2322 2018-02-06 Michael Collison <michael.collison@arm.com>
2324 * config/arm/thumb2.md:
2325 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2326 (*thumb_mov_notscc): Ditto.
2328 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2331 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2332 unsigned_fix, not "s".
2334 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2336 * configure.ac (gcc_fn_eh_frame_ro): New function.
2337 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2338 correct .eh_frame permissions.
2339 * configure: Regenerate.
2341 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2343 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2346 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2348 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2349 Display warning message for -mno-speculate-indirect-jumps.
2351 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2353 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2355 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2357 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2359 PR tree-optimization/84225
2360 * tree-eh.c (find_trapping_overflow): Only call
2361 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2363 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2366 * config/i386/i386.c: Reimplement the check of possible options
2367 -mibt/-mshstk conbination. Change error messages.
2368 * doc/invoke.texi: Fix a typo: remove extra '='.
2370 2018-02-06 Marek Polacek <polacek@redhat.com>
2372 PR tree-optimization/84228
2373 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2375 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2378 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2379 emitted arch directives.
2380 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2381 __ARM_FEATURE_COPROC before changing architectures.
2383 2018-02-06 Richard Biener <rguenther@suse.de>
2385 * config/i386/i386.c (print_reg): Fix typo.
2386 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2388 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2390 * configure: Regenerate.
2392 2018-02-05 Martin Sebor <msebor@redhat.com>
2394 PR tree-optimization/83369
2395 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2398 2018-02-05 Martin Liska <mliska@suse.cz>
2400 * doc/invoke.texi: Cherry-pick upstream r323995.
2402 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2404 * ira.c (ira_init_register_move_cost): Adjust comment.
2406 2018-02-05 Martin Liska <mliska@suse.cz>
2408 PR gcov-profile/84137
2409 * doc/gcov.texi: Fix typo in documentation.
2411 2018-02-05 Martin Liska <mliska@suse.cz>
2413 PR gcov-profile/83879
2414 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2415 using dlopen functionality.
2417 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2419 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2420 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2421 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2422 _mm_maskz_range_round_ss): New intrinsics.
2423 (__builtin_ia32_rangesd128_round)
2424 (__builtin_ia32_rangess128_round): Remove.
2425 (__builtin_ia32_rangesd128_mask_round,
2426 __builtin_ia32_rangess128_mask_round): New builtins.
2427 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2428 __builtin_ia32_rangess128_round): Remove.
2429 (__builtin_ia32_rangesd128_mask_round,
2430 __builtin_ia32_rangess128_mask_round): New builtins.
2431 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2432 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2433 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2434 "<round_saeonly_constraint>")): Changed to ...
2435 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2436 "<round_saeonly_scalar_constraint>")): ... this.
2437 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2438 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2439 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2440 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2441 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2443 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2445 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2447 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2448 Remove all values except native, 8540 and 8548.
2450 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2452 * config/i386/i386.c (ix86_output_function_return): Pass
2453 INVALID_REGNUM, instead of -1, as invalid register number to
2454 indirect_thunk_name and output_indirect_thunk.
2456 2018-02-02 Julia Koval <julia.koval@intel.com>
2458 * config.gcc: Add -march=icelake.
2459 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2460 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2461 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2462 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2463 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2464 (processor_target_table): Add icelake.
2465 (ix86_option_override_internal): Handle new PTAs.
2466 (get_builtin_code_for_version): Handle icelake.
2467 (M_INTEL_COREI7_ICELAKE): New.
2468 (fold_builtin_cpu): Handle icelake.
2469 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2470 * doc/invoke.texi: Add -march=icelake.
2472 2018-02-02 Julia Koval <julia.koval@intel.com>
2474 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2475 to wide_int_bitmask.
2476 * wide-int-bitmask.h: New.
2478 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2481 * config/i386/i386.md: Replace Pmode with word_mode in
2482 builtin_setjmp_setup and builtin_longjmp to support x32.
2484 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2488 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2490 (rs6000_supported_cpu_names): New static variable.
2491 (linux_cpu_translation_table): Likewise.
2492 (elf_platform) <cpu>: Define new static variable and use it.
2493 Translate kernel AT_PLATFORM name to canonical name if needed.
2494 Error if platform name is unknown.
2496 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2499 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2501 2018-02-01 Jeff Law <law@redhat.com>
2504 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2505 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2506 the scratch if RELEASE_VIA_POP is false.
2507 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2508 If we have to save a temporary register, decrement SIZE appropriately.
2509 Pass new arguments to release_scratch_register_on_entry.
2510 (ix86_adjust_stack_and_probe): Likewise.
2511 (ix86_emit_probe_stack_range): Pass new arguments to
2512 release_scratch_register_on_entry.
2514 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2516 PR rtl-optimization/84157
2517 * combine.c (change_zero_ext): Use REG_P predicate in
2518 front of HARD_REGISTER_P predicate.
2520 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2522 * config/avr/avr.c (avr_option_override): Move disabling of
2523 -fdelete-null-pointer-checks to...
2524 * common/config/avr/avr-common.c (avr_option_optimization_table):
2527 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2529 PR tree-optimization/81635
2530 * tree-data-ref.c (split_constant_offset_1): For types that
2531 wrap on overflow, try to use range info to prove that wrapping
2534 2018-02-01 Renlin Li <renlin.li@arm.com>
2537 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2539 (aarch64_register_move_cost): Likewise.
2540 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2542 (REG_CLASS_NAMES): Likewise.
2543 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2544 TAILCALL_ADDR_REGS. Remove IP registers.
2545 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2547 2018-02-01 Richard Biener <rguenther@suse.de>
2549 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2550 for specifying RPO order and allow NULL for that.
2551 * domwalk.c (dom_walker::dom_walker): Likewise.
2552 (dom_walker::walk): Handle NULL RPO order.
2553 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2555 (rewrite_update_dom_walker): Likewise.
2556 (mark_def_dom_walker): Likewise.
2558 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2560 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2561 (aarch64_maybe_expand_sve_subreg_move): Declare.
2562 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2563 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2565 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2566 that are semantically a reverse operation.
2567 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2568 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2569 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2571 (aarch64_can_change_mode_class): For big-endian, forbid changes
2572 between two SVE modes if they have different element sizes.
2574 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2576 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2577 the TImode handling for big-endian targets.
2579 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2581 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2582 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2584 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2585 Remove BSWAP handing for big-endian targets and use the form of
2586 LD1RQ appropariate for the mode.
2588 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2590 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2591 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2594 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2597 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2598 check for operands that need to go through aarch64_sve_reload_be.
2600 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2602 PR tree-optimization/81661
2603 PR tree-optimization/84117
2604 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2605 * tree-eh.c: Include gimplify.h.
2606 (find_trapping_overflow, replace_trapping_overflow,
2607 rewrite_to_non_trapping_overflow): New functions.
2608 * tree-vect-loop.c: Include tree-eh.h.
2609 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2610 * tree-data-ref.c: Include tree-eh.h.
2611 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2613 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2615 PR rtl-optimization/84123
2616 * combine.c (change_zero_ext): Check if hard register satisfies
2617 can_change_dest_mode before calling gen_lowpart_SUBREG.
2619 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2622 * ira.c (ira_init_register_move_cost): Remove assert.
2624 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2626 PR rtl-optimization/84071
2627 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2628 * doc/tm.texi: Regenerate.
2630 2018-01-31 Richard Biener <rguenther@suse.de>
2632 PR tree-optimization/84132
2633 * tree-data-ref.c (analyze_miv_subscript): Properly
2634 check whether evolution_function_is_affine_multivariate_p
2635 before calling gcd_of_steps_may_divide_p.
2637 2018-01-31 Julia Koval <julia.koval@intel.com>
2640 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2641 * config/i386/i386.md (rdpid_rex64) New.
2642 (rdpid): Make 32bit only.
2644 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2647 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2648 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2650 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2653 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2655 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2657 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2659 PR rtl-optimization/84071
2660 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2661 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2663 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2665 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2666 (arc_attribute_table): Add 'aux' attribute.
2667 (arc_in_small_data_p): Consider aux like variables.
2668 (arc_is_aux_reg_p): New function.
2669 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2670 (arc_get_aux_arg): New function.
2671 (prepare_move_operands): Handle aux-register access.
2672 (arc_handle_aux_attribute): New function.
2673 * doc/extend.texi (ARC Variable attributes): Add subsection.
2675 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2677 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2678 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2679 (arc_attribute_table): Add 'uncached' attribute.
2680 (arc_print_operand): Print '.di' flag for uncached memory
2682 (arc_in_small_data_p): Do not consider for small data the uncached
2684 (arc_is_uncached_mem_p): New function.
2685 * config/arc/predicates.md (compact_store_memory_operand): Check
2686 for uncached memory accesses.
2687 (nonvol_nonimm_operand): Likewise.
2688 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2690 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2693 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2694 falign-loops=): Add Optimization flag.
2696 2018-01-30 Jeff Law <law@redhat.com>
2699 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2700 INT_REGISTERS_SAVED. Check it prior to calling
2701 get_scratch_register_on_entry.
2702 (ix86_adjust_stack_and_probe): Similarly.
2703 (ix86_emit_probe_stack_range): Similarly.
2704 (ix86_expand_prologue): Corresponding changes.
2706 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2709 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2710 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2712 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2715 * lra-constraints.c (curr_insn_transform): Process AND in the
2718 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2720 PR rtl-optimization/83986
2721 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2722 dependence against last_pending_memory_flush in addition to
2725 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2727 PR tree-optimization/81611
2728 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2731 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2734 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2737 2018-01-30 Richard Biener <rguenther@suse.de>
2738 Jakub Jelinek <jakub@redhat.com>
2740 PR tree-optimization/84111
2741 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2742 inner loops added during recursion, as they don't have up-to-date
2745 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2748 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2749 (can_inline_edge_by_limits_p): ... here.
2750 (can_early_inline_edge_p, check_callers,
2751 update_caller_keys, update_callee_keys, recursive_inlining,
2752 add_new_edges_to_heap, speculation_useful_p,
2753 inline_small_functions,
2754 inline_small_functions, flatten_function,
2755 inline_to_all_callers_1): Update.
2757 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2759 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2762 2018-01-30 Richard Biener <rguenther@suse.de>
2764 PR tree-optimization/83008
2765 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2766 invariant and constant vector uses in stmts when they need
2769 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2772 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2773 * configure: Regenerate.
2775 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2777 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2779 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2780 Use gen_rtx_REG rather than gen_lowpart.
2782 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2784 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2785 rather than 0 when creating partial subregs.
2787 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2789 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2792 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2795 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2796 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2797 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2798 flags. This restores the settings used before the 2017-07-24.
2799 Turning off pre increment/decrement/modify allows IVOPTS to
2800 optimize DF/SF loops where the index is an int.
2802 2018-01-29 Richard Biener <rguenther@suse.de>
2803 Kelvin Nilsen <kelvin@gcc.gnu.org>
2806 * tree-vect-stmts.c (vectorizable_call): Don't call
2807 targetm.vectorize_builtin_md_vectorized_function if callee is
2810 2018-01-22 Carl Love <cel@us.ibm.com>
2812 * doc/extend.tex: Fix typo in second arg in
2813 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2815 2018-01-29 Richard Biener <rguenther@suse.de>
2817 PR tree-optimization/84086
2818 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2819 (flush_ssaname_freelist): When SSA names were released reset
2820 the SCEV hash table.
2822 2018-01-29 Richard Biener <rguenther@suse.de>
2824 PR tree-optimization/84057
2825 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2826 removed paths when removing edges.
2828 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2830 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2831 -mfunction-return=@var{choice}.
2833 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2836 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2838 (layout::print_source_line): Likewise.
2839 (test_get_line_width_without_trailing_whitespace): Add test cases.
2841 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2844 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2847 2018-01-26 Jim Wilson <jimw@sifive.com>
2849 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2851 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2854 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2856 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2857 and CMP + SUB-immediate -> SUBS.
2859 2018-01-26 Martin Sebor <msebor@redhat.com>
2861 PR tree-optimization/83896
2862 * tree-ssa-strlen.c (get_string_len): Rename...
2863 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2864 Avoid assuming length is constant.
2865 (handle_char_store): Use HOST_WIDE_INT for string length.
2867 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2870 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2871 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2873 2018-01-26 Richard Biener <rguenther@suse.de>
2875 PR rtl-optimization/84003
2876 * dse.c (record_store): Only record redundant stores when
2877 the earlier store aliases at least all accesses the later one does.
2879 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2881 PR rtl-optimization/83985
2882 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2883 REG_CFA_RESTORE insns.
2884 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2885 REG_CFA_RESTORE insns here.
2888 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2889 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2891 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2893 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2894 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2895 (arc_init): Likewise.
2896 (arc_override_options): Likewise.
2897 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2899 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2901 * config/arc/arc.h (TARGET_DBNZ): Define.
2902 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2903 properly set the tune attribute.
2904 (dbnz): Use TARGET_DBNZ guard.
2905 * config/arc/arc.opt (mtune): Add core3 option.
2907 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2909 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2910 recognize new pic like addresses.
2911 (arc_delegitimize_address): Clean up.
2913 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2915 * config/arc/arc-arches.def: Option mrf16 valid for all
2917 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2918 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2919 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2920 * config/arc/arc-tables.opt: Regenerate.
2921 * config/arc/arc.c (arc_conditional_register_usage): Handle
2922 reduced register file case.
2923 (arc_file_start): Set must have build attributes.
2924 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2926 * config/arc/arc.opt (mrf16): Add new option.
2927 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2928 * config/arc/genmultilib.awk: Handle new mrf16 option.
2929 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2930 * config/arc/t-multilib: Regenerate.
2931 * doc/invoke.texi (ARC Options): Document mrf16 option.
2933 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2935 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2936 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2937 (arc_attribute_table): Add 'secure_call' attribute.
2938 (arc_print_operand): Print secure call operand.
2939 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2941 (arc_is_secure_call_p): New function. * config/arc/arc.md
2942 (call_i): Add support for sjli instruction.
2943 (call_value_i): Likewise.
2944 * config/arc/constraints.md (Csc): New constraint.
2946 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2947 John Eric Martin <John.Martin@emmicro-us.com>
2949 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2950 * config/arc/arc.c (_arc_jli_section): New struct.
2951 (arc_jli_section): New type.
2952 (rc_jli_sections): New static variable.
2953 (arc_handle_jli_attribute): New function.
2954 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2955 (arc_file_end): New function.
2956 (TARGET_ASM_FILE_END): Define.
2957 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2958 (arc_add_jli_section): New function.
2959 (jli_call_scan): Likewise.
2960 (arc_reorg): Call jli_call_scan.
2961 (arc_output_addsi): Remove 'S' from printing asm operand.
2962 (arc_is_jli_call_p): New function.
2963 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2965 (movhi_insn): Likewise.
2966 (movsi_insn): Likewise.
2967 (movsi_set_cc_insn): Likewise.
2968 (loadqi_update): Likewise.
2969 (load_zeroextendqisi_update): Likewise.
2970 (load_signextendqisi_update): Likewise.
2971 (loadhi_update): Likewise.
2972 (load_zeroextendhisi_update): Likewise.
2973 (load_signextendhisi_update): Likewise.
2974 (loadsi_update): Likewise.
2975 (loadsf_update): Likewise.
2976 (movsicc_insn): Likewise.
2977 (bset_insn): Likewise.
2978 (bxor_insn): Likewise.
2979 (bclr_insn): Likewise.
2980 (bmsk_insn): Likewise.
2981 (bicsi3_insn): Likewise.
2982 (cmpsi_cc_c_insn): Likewise.
2983 (movsi_ne): Likewise.
2984 (movsi_cond_exec): Likewise.
2985 (clrsbsi2): Likewise.
2996 (call_i): Remove 'S' asm letter, add jli instruction.
2997 (call_value_i): Likewise.
2998 * config/arc/arc.op (mjli-always): New option.
2999 * config/arc/constraints.md (Cji): New constraint.
3000 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3002 (subsf3_fpx): Likewise.
3003 (mulsf3_fpx): Likewise.
3004 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3006 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3008 * doc/invoke.texi (ARC): Document mjli-always option.
3010 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3012 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3013 avoid addition with 0 and use incw and decw where possible.
3015 2018-01-26 Richard Biener <rguenther@suse.de>
3017 PR tree-optimization/81082
3018 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3019 association if it requires casting to unsigned.
3020 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3021 from fold_plusminus_mult_expr to catch important cases late when
3022 range info is available.
3024 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3026 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3027 * configure.ac (hidden_linkonce): New test.
3028 * configure: Regenerate.
3029 * config.in: Regenerate.
3031 2018-01-26 Julia Koval <julia.koval@intel.com>
3033 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3034 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3035 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3036 _mm_mask_bitshuffle_epi64_mask): Fix type.
3037 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3038 USI_FTYPE_V4DI_V4DI_USI): Remove.
3039 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3040 __builtin_ia32_vpshufbitqmb256_mask,
3041 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3042 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3043 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3045 2018-01-26 Alan Modra <amodra@gmail.com>
3048 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3049 UNSPEC_VBPERMQ. Sort other unspecs.
3051 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3053 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3055 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3058 * predict.c (drop_profile): Do not push/pop cfun; update also
3060 (handle_missing_profiles): Fix logic looking for zero profiles.
3062 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3065 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3066 on functions with #pragma omp declare simd or functions with simd
3068 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3069 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3070 Remove trailing \n from warning_at calls.
3072 2018-01-25 Tom de Vries <tom@codesourcery.com>
3075 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3076 for neutered workers.
3078 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3081 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3082 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3084 2018-01-24 Jeff Law <law@redhat.com>
3087 * i386.c (get_probe_interval): Move to earlier point.
3088 (ix86_compute_frame_layout): If -fstack-clash-protection and
3089 the frame is larger than the probe interval, then use pushes
3090 to save registers rather than reg->mem moves.
3091 (ix86_expand_prologue): Remove conditional for int_registers_saved
3094 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3097 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3098 min/max for never referenced object.
3100 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3103 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3105 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3106 attributes from DECL_ATTRIBUTES (decl) without affecting
3107 DECL_ATTRIBUTES (current_function_decl).
3108 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3109 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3111 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3113 PR tree-optimization/83979
3114 * fold-const.c (fold_comparison): Use constant_boolean_node
3115 instead of boolean_{true,false}_node.
3117 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3119 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3122 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3124 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3125 Simplify the clause that sets the length attribute.
3126 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3127 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3128 clause that sets the length attribute.
3129 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3131 2018-01-24 Tom de Vries <tom@codesourcery.com>
3134 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3135 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3136 Add strict parameter.
3137 (prevent_branch_around_nothing): Insert dummy insn between branch to
3138 label and label with no ptx insn inbetween.
3139 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3141 2018-01-24 Tom de Vries <tom@codesourcery.com>
3144 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3145 for neutered threads in warp.
3146 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3148 2018-01-24 Richard Biener <rguenther@suse.de>
3150 PR tree-optimization/83176
3151 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3154 2018-01-24 Richard Biener <rguenther@suse.de>
3156 PR tree-optimization/82819
3157 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3158 code generating pluses that are no-ops in the target precision.
3160 2018-01-24 Richard Biener <rguenther@suse.de>
3163 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3165 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3167 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3168 to merge probabilities.
3169 * predict.c (probably_never_executed): Also mark as cold functions
3170 with global 0 profile and guessed local profile.
3171 * profile-count.c (profile_probability::combine_with_count): New
3173 * profile-count.h (profile_probability::operator*,
3174 profile_probability::operator*=, profile_probability::operator/,
3175 profile_probability::operator/=): Reduce precision to adjusted
3176 and set value to guessed on contradictory divisions.
3177 (profile_probability::combine_with_freq): Remove.
3178 (profile_probability::combine_wiht_count): Declare.
3179 (profile_count::force_nonzero):: Set to adjusted.
3180 (profile_count::probability_in):: Set quality to adjusted.
3181 * tree-ssa-tail-merge.c (replace_block_by): Use
3184 2018-01-23 Andrew Waterman <andrew@sifive.com>
3185 Jim Wilson <jimw@sifive.com>
3187 * config/riscv/riscv.c (riscv_stack_boundary): New.
3188 (riscv_option_override): Set riscv_stack_boundary. Handle
3189 riscv_preferred_stack_boundary_arg.
3190 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3191 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3192 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3193 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3194 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3195 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3197 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3200 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3201 of struct ix86_frame.
3202 (ix86_expand_epilogue): Likewise. Add a local variable for
3203 the reg_save_offset field in struct ix86_frame.
3205 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3207 PR tree-optimization/82604
3208 * tree-loop-distribution.c (enum partition_kind): New enum item
3209 PKIND_PARTIAL_MEMSET.
3210 (partition_builtin_p): Support above new enum item.
3211 (generate_code_for_partition): Ditto.
3212 (compute_access_range): Differentiate cases that equality can be
3213 proven at all loops, the innermost loops or no loops.
3214 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3215 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3216 (finalize_partitions, distribute_loop): Don't fuse partition of
3217 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3218 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3221 2018-01-23 Martin Liska <mliska@suse.cz>
3223 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3224 order to ignore the predictor.
3225 (PRED_POLYMORPHIC_CALL): Likewise.
3226 (PRED_RECURSIVE_CALL): Likewise.
3228 2018-01-23 Martin Liska <mliska@suse.cz>
3230 * tree-profile.c (tree_profiling): Print function header to
3231 aware reader which function we are working on.
3232 * value-prof.c (gimple_find_values_to_profile): Do not print
3233 not interesting value histograms.
3235 2018-01-23 Martin Liska <mliska@suse.cz>
3237 * profile-count.h (enum profile_quality): Add
3238 profile_uninitialized as the first value. Do not number values
3239 as they are zero based.
3240 (profile_count::verify): Update sanity check.
3241 (profile_probability::verify): Likewise.
3243 2018-01-23 Nathan Sidwell <nathan@acm.org>
3245 * doc/invoke.texi (ffor-scope): Deprecate.
3247 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3249 PR tree-optimization/83510
3250 * domwalk.c (set_all_edges_as_executable): New function.
3251 (dom_walker::dom_walker): Convert bool param
3252 "skip_unreachable_blocks" to enum reachability. Move setup of
3253 edge flags to set_all_edges_as_executable and only do it when
3254 reachability is REACHABLE_BLOCKS.
3255 * domwalk.h (enum dom_walker::reachability): New enum.
3256 (dom_walker::dom_walker): Convert bool param
3257 "skip_unreachable_blocks" to enum reachability.
3258 (set_all_edges_as_executable): New decl.
3259 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3260 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3262 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3263 but converting true to REACHABLE_BLOCKS.
3264 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3266 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3267 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3268 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3270 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3271 if check_all_array_refs will be called.
3273 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3275 * tree.c (selftest::test_location_wrappers): Add more test
3278 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3280 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3281 (selftest::test_bit_in_range): Likewise.
3283 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3286 * doc/sourcebuild.texi (vect_float): Say that the selector
3287 only describes the situation when -funsafe-math-optimizations is on.
3288 (vect_float_strict): Document.
3290 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3292 PR tree-optimization/83965
3293 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3294 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3295 instead of checking only for a reduction.
3296 (vect_recog_widen_sum_pattern): Likewise.
3298 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3300 * predict.c (probably_never_executed): Only use precise profile info.
3301 (compute_function_frequency): Skip after inlining hack since we now
3302 have quality checking.
3304 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3306 * profile-count.h (profile_probability::very_unlikely,
3307 profile_probability::unlikely, profile_probability::even): Set
3308 precision to guessed.
3310 2018-01-23 Richard Biener <rguenther@suse.de>
3312 PR tree-optimization/83963
3313 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3314 Properly terminate dominator walk when crossing the exit edge not
3315 when visiting its source block.
3317 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3320 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3321 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3323 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3325 PR tree-optimization/83957
3326 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3327 semicolon after for body surrounded by braces.
3329 PR tree-optimization/83081
3330 * profile-count.h (profile_probability::split): New method.
3331 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3332 Use profile_probability::split.
3333 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3334 when splitting a single conditional jump into 2.
3336 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3338 PR tree-optimization/69452
3339 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3342 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3344 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3345 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3346 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3348 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3350 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3351 * config/rl78/rl78.md (movdi): New define_expand.
3352 * config/rl78/rl78.c (rl78_split_movdi): New function.
3354 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3357 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3359 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3360 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3361 128-bit to produce an UNSPEC move to get the double word with the
3362 signbit and then a shift directly to do signbit.
3363 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3364 implementation with a new version that just does either a direct
3365 move or a regular move. Move memory interface to separate insns.
3366 Move insns so they are next to the expander.
3367 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3368 with signbit move. Split big and little endian case.
3369 (signbit<mode>2_dm_mem_le): Likewise.
3370 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3371 (signbit<mode>2_dm2): Likewise.
3373 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3375 * config/rl78/rl78.md (anddi3): New define_expand.
3377 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3379 * config/rl78/rl78.md (umindi3): New define_expand.
3381 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3383 * config/rl78/rl78.md (smindi3): New define_expand.
3385 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3387 * config/rl78/rl78.md (smaxdi3): New define_expand.
3389 2018-01-22 Carl Love <cel@us.ibm.com>
3391 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3392 LVX_V1TI): Add macro expansion.
3393 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3394 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3395 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3396 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3397 Change check to determine if the instruction is a byte reversing
3398 entry. Fix typo in comment.
3399 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3400 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3401 Add def_builtin calls for new builtins.
3402 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3403 Add define_insn expansion.
3405 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3407 * config/rl78/rl78.md (umaxdi3): New define_expand.
3409 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3411 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3412 for non-QImode registers.
3414 2018-01-22 Richard Biener <rguenther@suse.de>
3416 PR tree-optimization/83963
3417 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3418 including the loop exit block.
3419 (scop_detection::merge_sese): Likewise.
3420 (scop_detection::add_scop): Do it here instead.
3422 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3424 * doc/sourcebuild.texi (arm_softfloat): Document.
3426 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3429 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3430 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3431 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3433 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3434 David Edelsohn <dje.gcc@gmail.com>
3437 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3438 Change "crset eq" to "crset 2".
3439 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3440 (*call_indirect_aix<mode>_nospec): Likewise.
3441 (*call_value_indirect_aix<mode>_nospec): Likewise.
3442 (*call_indirect_elfv2<mode>_nospec): Likewise.
3443 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3444 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3445 change assembly output from . to $.
3446 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3447 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3448 (*tablejump<mode>_internal1_nospec): Likewise.
3450 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3453 * config/sh/sh_optimize_sett_clrt.cc:
3454 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3456 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3458 PR tree-optimization/83940
3459 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3460 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3461 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3462 use it to pass back the definition type.
3463 (vect_check_store_rhs): Likewise rhs_dt_out.
3464 (vect_build_gather_load_calls): Add a mask_dt argument and use
3465 it instead of a call to vect_is_simple_use.
3466 (vectorizable_store): Update calls to vect_check_load_store_mask
3467 and vect_check_store_rhs. Use the dt returned by the latter instead
3468 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3469 instead of calls to vect_is_simple_use. Pass the scalar rather
3470 than the vector operand to vect_is_simple_use when handling
3471 second and subsequent copies of an rhs value.
3472 (vectorizable_load): Update calls to vect_check_load_store_mask
3473 and vect_build_gather_load_calls. Use the cached mask_dt and
3474 gs_info.offset_dt instead of calls to vect_is_simple_use.
3476 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3479 * tree-emutls.c: Include gimplify.h.
3480 (lower_emutls_2): New function.
3481 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3482 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3483 it before further processing.
3486 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3487 UINTVAL (trueop1) instead of INTVAL (op1).
3489 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3493 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3494 INCOMING_FRAME_SP_OFFSET if not defined.
3495 (scan_trace): Add ENTRY argument. If true and
3496 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3497 emit a note to adjust the CFA offset.
3498 (create_cfi_notes): Adjust scan_trace callers.
3499 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3500 INCOMING_FRAME_SP_OFFSET in the CIE.
3501 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3502 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3504 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3505 * doc/tm.texi: Regenerated.
3507 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3509 PR rtl-optimization/83147
3510 * lra-constraints.c (remove_inheritance_pseudos): Use
3511 lra_substitute_pseudo_within_insn.
3513 2018-01-19 Tom de Vries <tom@codesourcery.com>
3514 Cesar Philippidis <cesar@codesourcery.com>
3517 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3519 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3522 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3523 spaces for function labels.
3525 2018-01-19 Martin Liska <mliska@suse.cz>
3527 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3528 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3529 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3530 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3531 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3532 (PRED_CONST_RETURN): Change from 69 to 65.
3533 (PRED_NULL_RETURN): Change from 91 to 71.
3534 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3535 (PRED_LOOP_GUARD): Change from 66 to 73.
3537 2018-01-19 Martin Liska <mliska@suse.cz>
3539 * predict.c (predict_insn_def): Add new assert.
3540 (struct branch_predictor): Change type to signed integer.
3541 (test_prediction_value_range): Amend test to cover
3543 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3544 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3545 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3546 (PRED_LOOP_IV_COMPARE): Likewise.
3547 * predict.h (PROB_UNINITIALIZED): Define new constant.
3549 2018-01-19 Martin Liska <mliska@suse.cz>
3551 * predict.c (dump_prediction): Add new format for
3552 analyze_brprob.py script which is enabled with -details
3554 * profile-count.h (precise_p): New function.
3556 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3558 PR tree-optimization/83922
3559 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3560 there are no statements that need masking.
3561 (vect_active_double_reduction_p): New function.
3562 (vect_analyze_loop_operations): Use it when handling phis that
3563 are not in the loop header.
3565 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3567 PR tree-optimization/83914
3568 * tree-vect-loop.c (vectorizable_induction): Don't convert
3569 init_expr or apply the peeling adjustment for inductions
3570 that are nested within the vectorized loop.
3572 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3574 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3577 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3581 * function.h (gimplify_parameters): Add gimple_seq * argument.
3582 * function.c: Include gimple.h and options.h.
3583 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3584 for the added local temporaries if needed.
3585 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3586 if there are any parameter cleanups, wrap whole body into a
3587 try/finally with the cleanups.
3589 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3592 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3593 Use GET_MODE_CLASS for scalar floating point.
3595 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3599 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3600 Fix call of call_cgraph_insertion_hooks.
3602 2018-01-18 Martin Sebor <msebor@redhat.com>
3604 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3606 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3609 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3612 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3615 * common.opt: (-ffile-prefix-map): New option.
3616 * opts.c (common_handle_option): Defer it.
3617 * opts-global.c (handle_common_deferred_options): Handle it.
3618 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3619 * file-prefix-map.h: New file.
3620 (remap_debug_filename, add_debug_prefix_map): ...here.
3621 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3622 * final.c (debug_prefix_map, add_debug_prefix_map
3623 remap_debug_filename): Move to...
3624 * file-prefix-map.c: New file.
3625 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3626 generalize, get rid of alloca(), use strrchr() instead of strchr().
3627 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3628 Implement in terms of add_prefix_map().
3629 (remap_macro_filename, remap_debug_filename): Implement in term of
3631 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3632 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3633 * dbxout.c: Include file-prefix-map.h.
3634 * varasm.c: Likewise.
3635 * vmsdbgout.c: Likewise.
3636 * xcoffout.c: Likewise.
3637 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3638 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3639 * doc/invoke.texi (-ffile-prefix-map): Document.
3640 (-fdebug-prefix-map): Update description.
3642 2018-01-18 Martin Liska <mliska@suse.cz>
3644 * config/i386/i386.c (indirect_thunk_name): Document that also
3646 (output_indirect_thunk): Document why both instructions
3647 (pause and lfence) are generated.
3649 2018-01-18 Richard Biener <rguenther@suse.de>
3651 PR tree-optimization/83887
3652 * graphite-scop-detection.c
3653 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3654 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3655 (scop_detection::merge_sese): Re-implement with a flood-fill
3656 algorithm that properly finds a SESE region if it exists.
3658 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3661 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3662 pointer_diff optimizations use view_convert instead of convert.
3664 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3666 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3667 Generate different code for -mno-speculate-indirect-jumps.
3668 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3669 (*call_indirect_aix<mode>): Disable for
3670 -mno-speculate-indirect-jumps.
3671 (*call_indirect_aix<mode>_nospec): New define_insn.
3672 (*call_value_indirect_aix<mode>): Disable for
3673 -mno-speculate-indirect-jumps.
3674 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3675 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3676 -mno-speculate-indirect-jumps.
3677 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3679 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3681 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3682 long double type, set the flags for noting the default long double
3683 type, even if we don't pass or return a long double type.
3685 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3688 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3691 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3693 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3694 support for merge[hl].
3695 (fold_mergehl_helper): New helper function.
3696 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3697 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3698 (altivec_vmrglw_direct): Add xxmrglw insn.
3700 2018-01-17 Andrew Waterman <andrew@sifive.com>
3702 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3703 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3705 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3708 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3709 call the lto_location_cache before reading the
3710 DECL_SOURCE_LOCATION of the types.
3712 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3713 Richard Sandiford <richard.sandiford@linaro.org>
3715 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3716 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3717 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3718 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3719 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3721 * config/aarch64/constraints.md (aarch64_movti_operand):
3723 * config/aarch64/predicates.md (Uti): Add new constraint.
3725 2018-01-17 Carl Love <cel@us.ibm.com>
3727 * config/rs6000/vsx.md (define_expand xl_len_r,
3728 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3729 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3731 (define_expand, define_insn): Move the shift left from the
3732 define_insn to the define_expand for lxvl and stxvl instructions.
3733 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3734 and XL_LEN_R definitions to PURE.
3736 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3738 * config/i386/i386.c (indirect_thunk_name): Declare regno
3739 as unsigned int. Compare regno with INVALID_REGNUM.
3740 (output_indirect_thunk): Ditto.
3741 (output_indirect_thunk_function): Ditto.
3742 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3743 in the call to output_indirect_thunk_function.
3745 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3748 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3749 rather than the size of inner_type to determine the stack slot size
3750 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3752 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3755 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3758 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3760 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3761 endian Linux systems to optionally enable multilibs for selecting
3762 the long double type if the user configured an explicit type.
3763 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3764 have no long double multilibs if not defined.
3765 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3766 warn if the user used -mabi={ieee,ibm}longdouble and we built
3767 multilibs for long double.
3768 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3769 appropriate multilib option.
3770 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3772 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3773 for building long double multilibs.
3774 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3776 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3778 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3781 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3783 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3786 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3789 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3792 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3794 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3795 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3797 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3799 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3800 different rtl trees depending on TARGET_64BIT.
3801 (rs6000_gen_lvx): Likewise.
3803 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3805 * config/visium/visium.md (nop): Tweak comment.
3806 (hazard_nop): Likewise.
3808 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3810 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3811 -mspeculate-indirect-jumps.
3812 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3813 for -mno-speculate-indirect-jumps.
3814 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3815 (*call_value_indirect_elfv2<mode>): Disable for
3816 -mno-speculate-indirect-jumps.
3817 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3818 (indirect_jump): Emit different RTL for
3819 -mno-speculate-indirect-jumps.
3820 (*indirect_jump<mode>): Disable for
3821 -mno-speculate-indirect-jumps.
3822 (*indirect_jump<mode>_nospec): New define_insn.
3823 (tablejump): Emit different RTL for
3824 -mno-speculate-indirect-jumps.
3825 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3826 (tablejumpsi_nospec): New define_expand.
3827 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3828 (tablejumpdi_nospec): New define_expand.
3829 (*tablejump<mode>_internal1): Disable for
3830 -mno-speculate-indirect-jumps.
3831 (*tablejump<mode>_internal1_nospec): New define_insn.
3832 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3835 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3837 * caller-save.c (insert_save): Drop unnecessary parameter. All
3840 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3841 Richard Biener <rguenth@suse.de>
3844 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3845 return early, inline manually is_gimple_sizepos. Make sure if we
3846 call gimplify_expr we don't end up with a gimple constant.
3847 * tree.c (variably_modified_type_p): Don't return true for
3848 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3849 * gimplify.h (is_gimple_sizepos): Remove.
3851 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3853 PR tree-optimization/83857
3854 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3855 vectorizable_live_operation for pure SLP statements.
3856 (vectorizable_live_operation): Handle PHIs.
3858 2018-01-16 Richard Biener <rguenther@suse.de>
3860 PR tree-optimization/83867
3861 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3862 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3864 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3867 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3868 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3869 If off is not INTEGER_CST, issue a may not be aligned warning
3870 rather than isn't aligned. Use isn%'t rather than isn't.
3871 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3873 <case MULT_EXPR>: Improve the case when bottom and one of the
3874 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3875 operand, in that case check if the other operand is multiple of
3876 bottom divided by the INTEGER_CST operand.
3878 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3881 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3882 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3883 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3884 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3885 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3886 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3887 (pa_function_arg_size): New function.
3889 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3891 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3892 in a separate statement.
3894 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3896 PR tree-optimization/83847
3897 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3898 group gathers and scatters.
3900 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3902 PR rtl-optimization/86620
3903 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3905 PR rtl-optimization/83213
3906 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3907 to last if both are JUMP_INSNs.
3909 PR tree-optimization/83843
3910 * gimple-ssa-store-merging.c
3911 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3912 store_immediate_info for bswap/nop orig_stores.
3914 2018-01-15 Andrew Waterman <andrew@sifive.com>
3916 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3918 <UDIV>: Increase cost if !TARGET_DIV.
3920 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3922 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3923 (define_attr "cr_logical_3op"): New.
3924 (cceq_ior_compare): Adjust.
3925 (cceq_ior_compare_complement): Adjust.
3926 (*cceq_rev_compare): Adjust.
3927 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3928 (is_cracked_insn): Adjust.
3929 (insn_must_be_first_in_group): Adjust.
3930 * config/rs6000/40x.md: Adjust.
3931 * config/rs6000/440.md: Adjust.
3932 * config/rs6000/476.md: Adjust.
3933 * config/rs6000/601.md: Adjust.
3934 * config/rs6000/603.md: Adjust.
3935 * config/rs6000/6xx.md: Adjust.
3936 * config/rs6000/7450.md: Adjust.
3937 * config/rs6000/7xx.md: Adjust.
3938 * config/rs6000/8540.md: Adjust.
3939 * config/rs6000/cell.md: Adjust.
3940 * config/rs6000/e300c2c3.md: Adjust.
3941 * config/rs6000/e500mc.md: Adjust.
3942 * config/rs6000/e500mc64.md: Adjust.
3943 * config/rs6000/e5500.md: Adjust.
3944 * config/rs6000/e6500.md: Adjust.
3945 * config/rs6000/mpc.md: Adjust.
3946 * config/rs6000/power4.md: Adjust.
3947 * config/rs6000/power5.md: Adjust.
3948 * config/rs6000/power6.md: Adjust.
3949 * config/rs6000/power7.md: Adjust.
3950 * config/rs6000/power8.md: Adjust.
3951 * config/rs6000/power9.md: Adjust.
3952 * config/rs6000/rs64.md: Adjust.
3953 * config/rs6000/titan.md: Adjust.
3955 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3957 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3958 ix86_indirect_branch_register logic.
3960 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3962 * config/i386/constraints.md (Bs): Update
3963 ix86_indirect_branch_register check. Don't check
3964 ix86_indirect_branch_register with GOT_memory_operand.
3966 * config/i386/predicates.md (GOT_memory_operand): Don't check
3967 ix86_indirect_branch_register here.
3968 (GOT32_symbol_operand): Likewise.
3970 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3972 * config/i386/predicates.md (constant_call_address_operand):
3973 Rewrite ix86_indirect_branch_register logic.
3974 (sibcall_insn_operand): Likewise.
3976 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3978 * config/i386/constraints.md (Bs): Replace
3979 ix86_indirect_branch_thunk_register with
3980 ix86_indirect_branch_register.
3982 * config/i386/i386.md (indirect_jump): Likewise.
3983 (tablejump): Likewise.
3984 (*sibcall_memory): Likewise.
3985 (*sibcall_value_memory): Likewise.
3986 Peepholes of indirect call and jump via memory: Likewise.
3987 * config/i386/i386.opt: Likewise.
3988 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3989 (GOT_memory_operand): Likewise.
3990 (call_insn_operand): Likewise.
3991 (sibcall_insn_operand): Likewise.
3992 (GOT32_symbol_operand): Likewise.
3994 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3997 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3998 type rather than type addr's type points to.
3999 (expand_omp_atomic_mutex): Likewise.
4000 (expand_omp_atomic): Likewise.
4002 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4005 * config/i386/i386.c (output_indirect_thunk_function): Use
4006 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4007 for __x86_return_thunk.
4009 2018-01-15 Richard Biener <rguenther@suse.de>
4012 * expmed.c (extract_bit_field_1): Fix typo.
4014 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4017 * config/arm/iterators.md (VF): New mode iterator.
4018 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4019 Remove integer-related logic from pattern.
4020 (neon_vabd<mode>_3): Likewise.
4022 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4025 * common.opt (fstrict-overflow): No longer an alias.
4026 (fwrapv-pointer): New option.
4027 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4028 also for pointer types based on flag_wrapv_pointer.
4029 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4030 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4031 opts->x_flag_wrapv got set.
4032 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4033 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4034 POINTER_TYPE_OVERFLOW_UNDEFINED.
4035 * match.pd: Likewise in address comparison pattern.
4036 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4038 2018-01-15 Richard Biener <rguenther@suse.de>
4041 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4042 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4043 Reset type names to their identifier if their TYPE_DECL doesn't
4044 have linkage (and thus is used for ODR and devirt).
4045 (save_debug_info_for_decl): Remove.
4046 (save_debug_info_for_type): Likewise.
4047 (add_tree_to_fld_list): Adjust.
4048 * tree-pretty-print.c (dump_generic_node): Make dumping of
4049 type names more robust.
4051 2018-01-15 Richard Biener <rguenther@suse.de>
4053 * BASE-VER: Bump to 8.0.1.
4055 2018-01-14 Martin Sebor <msebor@redhat.com>
4058 * builtins.c (check_access): Avoid warning when the no-warning bit
4061 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4063 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4064 * ira-color (allocno_hard_regs_compare): Likewise.
4066 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4069 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4070 Use .pushsection/.popsection.
4072 2018-01-14 Martin Sebor <msebor@redhat.com>
4075 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4077 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4079 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4080 entry from extra_headers.
4081 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4082 extra_headers, make the list bitwise identical to the i?86-*-* one.
4084 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4086 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4087 -mcmodel=large with -mindirect-branch=thunk,
4088 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4089 -mfunction-return=thunk-extern.
4090 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4091 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4092 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4094 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4096 * config/i386/i386.c (print_reg): Print the name of the full
4097 integer register without '%'.
4098 (ix86_print_operand): Handle 'V'.
4099 * doc/extend.texi: Document 'V' modifier.
4101 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4103 * config/i386/constraints.md (Bs): Disallow memory operand for
4104 -mindirect-branch-register.
4106 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4107 (GOT_memory_operand): Likewise.
4108 (call_insn_operand): Likewise.
4109 (sibcall_insn_operand): Likewise.
4110 (GOT32_symbol_operand): Likewise.
4111 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4112 for -mindirect-branch-register.
4113 (tablejump): Likewise.
4114 (*sibcall_memory): Likewise.
4115 (*sibcall_value_memory): Likewise.
4116 Disallow peepholes of indirect call and jump via memory for
4117 -mindirect-branch-register.
4118 (*call_pop): Replace m with Bw.
4119 (*call_value_pop): Likewise.
4120 (*sibcall_pop_memory): Replace m with Bs.
4121 * config/i386/i386.opt (mindirect-branch-register): New option.
4122 * doc/invoke.texi: Document -mindirect-branch-register option.
4124 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4126 * config/i386/i386-protos.h (ix86_output_function_return): New.
4127 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4128 set function_return_type.
4129 (indirect_thunk_name): Add ret_p to indicate thunk for function
4131 (output_indirect_thunk_function): Pass false to
4132 indirect_thunk_name.
4133 (ix86_output_indirect_branch_via_reg): Likewise.
4134 (ix86_output_indirect_branch_via_push): Likewise.
4135 (output_indirect_thunk_function): Create alias for function
4136 return thunk if regno < 0.
4137 (ix86_output_function_return): New function.
4138 (ix86_handle_fndecl_attribute): Handle function_return.
4139 (ix86_attribute_table): Add function_return.
4140 * config/i386/i386.h (machine_function): Add
4141 function_return_type.
4142 * config/i386/i386.md (simple_return_internal): Use
4143 ix86_output_function_return.
4144 (simple_return_internal_long): Likewise.
4145 * config/i386/i386.opt (mfunction-return=): New option.
4146 (indirect_branch): Mention -mfunction-return=.
4147 * doc/extend.texi: Document function_return function attribute.
4148 * doc/invoke.texi: Document -mfunction-return= option.
4150 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4152 * config/i386/i386-opts.h (indirect_branch): New.
4153 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4154 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4155 with local indirect jump when converting indirect call and jump.
4156 (ix86_set_indirect_branch_type): New.
4157 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4158 (indirectlabelno): New.
4159 (indirect_thunk_needed): Likewise.
4160 (indirect_thunk_bnd_needed): Likewise.
4161 (indirect_thunks_used): Likewise.
4162 (indirect_thunks_bnd_used): Likewise.
4163 (INDIRECT_LABEL): Likewise.
4164 (indirect_thunk_name): Likewise.
4165 (output_indirect_thunk): Likewise.
4166 (output_indirect_thunk_function): Likewise.
4167 (ix86_output_indirect_branch_via_reg): Likewise.
4168 (ix86_output_indirect_branch_via_push): Likewise.
4169 (ix86_output_indirect_branch): Likewise.
4170 (ix86_output_indirect_jmp): Likewise.
4171 (ix86_code_end): Call output_indirect_thunk_function if needed.
4172 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4174 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4175 (ix86_attribute_table): Add indirect_branch.
4176 * config/i386/i386.h (machine_function): Add indirect_branch_type
4177 and has_local_indirect_jump.
4178 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4180 (tablejump): Likewise.
4181 (*indirect_jump): Use ix86_output_indirect_jmp.
4182 (*tablejump_1): Likewise.
4183 (simple_return_indirect_internal): Likewise.
4184 * config/i386/i386.opt (mindirect-branch=): New option.
4185 (indirect_branch): New.
4188 (thunk-inline): Likewise.
4189 (thunk-extern): Likewise.
4190 * doc/extend.texi: Document indirect_branch function attribute.
4191 * doc/invoke.texi: Document -mindirect-branch= option.
4193 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4196 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4198 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4200 * ipa-inline.c (want_inline_small_function_p): Return false if
4201 inlining has already failed with CIF_FINAL_ERROR.
4202 (update_caller_keys): Call want_inline_small_function_p before
4204 (update_callee_keys): Likewise.
4206 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4208 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4210 (rs6000_quadword_masked_address_p): Likewise.
4211 (quad_aligned_load_p): Likewise.
4212 (quad_aligned_store_p): Likewise.
4213 (const_load_sequence_p): Add comment to describe the outer-most loop.
4214 (mimic_memory_attributes_and_flags): New function.
4215 (rs6000_gen_stvx): Likewise.
4216 (replace_swapped_aligned_store): Likewise.
4217 (rs6000_gen_lvx): Likewise.
4218 (replace_swapped_aligned_load): Likewise.
4219 (replace_swapped_load_constant): Capitalize argument name in
4220 comment describing this function.
4221 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4222 and stores that access quad-word aligned addresses and replace
4223 with stvx or lvx instructions when appropriate.
4224 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4225 New function prototype.
4226 (rs6000_quadword_masked_address_p): Likewise.
4227 (rs6000_gen_lvx): Likewise.
4228 (rs6000_gen_stvx): Likewise.
4229 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4230 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4231 when memory address is aligned.
4232 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4233 this split to select lvx instruction when memory address is aligned.
4234 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4235 instruction when memory address is aligned.
4236 (*vsx_le_perm_load_v16qi): Likewise.
4237 (four unnamed splitters): Modify to select the stvx instruction
4238 when memory is aligned.
4240 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4242 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4243 which appears in the queue multiple times.
4245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4246 Alan Hayward <alan.hayward@arm.com>
4247 David Sherwood <david.sherwood@arm.com>
4249 * tree-vectorizer.h (vec_lower_bound): New structure.
4250 (_loop_vec_info): Add check_nonzero and lower_bounds.
4251 (LOOP_VINFO_CHECK_NONZERO): New macro.
4252 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4253 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4254 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4255 fields. Make seg_len the distance travelled, not including the
4257 (dr_direction_indicator): Declare.
4258 (dr_zero_step_indicator): Likewise.
4259 (dr_known_forward_stride_p): Likewise.
4260 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4262 (runtime_alias_check_p): Allow runtime alias checks with
4264 (operator ==): Compare access_size and align.
4265 (prune_runtime_alias_test_list): Rework for new distinction between
4266 the access_size and seg_len.
4267 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4269 (get_segment_min_max): New function.
4270 (create_intersect_range_checks): Use it.
4271 (dr_step_indicator): New function.
4272 (dr_direction_indicator): Likewise.
4273 (dr_zero_step_indicator): Likewise.
4274 (dr_known_forward_stride_p): Likewise.
4275 * tree-loop-distribution.c (data_ref_segment_size): Return
4276 DR_STEP * (niters - 1).
4277 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4279 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4280 (vect_preserves_scalar_order_p): New function, split out from...
4281 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4282 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4283 (vect_vfa_access_size): New function.
4284 (vect_vfa_align): Likewise.
4285 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4286 (dump_lower_bound): New function.
4287 (vect_check_lower_bound): Likewise.
4288 (vect_small_gap_p): Likewise.
4289 (vectorizable_with_step_bound_p): Likewise.
4290 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4291 depencies if the vectorization factor is 1. Convert the checks
4292 for nonzero steps into checks on the bounds of DR_STEP. Try using
4293 a bunds check for variable steps if the minimum required step is
4294 relatively small. Update calls to the dr_with_seg_len
4295 constructor and to vect_compile_time_alias.
4296 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4298 (vect_loop_versioning): Call it.
4299 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4301 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4304 Alan Hayward <alan.hayward@arm.com>
4305 David Sherwood <david.sherwood@arm.com>
4307 * doc/sourcebuild.texi (vect_scatter_store): Document.
4308 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4310 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4312 * genopinit.c (main): Add supports_vec_scatter_store and
4313 supports_vec_scatter_store_cached to target_optabs.
4314 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4315 IFN_MASK_SCATTER_STORE.
4316 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4318 * internal-fn.h (internal_store_fn_p): Declare.
4319 (internal_fn_stored_value_index): Likewise.
4320 * internal-fn.c (scatter_store_direct): New macro.
4321 (expand_scatter_store_optab_fn): New function.
4322 (direct_scatter_store_optab_supported_p): New macro.
4323 (internal_store_fn_p): New function.
4324 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4325 IFN_MASK_SCATTER_STORE.
4326 (internal_fn_mask_index): Likewise.
4327 (internal_fn_stored_value_index): New function.
4328 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4330 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4331 * optabs-query.c (supports_vec_scatter_store_p): New function.
4332 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4333 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4334 true for scatter stores.
4335 (vect_gather_scatter_fn_p): Handle scatter stores too.
4336 (vect_check_gather_scatter): Consider using scatter stores if
4337 supports_vec_scatter_store_p.
4338 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4340 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4341 internal_fn_stored_value_index.
4342 (check_load_store_masking): Handle scatter stores too.
4343 (vect_get_store_rhs): Make public.
4344 (vectorizable_call): Use internal_store_fn_p.
4345 (vectorizable_store): Handle scatter store internal functions.
4346 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4347 when deciding whether the end of the group has been reached.
4348 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4349 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4350 (mask_scatter_store<mode>): New insns.
4352 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4353 Alan Hayward <alan.hayward@arm.com>
4354 David Sherwood <david.sherwood@arm.com>
4356 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4357 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4358 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4360 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4361 Use vect_truncate_gather_scatter_offset if we can't treat the
4362 operation as a normal gather load or scatter store.
4363 (get_group_load_store_type): Take the gather_scatter_info
4364 as argument. Try using a gather load or scatter store for
4365 single-element groups.
4366 (get_load_store_type): Update calls to get_group_load_store_type
4367 and vect_use_strided_gather_scatters_p.
4369 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4370 Alan Hayward <alan.hayward@arm.com>
4371 David Sherwood <david.sherwood@arm.com>
4373 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4374 optional tree argument.
4375 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4377 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4378 but continue to use the current value as a fallback.
4379 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4380 to compare the updates.
4381 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4382 (get_load_store_type): Use it when handling a strided access.
4383 (vect_get_strided_load_store_ops): New function.
4384 (vect_get_data_ptr_increment): Likewise.
4385 (vectorizable_load): Handle strided gather loads. Always pass
4386 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4388 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4389 Alan Hayward <alan.hayward@arm.com>
4390 David Sherwood <david.sherwood@arm.com>
4392 * doc/md.texi (gather_load@var{m}): Document.
4393 (mask_gather_load@var{m}): Likewise.
4394 * genopinit.c (main): Add supports_vec_gather_load and
4395 supports_vec_gather_load_cached to target_optabs.
4396 * optabs-tree.c (init_tree_optimization_optabs): Use
4397 ggc_cleared_alloc to allocate target_optabs.
4398 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4399 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4401 * internal-fn.h (internal_load_fn_p): Declare.
4402 (internal_gather_scatter_fn_p): Likewise.
4403 (internal_fn_mask_index): Likewise.
4404 (internal_gather_scatter_fn_supported_p): Likewise.
4405 * internal-fn.c (gather_load_direct): New macro.
4406 (expand_gather_load_optab_fn): New function.
4407 (direct_gather_load_optab_supported_p): New macro.
4408 (direct_internal_fn_optab): New function.
4409 (internal_load_fn_p): Likewise.
4410 (internal_gather_scatter_fn_p): Likewise.
4411 (internal_fn_mask_index): Likewise.
4412 (internal_gather_scatter_fn_supported_p): Likewise.
4413 * optabs-query.c (supports_at_least_one_mode_p): New function.
4414 (supports_vec_gather_load_p): Likewise.
4415 * optabs-query.h (supports_vec_gather_load_p): Declare.
4416 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4417 and memory_type field.
4418 (NUM_PATTERNS): Bump to 15.
4419 * tree-vect-data-refs.c: Include internal-fn.h.
4420 (vect_gather_scatter_fn_p): New function.
4421 (vect_describe_gather_scatter_call): Likewise.
4422 (vect_check_gather_scatter): Try using internal functions for
4423 gather loads. Recognize existing calls to a gather load function.
4424 (vect_analyze_data_refs): Consider using gather loads if
4425 supports_vec_gather_load_p.
4426 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4427 (vect_get_gather_scatter_offset_type): Likewise.
4428 (vect_convert_mask_for_vectype): Likewise.
4429 (vect_add_conversion_to_patterm): Likewise.
4430 (vect_try_gather_scatter_pattern): Likewise.
4431 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4432 (vect_vect_recog_func_ptrs): Add it.
4433 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4434 internal_fn_mask_index and internal_gather_scatter_fn_p.
4435 (check_load_store_masking): Take the gather_scatter_info as an
4436 argument and handle gather loads.
4437 (vect_get_gather_scatter_ops): New function.
4438 (vectorizable_call): Check internal_load_fn_p.
4439 (vectorizable_load): Likewise. Handle gather load internal
4441 (vectorizable_store): Update call to check_load_store_masking.
4442 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4443 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4444 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4445 (aarch64_gather_scale_operand_d): New predicates.
4446 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4447 (mask_gather_load<mode>): New insns.
4449 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4450 Alan Hayward <alan.hayward@arm.com>
4451 David Sherwood <david.sherwood@arm.com>
4453 * optabs.def (fold_left_plus_optab): New optab.
4454 * doc/md.texi (fold_left_plus_@var{m}): Document.
4455 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4456 * internal-fn.c (fold_left_direct): Define.
4457 (expand_fold_left_optab_fn): Likewise.
4458 (direct_fold_left_optab_supported_p): Likewise.
4459 * fold-const-call.c (fold_const_fold_left): New function.
4460 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4461 * tree-parloops.c (valid_reduction_p): New function.
4462 (gather_scalar_reductions): Use it.
4463 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4464 (vect_finish_replace_stmt): Declare.
4465 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4466 (needs_fold_left_reduction_p): New function, split out from...
4467 (vect_is_simple_reduction): ...here. Accept reductions that
4468 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4469 (vect_force_simple_reduction): Also store the reduction type in
4470 the assignment's STMT_VINFO_REDUC_TYPE.
4471 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4472 (merge_with_identity): New function.
4473 (vect_expand_fold_left): Likewise.
4474 (vectorize_fold_left_reduction): Likewise.
4475 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4476 scalar phi in place for it. Check for target support and reject
4477 cases that would reassociate the operation. Defer the transform
4478 phase to vectorize_fold_left_reduction.
4479 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4480 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4481 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4483 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4485 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4486 call to ifc_temp_var.
4488 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4489 Alan Hayward <alan.hayward@arm.com>
4490 David Sherwood <david.sherwood@arm.com>
4492 * target.def (legitimize_address_displacement): Take the original
4493 offset as a poly_int.
4494 * targhooks.h (default_legitimize_address_displacement): Update
4496 * targhooks.c (default_legitimize_address_displacement): Likewise.
4497 * doc/tm.texi: Regenerate.
4498 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4499 as an argument, moving assert of ad->disp == ad->disp_term to...
4500 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4501 Try calling targetm.legitimize_address_displacement before expanding
4502 the address rather than afterwards, and adjust for the new interface.
4503 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4504 Match the new hook interface. Handle SVE addresses.
4505 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4510 * Makefile.in (OBJS): Add early-remat.o.
4511 * target.def (select_early_remat_modes): New hook.
4512 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4513 * doc/tm.texi: Regenerate.
4514 * targhooks.h (default_select_early_remat_modes): Declare.
4515 * targhooks.c (default_select_early_remat_modes): New function.
4516 * timevar.def (TV_EARLY_REMAT): New timevar.
4517 * passes.def (pass_early_remat): New pass.
4518 * tree-pass.h (make_pass_early_remat): Declare.
4519 * early-remat.c: New file.
4520 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4522 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4525 Alan Hayward <alan.hayward@arm.com>
4526 David Sherwood <david.sherwood@arm.com>
4528 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4529 vfm1 with a bound_epilog parameter.
4530 (vect_do_peeling): Update calls accordingly, and move the prologue
4531 call earlier in the function. Treat the base bound_epilog as 0 for
4532 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4533 this base when peeling for gaps.
4534 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4535 with fully-masked loops.
4536 (vect_estimate_min_profitable_iters): Handle the single peeled
4537 iteration in that case.
4539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4540 Alan Hayward <alan.hayward@arm.com>
4541 David Sherwood <david.sherwood@arm.com>
4543 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4544 single-element interleaving even if the size is not a power of 2.
4545 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4546 accesses for single-element interleaving if the group size is
4549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4550 Alan Hayward <alan.hayward@arm.com>
4551 David Sherwood <david.sherwood@arm.com>
4553 * doc/md.texi (fold_extract_last_@var{m}): Document.
4554 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4555 * optabs.def (fold_extract_last_optab): New optab.
4556 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4557 * internal-fn.c (fold_extract_direct): New macro.
4558 (expand_fold_extract_optab_fn): Likewise.
4559 (direct_fold_extract_optab_supported_p): Likewise.
4560 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4561 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4562 EXTRACT_LAST_REDUCTION.
4563 (get_initial_def_for_reduction): Do not create an initial vector
4564 for EXTRACT_LAST_REDUCTION reductions.
4565 (vectorizable_reduction): Leave the scalar phi in place for
4566 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4567 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4568 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4569 transform phase to vectorizable_condition.
4570 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4572 (vect_finish_stmt_generation): ...here.
4573 (vect_finish_replace_stmt): New function.
4574 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4575 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4577 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4580 Alan Hayward <alan.hayward@arm.com>
4581 David Sherwood <david.sherwood@arm.com>
4583 * doc/md.texi (extract_last_@var{m}): Document.
4584 * optabs.def (extract_last_optab): New optab.
4585 * internal-fn.def (EXTRACT_LAST): New internal function.
4586 * internal-fn.c (cond_unary_direct): New macro.
4587 (expand_cond_unary_optab_fn): Likewise.
4588 (direct_cond_unary_optab_supported_p): Likewise.
4589 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4590 loops using EXTRACT_LAST.
4591 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4592 (extract_last_<mode>): ...this optab.
4593 (vec_extract<mode><Vel>): Update accordingly.
4595 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4596 Alan Hayward <alan.hayward@arm.com>
4597 David Sherwood <david.sherwood@arm.com>
4599 * target.def (empty_mask_is_expensive): New hook.
4600 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4601 * doc/tm.texi: Regenerate.
4602 * targhooks.h (default_empty_mask_is_expensive): Declare.
4603 * targhooks.c (default_empty_mask_is_expensive): New function.
4604 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4605 if the target says that empty masks are expensive.
4606 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4608 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4611 Alan Hayward <alan.hayward@arm.com>
4612 David Sherwood <david.sherwood@arm.com>
4614 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4615 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4616 (vect_use_loop_mask_for_alignment_p): New function.
4617 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4618 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4619 niters_skip argument. Make sure that the first niters_skip elements
4620 of the first iteration are inactive.
4621 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4622 Update call to vect_set_loop_masks_directly.
4623 (get_misalign_in_elems): New function, split out from...
4624 (vect_gen_prolog_loop_niters): ...here.
4625 (vect_update_init_of_dr): Take a code argument that specifies whether
4626 the adjustment should be added or subtracted.
4627 (vect_update_init_of_drs): Likewise.
4628 (vect_prepare_for_masked_peels): New function.
4629 (vect_do_peeling): Skip prologue peeling if we're using a mask
4630 instead. Update call to vect_update_inits_of_drs.
4631 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4633 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4634 alignment. Do not include the number of peeled iterations in
4635 the minimum threshold in that case.
4636 (vectorizable_induction): Adjust the start value down by
4637 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4638 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4639 Take the number of skipped iterations into account when calculating
4641 * tree-vect-stmts.c (vect_gen_while_not): New function.
4643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4644 Alan Hayward <alan.hayward@arm.com>
4645 David Sherwood <david.sherwood@arm.com>
4647 * doc/sourcebuild.texi (vect_fully_masked): Document.
4648 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4650 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4652 (vect_analyze_loop_2): ...here. Don't check the vectorization
4653 factor against the number of loop iterations if the loop is
4656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4657 Alan Hayward <alan.hayward@arm.com>
4658 David Sherwood <david.sherwood@arm.com>
4660 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4661 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4662 (dump_groups): Update accordingly.
4663 (iv_use::mem_type): New member variable.
4664 (address_p): New function.
4665 (record_use): Add a mem_type argument and initialize the new
4667 (record_group_use): Add a mem_type argument. Use address_p.
4668 Remove obsolete null checks of base_object. Update call to record_use.
4669 (find_interesting_uses_op): Update call to record_group_use.
4670 (find_interesting_uses_cond): Likewise.
4671 (find_interesting_uses_address): Likewise.
4672 (get_mem_type_for_internal_fn): New function.
4673 (find_address_like_use): Likewise.
4674 (find_interesting_uses_stmt): Try find_address_like_use before
4675 calling find_interesting_uses_op.
4676 (addr_offset_valid_p): Use the iv mem_type field as the type
4677 of the addressed memory.
4678 (add_autoinc_candidates): Likewise.
4679 (get_address_cost): Likewise.
4680 (split_small_address_groups_p): Use address_p.
4681 (split_address_groups): Likewise.
4682 (add_iv_candidate_for_use): Likewise.
4683 (autoinc_possible_for_pair): Likewise.
4684 (rewrite_groups): Likewise.
4685 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4686 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4687 (get_alias_ptr_type_for_ptr_address): New function.
4688 (rewrite_use_address): Rewrite address uses in calls that were
4689 identified by find_address_like_use.
4691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4692 Alan Hayward <alan.hayward@arm.com>
4693 David Sherwood <david.sherwood@arm.com>
4695 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4697 * gimple-expr.h (is_gimple_addressable: Likewise.
4698 * gimple-expr.c (is_gimple_address): Likewise.
4699 * internal-fn.c (expand_call_mem_ref): New function.
4700 (expand_mask_load_optab_fn): Use it.
4701 (expand_mask_store_optab_fn): Likewise.
4703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4704 Alan Hayward <alan.hayward@arm.com>
4705 David Sherwood <david.sherwood@arm.com>
4707 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4708 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4709 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4710 (cond_umax@var{mode}): Document.
4711 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4712 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4713 (cond_umin_optab, cond_umax_optab): New optabs.
4714 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4715 (COND_IOR, COND_XOR): New internal functions.
4716 * internal-fn.h (get_conditional_internal_fn): Declare.
4717 * internal-fn.c (cond_binary_direct): New macro.
4718 (expand_cond_binary_optab_fn): Likewise.
4719 (direct_cond_binary_optab_supported_p): Likewise.
4720 (get_conditional_internal_fn): New function.
4721 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4722 Cope with reduction statements that are vectorized as calls rather
4724 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4725 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4726 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4727 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4728 (UNSPEC_COND_EOR): New unspecs.
4729 (optab): Add mappings for them.
4730 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4731 (sve_int_op, sve_fp_op): New int attributes.
4733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4734 Alan Hayward <alan.hayward@arm.com>
4735 David Sherwood <david.sherwood@arm.com>
4737 * optabs.def (while_ult_optab): New optab.
4738 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4739 * internal-fn.def (WHILE_ULT): New internal function.
4740 * internal-fn.h (direct_internal_fn_supported_p): New override
4741 that takes two types as argument.
4742 * internal-fn.c (while_direct): New macro.
4743 (expand_while_optab_fn): New function.
4744 (convert_optab_supported_p): Likewise.
4745 (direct_while_optab_supported_p): New macro.
4746 * wide-int.h (wi::udiv_ceil): New function.
4747 * tree-vectorizer.h (rgroup_masks): New structure.
4748 (vec_loop_masks): New typedef.
4749 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4751 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4752 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4753 (vect_max_vf): New function.
4754 (slpeel_make_loop_iterate_ntimes): Delete.
4755 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4756 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4757 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4758 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4759 internal-fn.h, stor-layout.h and optabs-query.h.
4760 (vect_set_loop_mask): New function.
4761 (add_preheader_seq): Likewise.
4762 (add_header_seq): Likewise.
4763 (interleave_supported_p): Likewise.
4764 (vect_maybe_permute_loop_masks): Likewise.
4765 (vect_set_loop_masks_directly): Likewise.
4766 (vect_set_loop_condition_masked): Likewise.
4767 (vect_set_loop_condition_unmasked): New function, split out from
4768 slpeel_make_loop_iterate_ntimes.
4769 (slpeel_make_loop_iterate_ntimes): Rename to..
4770 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4771 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4772 (vect_do_peeling): Update call accordingly.
4773 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4775 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4776 mask_compare_type, can_fully_mask_p and fully_masked_p.
4777 (release_vec_loop_masks): New function.
4778 (_loop_vec_info): Use it to free the loop masks.
4779 (can_produce_all_loop_masks_p): New function.
4780 (vect_get_max_nscalars_per_iter): Likewise.
4781 (vect_verify_full_masking): Likewise.
4782 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4783 retries, and free the mask rgroups before retrying. Check loop-wide
4784 reasons for disallowing fully-masked loops. Make the final decision
4785 about whether use a fully-masked loop or not.
4786 (vect_estimate_min_profitable_iters): Do not assume that peeling
4787 for the number of iterations will be needed for fully-masked loops.
4788 (vectorizable_reduction): Disable fully-masked loops.
4789 (vectorizable_live_operation): Likewise.
4790 (vect_halve_mask_nunits): New function.
4791 (vect_double_mask_nunits): Likewise.
4792 (vect_record_loop_mask): Likewise.
4793 (vect_get_loop_mask): Likewise.
4794 (vect_transform_loop): Handle the case in which the final loop
4795 iteration might handle a partial vector. Call vect_set_loop_condition
4796 instead of slpeel_make_loop_iterate_ntimes.
4797 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4798 (check_load_store_masking): New function.
4799 (prepare_load_store_mask): Likewise.
4800 (vectorizable_store): Handle fully-masked loops.
4801 (vectorizable_load): Likewise.
4802 (supportable_widening_operation): Use vect_halve_mask_nunits for
4804 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4805 (vect_gen_while): New function.
4806 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4807 (aarch64_uqdec<mode>): New insn.
4809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4810 Alan Hayward <alan.hayward@arm.com>
4811 David Sherwood <david.sherwood@arm.com>
4813 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4814 (reduc_xor_scal_optab): New optabs.
4815 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4816 (reduc_xor_scal_@var{m}): Document.
4817 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4818 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4820 * fold-const-call.c (fold_const_call): Handle them.
4821 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4822 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4823 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4824 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4825 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4826 (UNSPEC_XORV): New unspecs.
4827 (optab): Add entries for them.
4828 (BITWISEV): New int iterator.
4829 (bit_reduc_op): New int attributes.
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 (vec_shl_insert_@var{m}): New optab.
4836 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4837 * optabs.def (vec_shl_insert_optab): New optab.
4838 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4839 (duplicate_and_interleave): Likewise.
4840 * tree-vect-loop.c: Include internal-fn.h.
4841 (neutral_op_for_slp_reduction): New function, split out from
4842 get_initial_defs_for_reduction.
4843 (get_initial_def_for_reduction): Handle option 2 for variable-length
4844 vectors by loading the neutral value into a vector and then shifting
4845 the initial value into element 0.
4846 (get_initial_defs_for_reduction): Replace the code argument with
4847 the neutral value calculated by neutral_op_for_slp_reduction.
4848 Use gimple_build_vector for constant-length vectors.
4849 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4850 but the first group_size elements have a neutral value.
4851 Use duplicate_and_interleave otherwise.
4852 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4853 Update call to get_initial_defs_for_reduction. Handle SLP
4854 reductions for variable-length vectors by creating one vector
4855 result for each scalar result, with the elements associated
4856 with other scalar results stubbed out with the neutral value.
4857 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4858 Require IFN_VEC_SHL_INSERT for double reductions on
4859 variable-length vectors, or SLP reductions that have
4860 a neutral value. Require can_duplicate_and_interleave_p
4861 support for variable-length unchained SLP reductions if there
4862 is no neutral value, such as for MIN/MAX reductions. Also require
4863 the number of vector elements to be a multiple of the number of
4864 SLP statements when doing variable-length unchained SLP reductions.
4865 Update call to vect_create_epilog_for_reduction.
4866 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4867 and remove initial values.
4868 (duplicate_and_interleave): Make public.
4869 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4870 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4873 Alan Hayward <alan.hayward@arm.com>
4874 David Sherwood <david.sherwood@arm.com>
4876 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4877 (can_duplicate_and_interleave_p): New function.
4878 (vect_get_and_check_slp_defs): Take the vector of statements
4879 rather than just the current one. Remove excess parentheses.
4880 Restriction rejectinon of vect_constant_def and vect_external_def
4881 for variable-length vectors to boolean types, or types for which
4882 can_duplicate_and_interleave_p is false.
4883 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4884 (duplicate_and_interleave): New function.
4885 (vect_get_constant_vectors): Use gimple_build_vector for
4886 constant-length vectors and suitable variable-length constant
4887 vectors. Use duplicate_and_interleave for other variable-length
4888 vectors. Don't defer the update when inserting new statements.
4890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4891 Alan Hayward <alan.hayward@arm.com>
4892 David Sherwood <david.sherwood@arm.com>
4894 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4895 min_profitable_iters doesn't go negative.
4897 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4898 Alan Hayward <alan.hayward@arm.com>
4899 David Sherwood <david.sherwood@arm.com>
4901 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4902 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4903 * optabs.def (vec_mask_load_lanes_optab): New optab.
4904 (vec_mask_store_lanes_optab): Likewise.
4905 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4906 (MASK_STORE_LANES): Likewise.
4907 * internal-fn.c (mask_load_lanes_direct): New macro.
4908 (mask_store_lanes_direct): Likewise.
4909 (expand_mask_load_optab_fn): Handle masked operations.
4910 (expand_mask_load_lanes_optab_fn): New macro.
4911 (expand_mask_store_optab_fn): Handle masked operations.
4912 (expand_mask_store_lanes_optab_fn): New macro.
4913 (direct_mask_load_lanes_optab_supported_p): Likewise.
4914 (direct_mask_store_lanes_optab_supported_p): Likewise.
4915 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4917 (vect_load_lanes_supported): Likewise.
4918 * tree-vect-data-refs.c (strip_conversion): New function.
4919 (can_group_stmts_p): Likewise.
4920 (vect_analyze_data_ref_accesses): Use it instead of checking
4921 for a pair of assignments.
4922 (vect_store_lanes_supported): Take a masked_p parameter.
4923 (vect_load_lanes_supported): Likewise.
4924 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4925 vect_store_lanes_supported and vect_load_lanes_supported.
4926 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4927 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4928 parameter. Don't allow gaps for masked accesses.
4929 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4930 and vect_load_lanes_supported.
4931 (get_load_store_type): Take a masked_p parameter and update
4932 call to get_group_load_store_type.
4933 (vectorizable_store): Update call to get_load_store_type.
4934 Handle IFN_MASK_STORE_LANES.
4935 (vectorizable_load): Update call to get_load_store_type.
4936 Handle IFN_MASK_LOAD_LANES.
4938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4939 Alan Hayward <alan.hayward@arm.com>
4940 David Sherwood <david.sherwood@arm.com>
4942 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4944 * config/aarch64/aarch64-protos.h
4945 (aarch64_sve_struct_memory_operand_p): Declare.
4946 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4947 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4948 (VPRED, vpred): Handle SVE structure modes.
4949 * config/aarch64/constraints.md (Utx): New constraint.
4950 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4951 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4952 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4953 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4954 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4955 structure modes. Split into pieces after RA.
4956 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4957 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4959 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4960 SVE structure modes.
4961 (aarch64_classify_address): Likewise.
4962 (sizetochar): Move earlier in file.
4963 (aarch64_print_operand): Handle SVE register lists.
4964 (aarch64_array_mode): New function.
4965 (aarch64_sve_struct_memory_operand_p): Likewise.
4966 (TARGET_ARRAY_MODE): Redefine.
4968 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4969 Alan Hayward <alan.hayward@arm.com>
4970 David Sherwood <david.sherwood@arm.com>
4972 * target.def (array_mode): New target hook.
4973 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4974 * doc/tm.texi: Regenerate.
4975 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4976 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4977 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4979 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4982 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4983 Alan Hayward <alan.hayward@arm.com>
4984 David Sherwood <david.sherwood@arm.com>
4986 * fold-const.c (fold_binary_loc): Check the argument types
4987 rather than the result type when testing for a vector operation.
4989 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4991 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4992 * doc/tm.texi: Regenerate.
4994 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4995 Alan Hayward <alan.hayward@arm.com>
4996 David Sherwood <david.sherwood@arm.com>
4998 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4999 (sve): Document new AArch64 extension.
5000 * doc/md.texi (w): Extend the description of the AArch64
5001 constraint to include SVE vectors.
5002 (Upl, Upa): Document new AArch64 predicate constraints.
5003 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5005 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5006 (msve-vector-bits=): New option.
5007 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5008 SVE when these are disabled.
5009 (sve): New extension.
5010 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5011 modes. Adjust their number of units based on aarch64_sve_vg.
5012 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5013 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5014 aarch64_addr_query_type.
5015 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5016 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5017 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5018 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5019 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5020 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5021 (aarch64_simd_imm_zero_p): Delete.
5022 (aarch64_check_zero_based_sve_index_immediate): Declare.
5023 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5024 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5025 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5026 (aarch64_sve_float_mul_immediate_p): Likewise.
5027 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5029 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5030 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5031 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5032 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5033 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5034 (aarch64_regmode_natural_size): Likewise.
5035 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5036 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5038 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5039 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5040 for VG and the SVE predicate registers.
5041 (V_ALIASES): Add a "z"-prefixed alias.
5042 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5043 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5044 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5045 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5046 (REG_CLASS_NAMES): Add entries for them.
5047 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5048 and the predicate registers.
5049 (aarch64_sve_vg): Declare.
5050 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5051 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5052 (REGMODE_NATURAL_SIZE): Define.
5053 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5055 * config/aarch64/aarch64.c: Include cfgrtl.h.
5056 (simd_immediate_info): Add a constructor for series vectors,
5057 and an associated step field.
5058 (aarch64_sve_vg): New variable.
5059 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5060 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5061 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5062 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5063 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5064 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5065 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5066 (aarch64_get_mask_mode): New functions.
5067 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5068 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5069 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5070 predicate modes and predicate registers. Explicitly restrict
5071 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5072 to store a vector mode if it is recognized by
5073 aarch64_classify_vector_mode.
5074 (aarch64_regmode_natural_size): New function.
5075 (aarch64_hard_regno_caller_save_mode): Return the original mode
5077 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5078 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5079 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5080 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5082 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5083 does not overlap dest if the function is frame-related. Handle
5085 (aarch64_split_add_offset): New function.
5086 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5087 them aarch64_add_offset.
5088 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5089 and update call to aarch64_sub_sp.
5090 (aarch64_add_cfa_expression): New function.
5091 (aarch64_expand_prologue): Pass extra temporary registers to the
5092 functions above. Handle the case in which we need to emit new
5093 DW_CFA_expressions for registers that were originally saved
5094 relative to the stack pointer, but now have to be expressed
5095 relative to the frame pointer.
5096 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5098 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5099 IP0 and IP1 values for SVE frames.
5100 (aarch64_expand_vec_series): New function.
5101 (aarch64_expand_sve_widened_duplicate): Likewise.
5102 (aarch64_expand_sve_const_vector): Likewise.
5103 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5104 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5105 into the register, rather than emitting a SET directly.
5106 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5107 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5108 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5109 (offset_9bit_signed_scaled_p): New functions.
5110 (aarch64_replicate_bitmask_imm): New function.
5111 (aarch64_bitmask_imm): Use it.
5112 (aarch64_cannot_force_const_mem): Reject expressions involving
5113 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5114 (aarch64_classify_index): Handle SVE indices, by requiring
5115 a plain register index with a scale that matches the element size.
5116 (aarch64_classify_address): Handle SVE addresses. Assert that
5117 the mode of the address is VOIDmode or an integer mode.
5118 Update call to aarch64_classify_symbol.
5119 (aarch64_classify_symbolic_expression): Update call to
5120 aarch64_classify_symbol.
5121 (aarch64_const_vec_all_in_range_p): New function.
5122 (aarch64_print_vector_float_operand): Likewise.
5123 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5124 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5125 and the FP immediates 1.0 and 0.5.
5126 (aarch64_print_address_internal): Handle SVE addresses.
5127 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5128 (aarch64_regno_regclass): Handle predicate registers.
5129 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5131 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5132 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5133 (aarch64_convert_sve_vector_bits): New function.
5134 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5135 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5137 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5138 Handle SVE vector and predicate modes. Accept VL-based constants
5139 that need only one temporary register, and VL offsets that require
5140 no temporary registers.
5141 (aarch64_conditional_register_usage): Mark the predicate registers
5142 as fixed if SVE isn't available.
5143 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5144 Return true for SVE vector and predicate modes.
5145 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5146 rather than an unsigned int. Handle SVE modes.
5147 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5149 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
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): New functions.
5155 (aarch64_sve_valid_immediate): New function.
5156 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5157 Explicitly reject structure modes. Check for INDEX constants.
5158 Handle PTRUE and PFALSE constants.
5159 (aarch64_check_zero_based_sve_index_immediate): New function.
5160 (aarch64_simd_imm_zero_p): Delete.
5161 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5162 vector modes. Accept constants in the range of CNT[BHWD].
5163 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5164 ask for an Advanced SIMD mode.
5165 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5166 (aarch64_simd_vector_alignment): Handle SVE predicates.
5167 (aarch64_vectorize_preferred_vector_alignment): New function.
5168 (aarch64_simd_vector_alignment_reachable): Use it instead of
5170 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5171 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5173 (MAX_VECT_LEN): Delete.
5174 (expand_vec_perm_d): Add a vec_flags field.
5175 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5176 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5177 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5179 (aarch64_evpc_rev): Rename to...
5180 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5181 (aarch64_evpc_rev_global): New function.
5182 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5183 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5185 (aarch64_evpc_sve_tbl): New function.
5186 (aarch64_expand_vec_perm_const_1): Update after rename of
5187 aarch64_evpc_rev. Handle SVE permutes too, trying
5188 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5189 than aarch64_evpc_tbl.
5190 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5191 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5192 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5193 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5194 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5195 (aarch64_expand_sve_vcond): New functions.
5196 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5197 of aarch64_vector_mode_p.
5198 (aarch64_dwarf_poly_indeterminate_value): New function.
5199 (aarch64_compute_pressure_classes): Likewise.
5200 (aarch64_can_change_mode_class): Likewise.
5201 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5202 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5203 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5204 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5205 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5206 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5207 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5208 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5210 (Dn, Dl, Dr): Accept const as well as const_vector.
5211 (Dz): Likewise. Compare against CONST0_RTX.
5212 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5213 of "vector" where appropriate.
5214 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5215 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5216 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5217 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5218 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5219 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5220 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5221 (v_int_equiv): Extend to SVE modes.
5222 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5224 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5225 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5226 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5227 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5228 (SVE_COND_FP_CMP): New int iterators.
5229 (perm_hilo): Handle the new unpack unspecs.
5230 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5232 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5233 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5234 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5235 (aarch64_equality_operator, aarch64_constant_vector_operand)
5236 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5237 (aarch64_sve_nonimmediate_operand): Likewise.
5238 (aarch64_sve_general_operand): Likewise.
5239 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5240 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5241 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5242 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5243 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5244 (aarch64_sve_float_arith_immediate): Likewise.
5245 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5246 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5247 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5248 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5249 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5250 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5251 (aarch64_sve_float_arith_operand): Likewise.
5252 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5253 (aarch64_sve_float_mul_operand): Likewise.
5254 (aarch64_sve_vec_perm_operand): Likewise.
5255 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5256 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5257 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5258 as well as const_vector.
5259 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5260 in file. Use CONST0_RTX and CONSTM1_RTX.
5261 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5262 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5263 Use aarch64_simd_imm_zero.
5264 * config/aarch64/aarch64-sve.md: New file.
5265 * config/aarch64/aarch64.md: Include it.
5266 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5267 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5268 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5269 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5270 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5271 (sve): New attribute.
5272 (enabled): Disable instructions with the sve attribute unless
5274 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5275 aarch64_expand_mov_immediate.
5276 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5277 CNT[BHSD] immediates.
5278 (movti): Split CONST_POLY_INT moves into two halves.
5279 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5280 Split additions that need a temporary here if the destination
5281 is the stack pointer.
5282 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5283 (*add<mode>3_poly_1): New instruction.
5284 (set_clobber_cc): New expander.
5286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5288 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5289 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5290 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5291 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5292 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5293 Change innermode from fixed_mode_size to machine_mode.
5294 (simplify_subreg): Update call accordingly. Handle a constant-sized
5295 subreg of a variable-length CONST_VECTOR.
5297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5298 Alan Hayward <alan.hayward@arm.com>
5299 David Sherwood <david.sherwood@arm.com>
5301 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5302 (add_offset_to_base): New function, split out from...
5303 (create_mem_ref): ...here. When handling a scale other than 1,
5304 check first whether the address is valid without the offset.
5305 Add it into the base if so, leaving the index and scale as-is.
5307 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5310 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5311 fold_for_warn before checking if arg2 is INTEGER_CST.
5313 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5315 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5316 (store_multiple_operation): Delete.
5317 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5318 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5319 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5320 guarded by TARGET_STRING.
5321 (rs6000_output_load_multiple): Delete.
5322 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5323 OPTION_MASK_STRING / TARGET_STRING handling.
5324 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5325 (const rs6000_opt_masks) <"string">: Change mask to 0.
5326 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5327 (MASK_STRING): Delete.
5328 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5330 (load_multiple): Delete.
5337 (store_multiple): Delete.
5344 (movmemsi_8reg): Delete.
5345 (corresponding unnamed define_insn): Delete.
5346 (movmemsi_6reg): Delete.
5347 (corresponding unnamed define_insn): Delete.
5348 (movmemsi_4reg): Delete.
5349 (corresponding unnamed define_insn): Delete.
5350 (movmemsi_2reg): Delete.
5351 (corresponding unnamed define_insn): Delete.
5352 (movmemsi_1reg): Delete.
5353 (corresponding unnamed define_insn): Delete.
5354 * config/rs6000/rs6000.opt (mno-string): New.
5355 (mstring): Replace by deprecation warning stub.
5356 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5358 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5360 * regrename.c (regrename_do_replace): If replacing the same
5361 reg multiple times, try to reuse last created gen_raw_REG.
5364 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5365 main to workaround a bug in GDB.
5367 2018-01-12 Tom de Vries <tom@codesourcery.com>
5370 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5372 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5374 PR rtl-optimization/80481
5375 * ira-color.c (get_cap_member): New function.
5376 (allocnos_conflict_by_live_ranges_p): Use it.
5377 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5378 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5380 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5383 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5384 (*saddl_se_1): Ditto.
5386 (*ssubl_se_1): Ditto.
5388 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5390 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5391 rather than wi::to_widest for DR_INITs.
5392 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5393 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5394 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5396 (vect_analyze_group_access_1): Note that here.
5398 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5400 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5401 polynomial type sizes.
5403 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5405 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5406 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5407 (gimple_add_tmp_var): Likewise.
5409 2018-01-12 Martin Liska <mliska@suse.cz>
5411 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5412 (gimple_alloc_sizes): Likewise.
5413 (dump_gimple_statistics): Use PRIu64 in printf format.
5414 * gimple.h: Change uint64_t to int.
5416 2018-01-12 Martin Liska <mliska@suse.cz>
5418 * tree-core.h: Use uint64_t instead of int.
5419 * tree.c (tree_node_counts): Likewise.
5420 (tree_node_sizes): Likewise.
5421 (dump_tree_statistics): Use PRIu64 in printf format.
5423 2018-01-12 Martin Liska <mliska@suse.cz>
5425 * Makefile.in: As qsort_chk is implemented in vec.c, add
5426 vec.o to linkage of gencfn-macros.
5427 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5428 passing the info to record_node_allocation_statistics.
5429 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5431 * ggc-common.c (struct ggc_usage): Add operator== and use
5432 it in operator< and compare function.
5433 * mem-stats.h (struct mem_usage): Likewise.
5434 * vec.c (struct vec_usage): Remove operator< and compare
5435 function. Can be simply inherited.
5437 2018-01-12 Martin Jambor <mjambor@suse.cz>
5440 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5441 * tree-ssa-math-opts.c: Include domwalk.h.
5442 (convert_mult_to_fma_1): New function.
5443 (fma_transformation_info): New type.
5444 (fma_deferring_state): Likewise.
5445 (cancel_fma_deferring): New function.
5446 (result_of_phi): Likewise.
5447 (last_fma_candidate_feeds_initial_phi): Likewise.
5448 (convert_mult_to_fma): Added deferring logic, split actual
5449 transformation to convert_mult_to_fma_1.
5450 (math_opts_dom_walker): New type.
5451 (math_opts_dom_walker::after_dom_children): New method, body moved
5452 here from pass_optimize_widening_mul::execute, added deferring logic
5454 (pass_optimize_widening_mul::execute): Moved most of code to
5455 math_opts_dom_walker::after_dom_children.
5456 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5457 * config/i386/i386.c (ix86_option_override_internal): Added
5458 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5460 2018-01-12 Richard Biener <rguenther@suse.de>
5463 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5464 inline instance vars.
5466 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5469 * config/rx/rx.c (rx_is_restricted_memory_address):
5472 2018-01-12 Richard Biener <rguenther@suse.de>
5474 PR tree-optimization/80846
5475 * target.def (split_reduction): New target hook.
5476 * targhooks.c (default_split_reduction): New function.
5477 * targhooks.h (default_split_reduction): Declare.
5478 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5479 target requests first reduce vectors by combining low and high
5481 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5482 (get_vectype_for_scalar_type_and_size): Export.
5483 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5484 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5485 * doc/tm.texi: Regenerate.
5486 * config/i386/i386.c (ix86_split_reduction): Implement
5487 TARGET_VECTORIZE_SPLIT_REDUCTION.
5489 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5492 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5493 in PIC mode except for TARGET_VXWORKS_RTP.
5494 * config/sparc/sparc.c: Include cfgrtl.h.
5495 (TARGET_INIT_PIC_REG): Define.
5496 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5497 (sparc_pic_register_p): New predicate.
5498 (sparc_legitimate_address_p): Use it.
5499 (sparc_legitimize_pic_address): Likewise.
5500 (sparc_delegitimize_address): Likewise.
5501 (sparc_mode_dependent_address_p): Likewise.
5502 (gen_load_pcrel_sym): Remove 4th parameter.
5503 (load_got_register): Adjust call to above. Remove obsolete stuff.
5504 (sparc_expand_prologue): Do not call load_got_register here.
5505 (sparc_flat_expand_prologue): Likewise.
5506 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5507 (sparc_use_pseudo_pic_reg): New function.
5508 (sparc_init_pic_reg): Likewise.
5509 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5510 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5512 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5514 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5515 Add item for branch_cost.
5517 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5519 PR rtl-optimization/83565
5520 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5521 not extend the result to a larger mode for rotate operations.
5522 (num_sign_bit_copies1): Likewise.
5524 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5527 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5529 Use values-Xc.o for -pedantic.
5530 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5532 2018-01-12 Martin Liska <mliska@suse.cz>
5535 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5537 (possible_polymorphic_call_targets): Use it.
5538 (ipa_devirt): Likewise.
5540 2018-01-12 Martin Liska <mliska@suse.cz>
5542 * profile-count.h (enum profile_quality): Use 0 as invalid
5543 enum value of profile_quality.
5545 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5547 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5548 -mext-string options.
5550 2018-01-12 Richard Biener <rguenther@suse.de>
5552 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5553 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5554 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5556 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5558 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5560 * configure.ac (--with-long-double-format): Add support for the
5561 configuration option to change the default long double format on
5563 * config.gcc (powerpc*-linux*-*): Likewise.
5564 * configure: Regenerate.
5565 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5566 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5567 used without modification.
5569 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5571 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5572 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5573 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5574 MISC_BUILTIN_SPEC_BARRIER.
5575 (rs6000_init_builtins): Likewise.
5576 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5578 (speculation_barrier): New define_insn.
5579 * doc/extend.texi: Document __builtin_speculation_barrier.
5581 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5584 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5585 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5586 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5588 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5589 integral modes instead of "ss" and "sd".
5590 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5591 vectors with 32-bit and 64-bit elements.
5592 (vecdupssescalarmodesuffix): New mode attribute.
5593 (vec_dup<mode>): Use it.
5595 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5598 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5599 frame if argument is passed on stack.
5601 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5604 * ree.c (combine_reaching_defs): Optimize also
5605 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5606 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5608 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5611 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5613 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5616 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5617 after they are computed.
5619 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5621 PR tree-optimization/83695
5622 * gimple-loop-linterchange.cc
5623 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5624 reset cached scev information after interchange.
5625 (pass_linterchange::execute): Remove call to scev_reset_htab.
5627 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5629 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5630 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5631 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5632 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5633 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5634 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5635 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5636 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5637 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5638 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5639 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5640 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5641 (V_lane_reg): Likewise.
5642 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5644 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5645 (vfmal_lane_low<mode>_intrinsic,
5646 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5647 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5648 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5649 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5650 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5651 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5653 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5655 * config/arm/arm-cpus.in (fp16fml): New feature.
5656 (ALL_SIMD): Add fp16fml.
5657 (armv8.2-a): Add fp16fml as an option.
5658 (armv8.3-a): Likewise.
5659 (armv8.4-a): Add fp16fml as part of fp16.
5660 * config/arm/arm.h (TARGET_FP16FML): Define.
5661 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5663 * config/arm/arm-modes.def (V2HF): Define.
5664 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5665 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5666 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5667 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5668 vfmsl_low, vfmsl_high): New set of builtins.
5669 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5670 (vfml_op): New code attribute.
5671 (VFMLHALVES): New int iterator.
5672 (VFML, VFMLSEL): New mode attributes.
5673 (V_reg): Define mapping for V2HF.
5674 (V_hi, V_lo): New mode attributes.
5675 (VF_constraint): Likewise.
5676 (vfml_half, vfml_half_selector): New int attributes.
5677 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5679 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5680 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5682 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5683 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5684 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5685 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5687 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5688 Document new effective target and option set.
5690 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5692 * config/arm/arm-cpus.in (armv8_4): New feature.
5693 (ARMv8_4a): New fgroup.
5694 (armv8.4-a): New arch.
5695 * config/arm/arm-tables.opt: Regenerate.
5696 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5697 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5698 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5699 Add matching rules for -march=armv8.4-a and extensions.
5700 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5702 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5705 * config/rx/rx.md (BW): New mode attribute.
5706 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5708 2018-01-11 Richard Biener <rguenther@suse.de>
5710 PR tree-optimization/83435
5711 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5712 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5713 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5715 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5716 Alan Hayward <alan.hayward@arm.com>
5717 David Sherwood <david.sherwood@arm.com>
5719 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5721 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5722 (aarch64_print_address_internal): Use it to check for a zero offset.
5724 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5725 Alan Hayward <alan.hayward@arm.com>
5726 David Sherwood <david.sherwood@arm.com>
5728 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5729 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5730 Return a poly_int64 rather than a HOST_WIDE_INT.
5731 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5732 rather than a HOST_WIDE_INT.
5733 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5734 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5735 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5736 final_offset from HOST_WIDE_INT to poly_int64.
5737 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5738 to_constant when getting the number of units in an Advanced SIMD
5740 (aarch64_builtin_vectorized_function): Check for a constant number
5742 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5744 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5745 attribute instead of GET_MODE_NUNITS.
5746 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5747 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5748 GET_MODE_SIZE for fixed-size registers.
5749 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5750 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5751 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5752 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5753 (aarch64_print_operand, aarch64_print_address_internal)
5754 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5755 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5756 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5757 Handle polynomial GET_MODE_SIZE.
5758 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5759 wider than SImode without modification.
5760 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5761 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5762 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5763 passing and returning SVE modes.
5764 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5765 rather than GEN_INT.
5766 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5767 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5768 (aarch64_allocate_and_probe_stack_space): Likewise.
5769 (aarch64_layout_frame): Cope with polynomial offsets.
5770 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5771 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5773 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5774 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5775 poly_int64 rather than a HOST_WIDE_INT.
5776 (aarch64_get_separate_components, aarch64_process_components)
5777 (aarch64_expand_prologue, aarch64_expand_epilogue)
5778 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5779 (aarch64_anchor_offset): New function, split out from...
5780 (aarch64_legitimize_address): ...here.
5781 (aarch64_builtin_vectorization_cost): Handle polynomial
5782 TYPE_VECTOR_SUBPARTS.
5783 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5785 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5786 number of elements from the PARALLEL rather than the mode.
5787 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5788 rather than GET_MODE_BITSIZE.
5789 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5790 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5791 (aarch64_expand_vec_perm_const_1): Handle polynomial
5792 d->perm.length () and d->perm elements.
5793 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5794 Apply to_constant to d->perm elements.
5795 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5796 polynomial CONST_VECTOR_NUNITS.
5797 (aarch64_move_pointer): Take amount as a poly_int64 rather
5799 (aarch64_progress_pointer): Avoid temporary variable.
5800 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5801 the mode attribute instead of GET_MODE.
5803 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5804 Alan Hayward <alan.hayward@arm.com>
5805 David Sherwood <david.sherwood@arm.com>
5807 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5808 x exists before using it.
5809 (aarch64_add_constant_internal): Rename to...
5810 (aarch64_add_offset_1): ...this. Replace regnum with separate
5811 src and dest rtxes. Handle the case in which they're different,
5812 including when the offset is zero. Replace scratchreg with an rtx.
5813 Use 2 additions if there is no spare register into which we can
5814 move a 16-bit constant.
5815 (aarch64_add_constant): Delete.
5816 (aarch64_add_offset): Replace reg with separate src and dest
5817 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5818 Use aarch64_add_offset_1.
5819 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5820 an rtx rather than an int. Take the delta as a poly_int64
5821 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5822 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5823 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5824 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5825 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5827 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5828 aarch64_add_constant.
5830 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5832 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5833 Use scalar_float_mode.
5835 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5837 * config/aarch64/aarch64-simd.md
5838 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5839 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5840 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5841 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5842 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5843 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5844 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5845 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5846 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5847 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5849 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5852 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5853 targ_options->x_arm_arch_string is non NULL.
5855 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5857 * config/aarch64/aarch64.h
5858 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5860 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5863 * expmed.c (emit_store_flag_force): Swap if const op0
5864 and change VOIDmode to mode of op0.
5866 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5868 PR rtl-optimization/83761
5869 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5870 than bytes to mode_for_size.
5872 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5875 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5876 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5879 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5882 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5883 when in layout mode.
5884 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5885 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5888 2018-01-10 Michael Collison <michael.collison@arm.com>
5890 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5891 * config/aarch64/aarch64-option-extension.def: Add
5892 AARCH64_OPT_EXTENSION of 'fp16fml'.
5893 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5894 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5895 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5896 * config/aarch64/constraints.md (Ui7): New constraint.
5897 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5898 (VFMLA_SEL_W): Ditto.
5901 (VFMLA16_LOW): New int iterator.
5902 (VFMLA16_HIGH): Ditto.
5903 (UNSPEC_FMLAL): New unspec.
5904 (UNSPEC_FMLSL): Ditto.
5905 (UNSPEC_FMLAL2): Ditto.
5906 (UNSPEC_FMLSL2): Ditto.
5907 (f16mac): New code attribute.
5908 * config/aarch64/aarch64-simd-builtins.def
5909 (aarch64_fmlal_lowv2sf): Ditto.
5910 (aarch64_fmlsl_lowv2sf): Ditto.
5911 (aarch64_fmlalq_lowv4sf): Ditto.
5912 (aarch64_fmlslq_lowv4sf): Ditto.
5913 (aarch64_fmlal_highv2sf): Ditto.
5914 (aarch64_fmlsl_highv2sf): Ditto.
5915 (aarch64_fmlalq_highv4sf): Ditto.
5916 (aarch64_fmlslq_highv4sf): Ditto.
5917 (aarch64_fmlal_lane_lowv2sf): Ditto.
5918 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5919 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5920 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5921 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5922 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5923 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5924 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5925 (aarch64_fmlal_lane_highv2sf): Ditto.
5926 (aarch64_fmlsl_lane_highv2sf): Ditto.
5927 (aarch64_fmlal_laneq_highv2sf): Ditto.
5928 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5929 (aarch64_fmlalq_lane_highv4sf): Ditto.
5930 (aarch64_fmlsl_lane_highv4sf): Ditto.
5931 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5932 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5933 * config/aarch64/aarch64-simd.md:
5934 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5935 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5936 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5937 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5938 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5939 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5940 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5941 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5942 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5943 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5944 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5945 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5946 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5947 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5948 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5949 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5950 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5951 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5952 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5953 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5954 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5955 (vfmlsl_low_u32): Ditto.
5956 (vfmlalq_low_u32): Ditto.
5957 (vfmlslq_low_u32): Ditto.
5958 (vfmlal_high_u32): Ditto.
5959 (vfmlsl_high_u32): Ditto.
5960 (vfmlalq_high_u32): Ditto.
5961 (vfmlslq_high_u32): Ditto.
5962 (vfmlal_lane_low_u32): Ditto.
5963 (vfmlsl_lane_low_u32): Ditto.
5964 (vfmlal_laneq_low_u32): Ditto.
5965 (vfmlsl_laneq_low_u32): Ditto.
5966 (vfmlalq_lane_low_u32): Ditto.
5967 (vfmlslq_lane_low_u32): Ditto.
5968 (vfmlalq_laneq_low_u32): Ditto.
5969 (vfmlslq_laneq_low_u32): Ditto.
5970 (vfmlal_lane_high_u32): Ditto.
5971 (vfmlsl_lane_high_u32): Ditto.
5972 (vfmlal_laneq_high_u32): Ditto.
5973 (vfmlsl_laneq_high_u32): Ditto.
5974 (vfmlalq_lane_high_u32): Ditto.
5975 (vfmlslq_lane_high_u32): Ditto.
5976 (vfmlalq_laneq_high_u32): Ditto.
5977 (vfmlslq_laneq_high_u32): Ditto.
5978 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5979 (AARCH64_FL_FOR_ARCH8_4): New.
5980 (AARCH64_ISA_F16FML): New ISA flag.
5981 (TARGET_F16FML): New feature flag for fp16fml.
5982 (doc/invoke.texi): Document new fp16fml option.
5984 2018-01-10 Michael Collison <michael.collison@arm.com>
5986 * config/aarch64/aarch64-builtins.c:
5987 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5988 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5989 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5990 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5991 (AARCH64_ISA_SHA3): New ISA flag.
5992 (TARGET_SHA3): New feature flag for sha3.
5993 * config/aarch64/iterators.md (sha512_op): New int attribute.
5994 (CRYPTO_SHA512): New int iterator.
5995 (UNSPEC_SHA512H): New unspec.
5996 (UNSPEC_SHA512H2): Ditto.
5997 (UNSPEC_SHA512SU0): Ditto.
5998 (UNSPEC_SHA512SU1): Ditto.
5999 * config/aarch64/aarch64-simd-builtins.def
6000 (aarch64_crypto_sha512hqv2di): New builtin.
6001 (aarch64_crypto_sha512h2qv2di): Ditto.
6002 (aarch64_crypto_sha512su0qv2di): Ditto.
6003 (aarch64_crypto_sha512su1qv2di): Ditto.
6004 (aarch64_eor3qv8hi): Ditto.
6005 (aarch64_rax1qv2di): Ditto.
6006 (aarch64_xarqv2di): Ditto.
6007 (aarch64_bcaxqv8hi): Ditto.
6008 * config/aarch64/aarch64-simd.md:
6009 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6010 (aarch64_crypto_sha512su0qv2di): Ditto.
6011 (aarch64_crypto_sha512su1qv2di): Ditto.
6012 (aarch64_eor3qv8hi): Ditto.
6013 (aarch64_rax1qv2di): Ditto.
6014 (aarch64_xarqv2di): Ditto.
6015 (aarch64_bcaxqv8hi): Ditto.
6016 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6017 (vsha512h2q_u64): Ditto.
6018 (vsha512su0q_u64): Ditto.
6019 (vsha512su1q_u64): Ditto.
6020 (veor3q_u16): Ditto.
6021 (vrax1q_u64): Ditto.
6023 (vbcaxq_u16): Ditto.
6024 * config/arm/types.md (crypto_sha512): New type attribute.
6025 (crypto_sha3): Ditto.
6026 (doc/invoke.texi): Document new sha3 option.
6028 2018-01-10 Michael Collison <michael.collison@arm.com>
6030 * config/aarch64/aarch64-builtins.c:
6031 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6032 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6033 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6034 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6035 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6036 (AARCH64_ISA_SM4): New ISA flag.
6037 (TARGET_SM4): New feature flag for sm4.
6038 * config/aarch64/aarch64-simd-builtins.def
6039 (aarch64_sm3ss1qv4si): Ditto.
6040 (aarch64_sm3tt1aq4si): Ditto.
6041 (aarch64_sm3tt1bq4si): Ditto.
6042 (aarch64_sm3tt2aq4si): Ditto.
6043 (aarch64_sm3tt2bq4si): Ditto.
6044 (aarch64_sm3partw1qv4si): Ditto.
6045 (aarch64_sm3partw2qv4si): Ditto.
6046 (aarch64_sm4eqv4si): Ditto.
6047 (aarch64_sm4ekeyqv4si): Ditto.
6048 * config/aarch64/aarch64-simd.md:
6049 (aarch64_sm3ss1qv4si): Ditto.
6050 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6051 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6052 (aarch64_sm4eqv4si): Ditto.
6053 (aarch64_sm4ekeyqv4si): Ditto.
6054 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6055 (sm3part_op): Ditto.
6056 (CRYPTO_SM3TT): Ditto.
6057 (CRYPTO_SM3PART): Ditto.
6058 (UNSPEC_SM3SS1): New unspec.
6059 (UNSPEC_SM3TT1A): Ditto.
6060 (UNSPEC_SM3TT1B): Ditto.
6061 (UNSPEC_SM3TT2A): Ditto.
6062 (UNSPEC_SM3TT2B): Ditto.
6063 (UNSPEC_SM3PARTW1): Ditto.
6064 (UNSPEC_SM3PARTW2): Ditto.
6065 (UNSPEC_SM4E): Ditto.
6066 (UNSPEC_SM4EKEY): Ditto.
6067 * config/aarch64/constraints.md (Ui2): New constraint.
6068 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6069 * config/arm/types.md (crypto_sm3): New type attribute.
6070 (crypto_sm4): Ditto.
6071 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6072 (vsm3tt1aq_u32): Ditto.
6073 (vsm3tt1bq_u32): Ditto.
6074 (vsm3tt2aq_u32): Ditto.
6075 (vsm3tt2bq_u32): Ditto.
6076 (vsm3partw1q_u32): Ditto.
6077 (vsm3partw2q_u32): Ditto.
6078 (vsm4eq_u32): Ditto.
6079 (vsm4ekeyq_u32): Ditto.
6080 (doc/invoke.texi): Document new sm4 option.
6082 2018-01-10 Michael Collison <michael.collison@arm.com>
6084 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6085 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6086 (AARCH64_FL_FOR_ARCH8_4): New.
6087 (AARCH64_FL_V8_4): New flag.
6088 (doc/invoke.texi): Document new armv8.4-a option.
6090 2018-01-10 Michael Collison <michael.collison@arm.com>
6092 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6093 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6094 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6095 * config/aarch64/aarch64-option-extension.def: Add
6096 AARCH64_OPT_EXTENSION of 'sha2'.
6097 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6098 (crypto): Disable sha2 and aes if crypto disabled.
6099 (crypto): Enable aes and sha2 if enabled.
6100 (simd): Disable sha2 and aes if simd disabled.
6101 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6103 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6104 (TARGET_SHA2): New feature flag for sha2.
6105 (TARGET_AES): New feature flag for aes.
6106 * config/aarch64/aarch64-simd.md:
6107 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6108 conditional on TARGET_AES.
6109 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6110 (aarch64_crypto_sha1hsi): Make pattern conditional
6112 (aarch64_crypto_sha1hv4si): Ditto.
6113 (aarch64_be_crypto_sha1hv4si): Ditto.
6114 (aarch64_crypto_sha1su1v4si): Ditto.
6115 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6116 (aarch64_crypto_sha1su0v4si): Ditto.
6117 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6118 (aarch64_crypto_sha256su0v4si): Ditto.
6119 (aarch64_crypto_sha256su1v4si): Ditto.
6120 (doc/invoke.texi): Document new aes and sha2 options.
6122 2018-01-10 Martin Sebor <msebor@redhat.com>
6124 PR tree-optimization/83781
6125 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6128 2018-01-11 Martin Sebor <msebor@gmail.com>
6129 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6131 PR tree-optimization/83501
6132 PR tree-optimization/81703
6134 * tree-ssa-strlen.c (get_string_cst): Rename...
6135 (get_string_len): ...to this. Handle global constants.
6136 (handle_char_store): Adjust.
6138 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6139 Jim Wilson <jimw@sifive.com>
6141 * config/riscv/riscv-protos.h (riscv_output_return): New.
6142 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6143 (riscv_attribute_table, riscv_output_return),
6144 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6145 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6146 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6148 (riscv_expand_prologue): Add early return for naked function.
6149 (riscv_expand_epilogue): Likewise.
6150 (riscv_function_ok_for_sibcall): Return false for naked function.
6151 (riscv_set_current_function): New.
6152 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6153 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6154 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6155 * doc/extend.texi (RISC-V Function Attributes): New.
6157 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6159 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6160 check for 128-bit long double before checking TCmode.
6161 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6162 128-bit long doubles before checking TFmode or TCmode.
6163 (FLOAT128_IBM_P): Likewise.
6165 2018-01-10 Martin Sebor <msebor@redhat.com>
6167 PR tree-optimization/83671
6168 * builtins.c (c_strlen): Unconditionally return zero for the empty
6170 Use -Warray-bounds for warnings.
6171 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6172 for non-constant array indices with COMPONENT_REF, arrays of
6173 arrays, and pointers to arrays.
6174 (gimple_fold_builtin_strlen): Determine and set length range for
6175 non-constant character arrays.
6177 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6180 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6183 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6185 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6187 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6190 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6191 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6192 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6193 indexed_or_indirect_operand predicate.
6194 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6195 (*vsx_le_perm_load_v8hi): Likewise.
6196 (*vsx_le_perm_load_v16qi): Likewise.
6197 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6198 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6199 (*vsx_le_perm_store_v8hi): Likewise.
6200 (*vsx_le_perm_store_v16qi): Likewise.
6201 (eight unnamed splitters): Likewise.
6203 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6205 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6206 * config/rs6000/emmintrin.h: Likewise.
6207 * config/rs6000/mmintrin.h: Likewise.
6208 * config/rs6000/xmmintrin.h: Likewise.
6210 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6213 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6215 * tree.c (tree_nop_conversion): Return true for location wrapper
6217 (maybe_wrap_with_location): New function.
6218 (selftest::check_strip_nops): New function.
6219 (selftest::test_location_wrappers): New function.
6220 (selftest::tree_c_tests): Call it.
6221 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6222 (maybe_wrap_with_location): New decl.
6223 (EXPR_LOCATION_WRAPPER_P): New macro.
6224 (location_wrapper_p): New inline function.
6225 (tree_strip_any_location_wrapper): New inline function.
6227 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6230 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6231 stack_realign_offset for the largest alignment of stack slot
6233 (ix86_find_max_used_stack_alignment): New function.
6234 (ix86_finalize_stack_frame_flags): Use it. Set
6235 max_used_stack_alignment if we don't realign stack.
6236 * config/i386/i386.h (machine_function): Add
6237 max_used_stack_alignment.
6239 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6241 * config/arm/arm.opt (-mbranch-cost): New option.
6242 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6245 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6248 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6249 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6251 2018-01-10 Richard Biener <rguenther@suse.de>
6254 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6255 early out so it also covers the case where we have a non-NULL
6258 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6260 PR tree-optimization/83753
6261 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6262 for non-strided grouped accesses if the number of elements is 1.
6264 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6267 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6268 * i386.h (TARGET_USE_GATHER): Define.
6269 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6271 2018-01-10 Martin Liska <mliska@suse.cz>
6274 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6275 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6277 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6278 CLEANUP_NO_PARTITIONING is not set.
6280 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6282 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6283 for vectors, as a partial revert of r254296.
6284 * rtl.h (const_vec_p): Delete.
6285 (const_vec_duplicate_p): Don't test for vector CONSTs.
6286 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6287 * expmed.c (make_tree): Likewise.
6290 * common.md (E, F): Use CONSTANT_P instead of checking for
6292 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6293 checking for CONST_VECTOR.
6295 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6298 * predict.c (force_edge_cold): Handle in more sane way edges
6301 2018-01-09 Carl Love <cel@us.ibm.com>
6303 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6305 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6306 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6307 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6308 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6309 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6310 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6311 * config/rs6000/rs6000-protos.h: Add extern defition for
6312 rs6000_generate_float2_double_code.
6313 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6315 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6316 (float2_v2df): Add define_expand.
6318 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6321 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6322 op_mode in the force_to_mode call.
6324 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6326 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6327 instead of checking each element individually.
6328 (aarch64_evpc_uzp): Likewise.
6329 (aarch64_evpc_zip): Likewise.
6330 (aarch64_evpc_ext): Likewise.
6331 (aarch64_evpc_rev): Likewise.
6332 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6333 instead of checking each element individually. Return true without
6335 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6336 whether all selected elements come from the same input, instead of
6337 checking each element individually. Remove calls to gen_rtx_REG,
6338 start_sequence and end_sequence and instead assert that no rtl is
6341 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6343 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6344 order of HIGH and CONST checks.
6346 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6348 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6349 if the destination isn't an SSA_NAME.
6351 2018-01-09 Richard Biener <rguenther@suse.de>
6353 PR tree-optimization/83668
6354 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6356 (canonicalize_loop_form): ... here, renamed from ...
6357 (canonicalize_loop_closed_ssa_form): ... this and amended to
6358 swap successor edges for loop exit blocks to make us use
6359 the RPO order we need for initial schedule generation.
6361 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6363 PR tree-optimization/64811
6364 * match.pd: When optimizing comparisons with Inf, avoid
6365 introducing or losing exceptions from comparisons with NaN.
6367 2018-01-09 Martin Liska <mliska@suse.cz>
6370 * asan.c (shadow_mem_size): Add gcc_assert.
6372 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6374 Don't save registers in main().
6377 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6378 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6379 * config/avr/avr.c (avr_set_current_function): Don't error if
6380 naked, OS_task or OS_main are specified at the same time.
6381 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6383 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6385 * common/config/avr/avr-common.c (avr_option_optimization_table):
6386 Switch on -mmain-is-OS_task for optimizing compilations.
6388 2018-01-09 Richard Biener <rguenther@suse.de>
6390 PR tree-optimization/83572
6391 * graphite.c: Include cfganal.h.
6392 (graphite_transform_loops): Connect infinite loops to exit
6393 and remove fake edges at the end.
6395 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6397 * ipa-inline.c (edge_badness): Revert accidental checkin.
6399 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6402 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6403 symbols; not inline clones.
6405 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6408 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6409 hard registers. Formatting fixes.
6411 PR preprocessor/83722
6412 * gcc.c (try_generate_repro): Pass
6413 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6414 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6417 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6418 Kito Cheng <kito.cheng@gmail.com>
6420 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6421 (riscv_leaf_function_p): Delete.
6422 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6424 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6426 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6428 (do_ifelse): New function.
6429 (do_isel): New function.
6430 (do_sub3): New function.
6431 (do_add3): New function.
6432 (do_load_mask_compare): New function.
6433 (do_overlap_load_compare): New function.
6434 (expand_compare_loop): New function.
6435 (expand_block_compare): Call expand_compare_loop() when appropriate.
6436 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6438 (-mblock-compare-inline-loop-limit): New option.
6440 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6443 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6444 Reverse order of second and third operands in first alternative.
6445 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6446 of first and second elements in UNSPEC_VPERMR vector.
6447 (altivec_expand_vec_perm_le): Likewise.
6449 2018-01-08 Jeff Law <law@redhat.com>
6451 PR rtl-optimizatin/81308
6452 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6453 (process_switch): If group_case_labels makes a change, then set
6455 (pass_convert_switch::execute): If a switch is converted, then
6456 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6458 PR rtl-optimization/81308
6459 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6462 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6464 PR target/83663 - Revert r255946
6465 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6466 generation for cases where splatting a value is not useful.
6467 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6468 across a vec_duplicate and a paradoxical subreg forming a vector
6469 mode to a vec_concat.
6471 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6473 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6474 -march=armv8.3-a variants.
6475 * config/arm/t-multilib: Likewise.
6476 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6478 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6480 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6482 (cceq_ior_compare_complement): Give it a name so I can use it, and
6483 change boolean_or_operator predicate to boolean_operator so it can
6484 be used to generate a crand.
6485 (eqne): New code iterator.
6486 (bd/bd_neg): New code_attrs.
6487 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6488 a single define_insn.
6489 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6490 decrement (bdnzt/bdnzf/bdzt/bdzf).
6491 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6492 with the new names of the branch decrement patterns, and added the
6493 names of the branch decrement conditional patterns.
6495 2018-01-08 Richard Biener <rguenther@suse.de>
6497 PR tree-optimization/83563
6498 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6501 2018-01-08 Richard Biener <rguenther@suse.de>
6504 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6506 2018-01-08 Richard Biener <rguenther@suse.de>
6508 PR tree-optimization/83685
6509 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6510 references to abnormals.
6512 2018-01-08 Richard Biener <rguenther@suse.de>
6515 * dwarf2out.c (output_indirect_strings): Handle empty
6516 skeleton_debug_str_hash.
6517 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6519 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6521 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6522 (emit_store_direct): Likewise.
6523 (arc_trampoline_adjust_address): Likewise.
6524 (arc_asm_trampoline_template): New function.
6525 (arc_initialize_trampoline): Use asm_trampoline_template.
6526 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6527 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6528 * config/arc/arc.md (flush_icache): Delete pattern.
6530 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6532 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6533 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6536 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6539 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6540 by not USED_FOR_TARGET.
6541 (make_pass_resolve_sw_modes): Likewise.
6543 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6545 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6548 2018-01-08 Richard Biener <rguenther@suse.de>
6551 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6553 2018-01-08 Richard Biener <rguenther@suse.de>
6556 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6558 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6561 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6562 basic blocks with a small number of successors.
6563 (convert_control_dep_chain_into_preds): Improve handling of
6565 (dump_predicates): Split apart into...
6566 (dump_pred_chain): ...here...
6567 (dump_pred_info): ...and here.
6568 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6569 (can_chain_union_be_invalidated_p): Improve check for invalidation
6571 (uninit_uses_cannot_happen): Avoid unnecessary if
6572 convert_control_dep_chain_into_preds yielded nothing.
6574 2018-01-06 Martin Sebor <msebor@redhat.com>
6576 PR tree-optimization/83640
6577 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6578 subtracting negative offset from size.
6579 (builtin_access::overlap): Adjust offset bounds of the access to fall
6580 within the size of the object if possible.
6582 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6584 PR rtl-optimization/83699
6585 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6586 extract_bit_field_as_subreg to cases in which the extracted
6587 value is also a vector.
6589 * lra-constraints.c (process_alt_operands): Test for the equivalence
6590 substitutions when detecting a possible reload cycle.
6592 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6595 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6596 by default if flag_selective_schedling{,2}. Formatting fixes.
6598 PR rtl-optimization/83682
6599 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6600 if it has non-VECTOR_MODE element mode.
6601 (vec_duplicate_p): Likewise.
6604 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6605 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6607 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6610 * config/i386/i386-builtin.def
6611 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6612 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6613 Require also OPTION_MASK_ISA_AVX512F in addition to
6614 OPTION_MASK_ISA_GFNI.
6615 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6616 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6617 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6618 to OPTION_MASK_ISA_GFNI.
6619 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6620 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6621 OPTION_MASK_ISA_AVX512BW.
6622 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6623 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6624 addition to OPTION_MASK_ISA_GFNI.
6625 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6626 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6627 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6628 to OPTION_MASK_ISA_GFNI.
6629 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6630 a requirement for all ISAs rather than any of them with a few
6632 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6634 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6635 bitmasks to be enabled with 3 exceptions, instead of requiring any
6636 enabled ISA with lots of exceptions.
6637 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6638 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6639 Change avx512bw in isa attribute to avx512f.
6640 * config/i386/sgxintrin.h: Add license boilerplate.
6641 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6642 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6643 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6644 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6646 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6647 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6648 temporarily sse2 rather than sse if not enabled already.
6651 * config/i386/sse.md (VI248_VLBW): Rename to ...
6652 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6653 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6654 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6655 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6656 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6657 mode iterator instead of VI248_VLBW.
6659 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6661 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6662 (record_modified): Skip clobbers; add debug output.
6663 (param_change_prob): Use sreal frequencies.
6665 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6667 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6668 punt for user-aligned variables.
6670 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6672 * tree-chrec.c (chrec_contains_symbols): Return true for
6675 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6678 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6679 of (x|y) == x for BICS pattern.
6681 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6683 PR tree-optimization/83605
6684 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6685 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6688 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6690 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6691 * config/epiphany/rtems.h: New file.
6693 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6694 Uros Bizjak <ubizjak@gmail.com>
6697 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6698 QIreg_operand instead of register_operand predicate.
6699 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6700 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6701 comments instead of -fmitigate[-_]rop.
6703 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6706 * cgraphunit.c (symbol_table::compile): Switch to text_section
6707 before calling assembly_start debug hook.
6708 * run-rtl-passes.c (run_rtl_passes): Likewise.
6711 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6713 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6714 range_int_cst_p rather than !symbolic_range_p before calling
6715 extract_range_from_multiplicative_op_1.
6717 2018-01-04 Jeff Law <law@redhat.com>
6719 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6720 redundant test in assertion.
6722 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6724 * doc/rtl.texi: Document machine_mode wrapper classes.
6726 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6728 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6731 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6733 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6734 the VEC_PERM_EXPR fold to fail.
6736 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6739 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6740 to switched_sections.
6742 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6745 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6748 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6751 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6752 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6754 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6757 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6758 is BLKmode and bitpos not zero or mode change is needed.
6760 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6763 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6766 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6769 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6770 instead of MULT rtx. Update all corresponding splitters.
6772 (*ssub<modesuffix>): Ditto.
6774 (*cmp_sadd_di): Update split patterns.
6775 (*cmp_sadd_si): Ditto.
6776 (*cmp_sadd_sidi): Ditto.
6777 (*cmp_ssub_di): Ditto.
6778 (*cmp_ssub_si): Ditto.
6779 (*cmp_ssub_sidi): Ditto.
6780 * config/alpha/predicates.md (const23_operand): New predicate.
6781 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6782 Look for ASHIFT, not MULT inner operand.
6783 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6785 2018-01-04 Martin Liska <mliska@suse.cz>
6787 PR gcov-profile/83669
6788 * gcov.c (output_intermediate_file): Add version to intermediate
6790 * doc/gcov.texi: Document new field 'version' in intermediate
6791 file format. Fix location of '-k' option of gcov command.
6793 2018-01-04 Martin Liska <mliska@suse.cz>
6796 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6798 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6800 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6802 2018-01-03 Martin Sebor <msebor@redhat.com>
6804 PR tree-optimization/83655
6805 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6806 checking calls with invalid arguments.
6808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6810 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6811 (vectorizable_mask_load_store): Delete.
6812 (vectorizable_call): Return false for masked loads and stores.
6813 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6814 instead of gimple_assign_rhs1.
6815 (vectorizable_load): Handle IFN_MASK_LOAD.
6816 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6820 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6822 (vectorizable_mask_load_store): ...here.
6823 (vectorizable_load): ...and here.
6825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6827 * tree-vect-stmts.c (vect_build_all_ones_mask)
6828 (vect_build_zero_merge_argument): New functions, split out from...
6829 (vectorizable_load): ...here.
6831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6833 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6835 (vectorizable_mask_load_store): ...here.
6836 (vectorizable_store): ...and here.
6838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6840 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6842 (vectorizable_mask_load_store): ...here.
6844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6846 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6847 (vect_model_store_cost): Take a vec_load_store_type instead of a
6849 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6850 (vect_model_store_cost): Take a vec_load_store_type instead of a
6852 (vectorizable_mask_load_store): Update accordingly.
6853 (vectorizable_store): Likewise.
6854 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6858 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6859 IFN_MASK_LOAD calls here rather than...
6860 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6863 Alan Hayward <alan.hayward@arm.com>
6864 David Sherwood <david.sherwood@arm.com>
6866 * expmed.c (extract_bit_field_1): For vector extracts,
6867 fall back to extract_bit_field_as_subreg if vec_extract
6870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6871 Alan Hayward <alan.hayward@arm.com>
6872 David Sherwood <david.sherwood@arm.com>
6874 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6875 they are variable or constant sized.
6876 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6877 slots for constant-sized data.
6879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6880 Alan Hayward <alan.hayward@arm.com>
6881 David Sherwood <david.sherwood@arm.com>
6883 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6884 handling COND_EXPRs with boolean comparisons, try to find a better
6885 basis for the mask type than the boolean itself.
6887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6889 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6890 is calculated and how it can be overridden.
6891 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6892 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6894 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6898 Alan Hayward <alan.hayward@arm.com>
6899 David Sherwood <david.sherwood@arm.com>
6901 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6902 Remove the mode argument.
6903 (aarch64_simd_valid_immediate): Remove the mode and inverse
6905 * config/aarch64/iterators.md (bitsize): New iterator.
6906 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6907 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6908 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6909 aarch64_simd_valid_immediate.
6910 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6911 (aarch64_reg_or_bic_imm): Likewise.
6912 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6913 with an insn_type enum and msl with a modifier_type enum.
6914 Replace element_width with a scalar_mode. Change the shift
6915 to unsigned int. Add constructors for scalar_float_mode and
6916 scalar_int_mode elements.
6917 (aarch64_vect_float_const_representable_p): Delete.
6918 (aarch64_can_const_movi_rtx_p)
6919 (aarch64_simd_scalar_immediate_valid_for_move)
6920 (aarch64_simd_make_constant): Update call to
6921 aarch64_simd_valid_immediate.
6922 (aarch64_advsimd_valid_immediate_hs): New function.
6923 (aarch64_advsimd_valid_immediate): Likewise.
6924 (aarch64_simd_valid_immediate): Remove mode and inverse
6925 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6926 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6927 and aarch64_float_const_representable_p on the result.
6928 (aarch64_output_simd_mov_immediate): Remove mode argument.
6929 Update call to aarch64_simd_valid_immediate and use of
6930 simd_immediate_info.
6931 (aarch64_output_scalar_simd_mov_immediate): Update call
6934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6935 Alan Hayward <alan.hayward@arm.com>
6936 David Sherwood <david.sherwood@arm.com>
6938 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6939 (mode_nunits): Likewise CONST_MODE_NUNITS.
6940 * machmode.def (ADJUST_NUNITS): Document.
6941 * genmodes.c (mode_data::need_nunits_adj): New field.
6942 (blank_mode): Update accordingly.
6943 (adj_nunits): New variable.
6944 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6946 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6947 listed in adj_nunits.
6948 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6949 listed in adj_nunits. Don't emit case statements for such modes.
6950 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6951 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6952 nothing if adj_nunits is nonnull.
6953 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6954 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6955 (emit_mode_fbit): Update use of print_maybe_const_decl.
6956 (emit_move_size): Likewise. Treat the array as non-const
6958 (emit_mode_adjustments): Handle adj_nunits.
6960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6962 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6963 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6964 (VECTOR_MODES): Use it.
6965 (make_vector_modes): Take the prefix as an argument.
6967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6968 Alan Hayward <alan.hayward@arm.com>
6969 David Sherwood <david.sherwood@arm.com>
6971 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6972 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6973 for MODE_VECTOR_BOOL.
6974 * machmode.def (VECTOR_BOOL_MODE): Document.
6975 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6976 (make_vector_bool_mode): New function.
6977 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6979 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6980 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6982 * stor-layout.c (int_mode_for_mode): Likewise.
6983 * tree.c (build_vector_type_for_mode): Likewise.
6984 * varasm.c (output_constant_pool_2): Likewise.
6985 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6986 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6987 for MODE_VECTOR_BOOL.
6988 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6989 of mode class checks.
6990 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6991 instead of a list of mode class checks.
6992 (expand_vector_scalar_condition): Likewise.
6993 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6996 Alan Hayward <alan.hayward@arm.com>
6997 David Sherwood <david.sherwood@arm.com>
6999 * machmode.h (mode_size): Change from unsigned short to
7001 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7002 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7003 or if measurement_type is not polynomial.
7004 (fixed_size_mode::includes_p): Check for constant-sized modes.
7005 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7006 return a poly_uint16 rather than an unsigned short.
7007 (emit_mode_size): Change the type of mode_size from unsigned short
7008 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7009 (emit_mode_adjustments): Cope with polynomial vector sizes.
7010 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7012 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7014 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7015 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7016 * caller-save.c (setup_save_areas): Likewise.
7017 (replace_reg_with_saved_mem): Likewise.
7018 * calls.c (emit_library_call_value_1): Likewise.
7019 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7020 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7021 (gen_lowpart_for_combine): Likewise.
7022 * convert.c (convert_to_integer_1): Likewise.
7023 * cse.c (equiv_constant, cse_insn): Likewise.
7024 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7025 (cselib_subst_to_values): Likewise.
7026 * dce.c (word_dce_process_block): Likewise.
7027 * df-problems.c (df_word_lr_mark_ref): Likewise.
7028 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7029 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7030 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7031 (rtl_for_decl_location): Likewise.
7032 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7033 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7034 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7035 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7036 (expand_expr_real_1): Likewise.
7037 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7038 (pad_below): Likewise.
7039 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7040 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7041 * ira.c (get_subreg_tracking_sizes): Likewise.
7042 * ira-build.c (ira_create_allocno_objects): Likewise.
7043 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7044 (ira_sort_regnos_for_alter_reg): Likewise.
7045 * ira-costs.c (record_operand_costs): Likewise.
7046 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7047 (resolve_simple_move): Likewise.
7048 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7049 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7050 (lra_constraints): Likewise.
7051 (CONST_POOL_OK_P): Reject variable-sized modes.
7052 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7053 (add_pseudo_to_slot, lra_spill): Likewise.
7054 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7055 * optabs-query.c (get_best_extraction_insn): Likewise.
7056 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7057 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7058 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7059 * recog.c (offsettable_address_addr_space_p): Likewise.
7060 * regcprop.c (maybe_mode_change): Likewise.
7061 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7062 * regrename.c (build_def_use): Likewise.
7063 * regstat.c (dump_reg_info): Likewise.
7064 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7065 (find_reloads, find_reloads_subreg_address): Likewise.
7066 * reload1.c (eliminate_regs_1): Likewise.
7067 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7068 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7069 (simplify_binary_operation_1, simplify_subreg): Likewise.
7070 * targhooks.c (default_function_arg_padding): Likewise.
7071 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7072 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7073 (verify_gimple_assign_ternary): Likewise.
7074 * tree-inline.c (estimate_move_cost): Likewise.
7075 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7076 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7077 (get_address_cost_ainc): Likewise.
7078 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7079 (vect_supportable_dr_alignment): Likewise.
7080 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7081 (vectorizable_reduction): Likewise.
7082 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7083 (vectorizable_operation, vectorizable_load): Likewise.
7084 * tree.c (build_same_sized_truth_vector_type): Likewise.
7085 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7086 * var-tracking.c (emit_note_insn_var_location): Likewise.
7087 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7088 (ADDR_VEC_ALIGN): Likewise.
7090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7091 Alan Hayward <alan.hayward@arm.com>
7092 David Sherwood <david.sherwood@arm.com>
7094 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7096 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7097 or if measurement_type is polynomial.
7098 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7099 * combine.c (make_extraction): Likewise.
7100 * dse.c (find_shift_sequence): Likewise.
7101 * dwarf2out.c (mem_loc_descriptor): Likewise.
7102 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7103 (extract_bit_field, extract_low_bits): Likewise.
7104 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7105 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7106 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7107 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7108 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7109 * reload.c (find_reloads): Likewise.
7110 * reload1.c (alter_reg): Likewise.
7111 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7112 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7113 * tree-if-conv.c (predicate_mem_writes): Likewise.
7114 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7115 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7116 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7117 * valtrack.c (dead_debug_insert_temp): Likewise.
7118 * varasm.c (mergeable_constant_section): Likewise.
7119 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7122 Alan Hayward <alan.hayward@arm.com>
7123 David Sherwood <david.sherwood@arm.com>
7125 * expr.c (expand_assignment): Cope with polynomial mode sizes
7126 when assigning to a CONCAT.
7128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7129 Alan Hayward <alan.hayward@arm.com>
7130 David Sherwood <david.sherwood@arm.com>
7132 * machmode.h (mode_precision): Change from unsigned short to
7134 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7136 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7137 or if measurement_type is not polynomial.
7138 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7139 in which the mode is already known to be a scalar_int_mode.
7140 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7141 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7143 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7144 for GET_MODE_PRECISION.
7145 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7146 for GET_MODE_PRECISION.
7147 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7149 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7150 (expand_field_assignment, make_extraction): Likewise.
7151 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7152 (get_last_value): Likewise.
7153 * convert.c (convert_to_integer_1): Likewise.
7154 * cse.c (cse_insn): Likewise.
7155 * expr.c (expand_expr_real_1): Likewise.
7156 * lra-constraints.c (simplify_operand_subreg): Likewise.
7157 * optabs-query.c (can_atomic_load_p): Likewise.
7158 * optabs.c (expand_atomic_load): Likewise.
7159 (expand_atomic_store): Likewise.
7160 * ree.c (combine_reaching_defs): Likewise.
7161 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7162 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7163 * tree.h (type_has_mode_precision_p): Likewise.
7164 * ubsan.c (instrument_si_overflow): Likewise.
7166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7167 Alan Hayward <alan.hayward@arm.com>
7168 David Sherwood <david.sherwood@arm.com>
7170 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7171 polynomial numbers of units.
7172 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7173 (valid_vector_subparts_p): New function.
7174 (build_vector_type): Remove temporary shim and take the number
7175 of units as a poly_uint64 rather than an int.
7176 (build_opaque_vector_type): Take the number of units as a
7177 poly_uint64 rather than an int.
7178 * tree.c (build_vector_from_ctor): Handle polynomial
7179 TYPE_VECTOR_SUBPARTS.
7180 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7181 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7182 (build_vector_from_val): If the number of units is variable,
7183 use build_vec_duplicate_cst for constant operands and
7184 VEC_DUPLICATE_EXPR otherwise.
7185 (make_vector_type): Remove temporary is_constant ().
7186 (build_vector_type, build_opaque_vector_type): Take the number of
7187 units as a poly_uint64 rather than an int.
7188 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7190 * cfgexpand.c (expand_debug_expr): Likewise.
7191 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7192 (store_constructor, expand_expr_real_1): Likewise.
7193 (const_scalar_mask_from_tree): Likewise.
7194 * fold-const-call.c (fold_const_reduction): Likewise.
7195 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7196 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7197 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7198 (fold_relational_const): Likewise.
7199 (native_interpret_vector): Likewise. Change the size from an
7200 int to an unsigned int.
7201 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7202 TYPE_VECTOR_SUBPARTS.
7203 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7204 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7205 duplicating a non-constant operand into a variable-length vector.
7206 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7207 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7208 * ipa-icf.c (sem_variable::equals): Likewise.
7209 * match.pd: Likewise.
7210 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7211 * print-tree.c (print_node): Likewise.
7212 * stor-layout.c (layout_type): Likewise.
7213 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7214 * tree-cfg.c (verify_gimple_comparison): Likewise.
7215 (verify_gimple_assign_binary): Likewise.
7216 (verify_gimple_assign_ternary): Likewise.
7217 (verify_gimple_assign_single): Likewise.
7218 * tree-pretty-print.c (dump_generic_node): Likewise.
7219 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7220 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7221 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7222 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7223 (vect_shift_permute_load_chain): Likewise.
7224 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7225 (expand_vector_condition, optimize_vector_constructor): Likewise.
7226 (lower_vec_perm, get_compute_type): Likewise.
7227 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7228 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7229 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7230 (vect_recog_mask_conversion_pattern): Likewise.
7231 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7232 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7233 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7234 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7235 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7236 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7237 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7238 (supportable_widening_operation): Likewise.
7239 (supportable_narrowing_operation): Likewise.
7240 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7242 * varasm.c (output_constant): Likewise.
7244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7245 Alan Hayward <alan.hayward@arm.com>
7246 David Sherwood <david.sherwood@arm.com>
7248 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7249 so that both the length == 3 and length != 3 cases set up their
7250 own permute vectors. Add comments explaining why we know the
7251 number of elements is constant.
7252 (vect_permute_load_chain): Likewise.
7254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7255 Alan Hayward <alan.hayward@arm.com>
7256 David Sherwood <david.sherwood@arm.com>
7258 * machmode.h (mode_nunits): Change from unsigned char to
7260 (ONLY_FIXED_SIZE_MODES): New macro.
7261 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7262 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7263 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7265 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7266 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7267 or if measurement_type is not polynomial.
7268 * genmodes.c (ZERO_COEFFS): New macro.
7269 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7271 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7272 Use ZERO_COEFFS when emitting initializers.
7273 * data-streamer.h (bp_pack_poly_value): New function.
7274 (bp_unpack_poly_value): Likewise.
7275 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7276 for GET_MODE_NUNITS.
7277 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7278 for GET_MODE_NUNITS.
7279 * tree.c (make_vector_type): Remove temporary shim and make
7280 the real function take the number of units as a poly_uint64
7282 (build_vector_type_for_mode): Handle polynomial nunits.
7283 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7284 * emit-rtl.c (const_vec_series_p_1): Likewise.
7285 (gen_rtx_CONST_VECTOR): Likewise.
7286 * fold-const.c (test_vec_duplicate_folding): Likewise.
7287 * genrecog.c (validate_pattern): Likewise.
7288 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7289 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7290 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7291 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7292 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7293 * rtlanal.c (subreg_get_info): Likewise.
7294 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7295 (vect_grouped_load_supported): Likewise.
7296 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7297 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7298 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7299 (simplify_const_unary_operation, simplify_binary_operation_1)
7300 (simplify_const_binary_operation, simplify_ternary_operation)
7301 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7302 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7303 instead of CONST_VECTOR_NUNITS.
7304 * varasm.c (output_constant_pool_2): Likewise.
7305 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7306 explicit-encoded elements in the XVEC for variable-length vectors.
7308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7310 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7313 Alan Hayward <alan.hayward@arm.com>
7314 David Sherwood <david.sherwood@arm.com>
7316 * coretypes.h (fixed_size_mode): Declare.
7317 (fixed_size_mode_pod): New typedef.
7318 * builtins.h (target_builtins::x_apply_args_mode)
7319 (target_builtins::x_apply_result_mode): Change type to
7320 fixed_size_mode_pod.
7321 * builtins.c (apply_args_size, apply_result_size, result_vector)
7322 (expand_builtin_apply_args_1, expand_builtin_apply)
7323 (expand_builtin_return): Update accordingly.
7325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7327 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7328 * cselib.c (cselib_hash_rtx): Likewise.
7329 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7330 CONST_VECTOR encoding.
7332 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7333 Jeff Law <law@redhat.com>
7336 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7337 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7338 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7339 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7342 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7343 explicitly probe *sp in a noreturn function if there were any callee
7344 register saves or frame pointer is needed.
7346 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7349 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7350 BLKmode for ternary, binary or unary expressions.
7353 * var-tracking.c (delete_vta_debug_insn): New inline function.
7354 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7355 insns from get_insns () to NULL instead of each bb separately.
7356 Use delete_vta_debug_insn. No longer static.
7357 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7358 delete_vta_debug_insns callers.
7359 * rtl.h (delete_vta_debug_insns): Declare.
7360 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7361 instead of variable_tracking_main.
7363 2018-01-03 Martin Sebor <msebor@redhat.com>
7365 PR tree-optimization/83603
7366 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7367 arguments past the endof the argument list in functions declared
7368 without a prototype.
7369 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7370 Avoid checking when arguments are null.
7372 2018-01-03 Martin Sebor <msebor@redhat.com>
7375 * doc/extend.texi (attribute const): Fix a typo.
7376 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7377 issuing -Wsuggest-attribute for void functions.
7379 2018-01-03 Martin Sebor <msebor@redhat.com>
7381 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7382 offset_int::from instead of wide_int::to_shwi.
7383 (maybe_diag_overlap): Remove assertion.
7384 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7385 * gimple-ssa-sprintf.c (format_directive): Same.
7386 (parse_directive): Same.
7387 (sprintf_dom_walker::compute_format_length): Same.
7388 (try_substitute_return_value): Same.
7390 2018-01-03 Jeff Law <law@redhat.com>
7393 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7394 non-constant residual for zero at runtime and avoid probing in
7395 that case. Reorganize code for trailing problem to mirror handling
7398 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7400 PR tree-optimization/83501
7401 * tree-ssa-strlen.c (get_string_cst): New.
7402 (handle_char_store): Call get_string_cst.
7404 2018-01-03 Martin Liska <mliska@suse.cz>
7406 PR tree-optimization/83593
7407 * tree-ssa-strlen.c: Include tree-cfg.h.
7408 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7409 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7410 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7412 (strlen_dom_walker::before_dom_children): Call
7413 gimple_purge_dead_eh_edges. Dump tranformation with details
7415 (strlen_dom_walker::before_dom_children): Update call by adding
7416 new argument cleanup_eh.
7417 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7419 2018-01-03 Martin Liska <mliska@suse.cz>
7422 * cif-code.def (VARIADIC_THUNK): New enum value.
7423 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7426 2018-01-03 Jan Beulich <jbeulich@suse.com>
7428 * sse.md (mov<mode>_internal): Tighten condition for when to use
7429 vmovdqu<ssescalarsize> for TI and OI modes.
7431 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7433 Update copyright years.
7435 2018-01-03 Martin Liska <mliska@suse.cz>
7438 * ipa-visibility.c (function_and_variable_visibility): Skip
7439 functions with noipa attribure.
7441 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7443 * gcc.c (process_command): Update copyright notice dates.
7444 * gcov-dump.c (print_version): Ditto.
7445 * gcov.c (print_version): Ditto.
7446 * gcov-tool.c (print_version): Ditto.
7447 * gengtype.c (create_file): Ditto.
7448 * doc/cpp.texi: Bump @copying's copyright year.
7449 * doc/cppinternals.texi: Ditto.
7450 * doc/gcc.texi: Ditto.
7451 * doc/gccint.texi: Ditto.
7452 * doc/gcov.texi: Ditto.
7453 * doc/install.texi: Ditto.
7454 * doc/invoke.texi: Ditto.
7456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7458 * vector-builder.h (vector_builder::m_full_nelts): Change from
7459 unsigned int to poly_uint64.
7460 (vector_builder::full_nelts): Update prototype accordingly.
7461 (vector_builder::new_vector): Likewise.
7462 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7463 (vector_builder::operator ==): Likewise.
7464 (vector_builder::finalize): Likewise.
7465 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7466 Take the number of elements as a poly_uint64 rather than an
7468 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7469 from unsigned int to poly_uint64.
7470 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7471 (vec_perm_indices::new_vector): Likewise.
7472 (vec_perm_indices::length): Likewise.
7473 (vec_perm_indices::nelts_per_input): Likewise.
7474 (vec_perm_indices::input_nelts): Likewise.
7475 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7476 number of elements per input as a poly_uint64 rather than an
7477 unsigned int. Use the original encoding for variable-length
7478 vectors, rather than clamping each individual element.
7479 For the second and subsequent elements in each pattern,
7480 clamp the step and base before clamping their sum.
7481 (vec_perm_indices::series_p): Handle polynomial element counts.
7482 (vec_perm_indices::all_in_range_p): Likewise.
7483 (vec_perm_indices_to_tree): Likewise.
7484 (vec_perm_indices_to_rtx): Likewise.
7485 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7486 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7487 (tree_vector_builder::new_binary_operation): Handle polynomial
7488 element counts. Return false if we need to know the number
7489 of elements at compile time.
7490 * fold-const.c (fold_vec_perm): Punt if the number of elements
7491 isn't known at compile time.
7493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7495 * vec-perm-indices.h (vec_perm_builder): Change element type
7496 from HOST_WIDE_INT to poly_int64.
7497 (vec_perm_indices::element_type): Update accordingly.
7498 (vec_perm_indices::clamp): Handle polynomial element_types.
7499 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7500 (vec_perm_indices::all_in_range_p): Likewise.
7501 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7503 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7504 polynomial vec_perm_indices element types.
7505 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7506 * fold-const.c (fold_vec_perm): Likewise.
7507 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7508 * tree-vect-generic.c (lower_vec_perm): Likewise.
7509 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7510 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7511 element type to HOST_WIDE_INT.
7513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7514 Alan Hayward <alan.hayward@arm.com>
7515 David Sherwood <david.sherwood@arm.com>
7517 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7518 rather than an int. Use plus_constant.
7519 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7520 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7523 Alan Hayward <alan.hayward@arm.com>
7524 David Sherwood <david.sherwood@arm.com>
7526 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7527 a HOST_WIDE_INT to a poly_int64.
7529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7530 Alan Hayward <alan.hayward@arm.com>
7531 David Sherwood <david.sherwood@arm.com>
7533 * calls.c (load_register_parameters): Cope with polynomial
7534 mode sizes. Require a constant size for BLKmode parameters
7535 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7536 forces a parameter to be padded at the lsb end in order to
7537 fill a complete number of words, require the parameter size
7538 to be ordered wrt UNITS_PER_WORD.
7540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7541 Alan Hayward <alan.hayward@arm.com>
7542 David Sherwood <david.sherwood@arm.com>
7544 * reload1.c (spill_stack_slot_width): Change element type
7545 from unsigned int to poly_uint64_pod.
7546 (alter_reg): Treat mode sizes as polynomial.
7548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7549 Alan Hayward <alan.hayward@arm.com>
7550 David Sherwood <david.sherwood@arm.com>
7552 * reload.c (complex_word_subreg_p): New function.
7553 (reload_inner_reg_of_subreg, push_reload): Use it.
7555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7556 Alan Hayward <alan.hayward@arm.com>
7557 David Sherwood <david.sherwood@arm.com>
7559 * lra-constraints.c (process_alt_operands): Reject matched
7560 operands whose sizes aren't ordered.
7561 (match_reload): Refer to this check here.
7563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7564 Alan Hayward <alan.hayward@arm.com>
7565 David Sherwood <david.sherwood@arm.com>
7567 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7568 that the mode size is in the set {1, 2, 4, 8, 16}.
7570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7571 Alan Hayward <alan.hayward@arm.com>
7572 David Sherwood <david.sherwood@arm.com>
7574 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7575 Use plus_constant instead of gen_rtx_PLUS.
7577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7578 Alan Hayward <alan.hayward@arm.com>
7579 David Sherwood <david.sherwood@arm.com>
7581 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7582 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7583 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7584 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7585 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7586 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7587 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7588 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7589 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7590 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7592 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7593 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7594 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7595 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7596 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7597 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7598 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7599 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7600 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7601 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7603 * expr.c (emit_move_resolve_push): Treat the input and result
7604 of PUSH_ROUNDING as a poly_int64.
7605 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7606 (emit_push_insn): Likewise.
7607 * lra-eliminations.c (mark_not_eliminable): Likewise.
7608 * recog.c (push_operand): Likewise.
7609 * reload1.c (elimination_effects): Likewise.
7610 * rtlanal.c (nonzero_bits1): Likewise.
7611 * calls.c (store_one_arg): Likewise. Require the padding to be
7612 known at compile time.
7614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7615 Alan Hayward <alan.hayward@arm.com>
7616 David Sherwood <david.sherwood@arm.com>
7618 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7619 Use plus_constant instead of gen_rtx_PLUS.
7621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7622 Alan Hayward <alan.hayward@arm.com>
7623 David Sherwood <david.sherwood@arm.com>
7625 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7629 Alan Hayward <alan.hayward@arm.com>
7630 David Sherwood <david.sherwood@arm.com>
7632 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7633 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7634 via stack temporaries. Treat the mode size as polynomial too.
7636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7637 Alan Hayward <alan.hayward@arm.com>
7638 David Sherwood <david.sherwood@arm.com>
7640 * expr.c (expand_expr_real_2): When handling conversions involving
7641 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7642 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7643 as a poly_uint64 too.
7645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7646 Alan Hayward <alan.hayward@arm.com>
7647 David Sherwood <david.sherwood@arm.com>
7649 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7652 Alan Hayward <alan.hayward@arm.com>
7653 David Sherwood <david.sherwood@arm.com>
7655 * combine.c (can_change_dest_mode): Handle polynomial
7656 REGMODE_NATURAL_SIZE.
7657 * expmed.c (store_bit_field_1): Likewise.
7658 * expr.c (store_constructor): Likewise.
7659 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7660 and polynomial REGMODE_NATURAL_SIZE.
7661 (gen_lowpart_common): Likewise.
7662 * reginfo.c (record_subregs_of_mode): Likewise.
7663 * rtlanal.c (read_modify_subreg_p): Likewise.
7665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7666 Alan Hayward <alan.hayward@arm.com>
7667 David Sherwood <david.sherwood@arm.com>
7669 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7670 numbers of elements.
7672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7673 Alan Hayward <alan.hayward@arm.com>
7674 David Sherwood <david.sherwood@arm.com>
7676 * match.pd: Cope with polynomial numbers of vector elements.
7678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7679 Alan Hayward <alan.hayward@arm.com>
7680 David Sherwood <david.sherwood@arm.com>
7682 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7683 in a POINTER_PLUS_EXPR.
7685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7686 Alan Hayward <alan.hayward@arm.com>
7687 David Sherwood <david.sherwood@arm.com>
7689 * omp-simd-clone.c (simd_clone_subparts): New function.
7690 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7691 (ipa_simd_modify_function_body): Likewise.
7693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7694 Alan Hayward <alan.hayward@arm.com>
7695 David Sherwood <david.sherwood@arm.com>
7697 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7698 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7699 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7700 (expand_vector_condition, vector_element): Likewise.
7701 (subparts_gt): New function.
7702 (get_compute_type): Use subparts_gt.
7703 (count_type_subparts): Delete.
7704 (expand_vector_operations_1): Use subparts_gt instead of
7705 count_type_subparts.
7707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7708 Alan Hayward <alan.hayward@arm.com>
7709 David Sherwood <david.sherwood@arm.com>
7711 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7712 (vect_compile_time_alias): ...this new function. Do the calculation
7713 on poly_ints rather than trees.
7714 (vect_prune_runtime_alias_test_list): Update call accordingly.
7716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7717 Alan Hayward <alan.hayward@arm.com>
7718 David Sherwood <david.sherwood@arm.com>
7720 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7722 (vect_schedule_slp_instance): Likewise.
7724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7725 Alan Hayward <alan.hayward@arm.com>
7726 David Sherwood <david.sherwood@arm.com>
7728 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7729 constant and extern definitions for variable-length vectors.
7730 (vect_get_constant_vectors): Note that the number of units
7731 is known to be constant.
7733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7734 Alan Hayward <alan.hayward@arm.com>
7735 David Sherwood <david.sherwood@arm.com>
7737 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7738 of units as polynomial. Choose between WIDE and NARROW based
7741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7742 Alan Hayward <alan.hayward@arm.com>
7743 David Sherwood <david.sherwood@arm.com>
7745 * tree-vect-stmts.c (simd_clone_subparts): New function.
7746 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7749 Alan Hayward <alan.hayward@arm.com>
7750 David Sherwood <david.sherwood@arm.com>
7752 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7753 vectors as polynomial. Use build_index_vector for
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 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7761 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7762 for variable-length vectors.
7763 (vectorizable_mask_load_store): Treat the number of units as
7764 polynomial, asserting that it is constant if the condition has
7765 already been enforced.
7766 (vectorizable_store, vectorizable_load): Likewise.
7768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7769 Alan Hayward <alan.hayward@arm.com>
7770 David Sherwood <david.sherwood@arm.com>
7772 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7773 of units as polynomial. Punt if we can't tell at compile time
7774 which vector contains the final result.
7776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7777 Alan Hayward <alan.hayward@arm.com>
7778 David Sherwood <david.sherwood@arm.com>
7780 * tree-vect-loop.c (vectorizable_induction): Treat the number
7781 of units as polynomial. Punt on SLP inductions. Use an integer
7782 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7783 cast of such a series for variable-length floating-point
7786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7787 Alan Hayward <alan.hayward@arm.com>
7788 David Sherwood <david.sherwood@arm.com>
7790 * tree.h (build_index_vector): Declare.
7791 * tree.c (build_index_vector): New function.
7792 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7793 of units as polynomial, forcibly converting it to a constant if
7794 vectorizable_reduction has already enforced the condition.
7795 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7796 to create a {1,2,3,...} vector.
7797 (vectorizable_reduction): Treat the number of units as polynomial.
7798 Choose vectype_in based on the largest scalar element size rather
7799 than the smallest number of units. Enforce the restrictions
7802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7803 Alan Hayward <alan.hayward@arm.com>
7804 David Sherwood <david.sherwood@arm.com>
7806 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7807 number of units as polynomial.
7809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7810 Alan Hayward <alan.hayward@arm.com>
7811 David Sherwood <david.sherwood@arm.com>
7813 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7814 * target.def (autovectorize_vector_sizes): Return the vector sizes
7815 by pointer, using vector_sizes rather than a bitmask.
7816 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7817 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7818 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7820 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7821 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7822 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7823 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7824 * omp-general.c (omp_max_vf): Likewise.
7825 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7826 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7827 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7828 * tree-vect-slp.c (vect_slp_bb): Likewise.
7829 * doc/tm.texi: Regenerate.
7830 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7832 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7833 the vector size as a poly_uint64 rather than an unsigned int.
7834 (current_vector_size): Change from an unsigned int to a poly_uint64.
7835 (get_vectype_for_scalar_type): Update accordingly.
7836 * tree.h (build_truth_vector_type): Take the size and number of
7837 units as a poly_uint64 rather than an unsigned int.
7838 (build_vector_type): Add a temporary overload that takes
7839 the number of units as a poly_uint64 rather than an unsigned int.
7840 * tree.c (make_vector_type): Likewise.
7841 (build_truth_vector_type): Take the number of units as a poly_uint64
7842 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
7849 as poly_uint64s rather than unsigned ints.
7850 * targhooks.h (default_get_mask_mode): Update accordingly.
7851 * targhooks.c (default_get_mask_mode): Likewise.
7852 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7853 * doc/tm.texi: Regenerate.
7855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7856 Alan Hayward <alan.hayward@arm.com>
7857 David Sherwood <david.sherwood@arm.com>
7859 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7860 * omp-general.c (omp_max_vf): Likewise.
7861 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7862 (expand_omp_simd): Handle polynomial safelen.
7863 * omp-low.c (omplow_simd_context): Add a default constructor.
7864 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7865 (lower_rec_simd_input_clauses): Update accordingly.
7866 (lower_rec_input_clauses): Likewise.
7868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7869 Alan Hayward <alan.hayward@arm.com>
7870 David Sherwood <david.sherwood@arm.com>
7872 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7873 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7874 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7875 (vect_analyze_slp_cost): Likewise.
7876 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7877 (vect_model_load_cost): Likewise.
7879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7880 Alan Hayward <alan.hayward@arm.com>
7881 David Sherwood <david.sherwood@arm.com>
7883 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7884 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7885 from an unsigned int * to a poly_uint64_pod *.
7886 (calculate_unrolling_factor): New function.
7887 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7890 Alan Hayward <alan.hayward@arm.com>
7891 David Sherwood <david.sherwood@arm.com>
7893 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7894 from an unsigned int to a poly_uint64.
7895 (_loop_vec_info::slp_unrolling_factor): Likewise.
7896 (_loop_vec_info::vectorization_factor): Change from an int
7898 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7899 (vect_get_num_vectors): New function.
7900 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7901 (vect_get_num_copies): Use vect_get_num_vectors.
7902 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7903 to an unsigned int *.
7904 (vect_analyze_data_refs): Change min_vf from an int * to a
7906 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7907 than an unsigned HOST_WIDE_INT.
7908 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7909 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7910 to an unsigned int *.
7911 (vect_analyze_data_ref_dependences): Likewise.
7912 (vect_compute_data_ref_alignment): Handle polynomial vf.
7913 (vect_enhance_data_refs_alignment): Likewise.
7914 (vect_prune_runtime_alias_test_list): Likewise.
7915 (vect_shift_permute_load_chain): Likewise.
7916 (vect_supportable_dr_alignment): Likewise.
7917 (dependence_distance_ge_vf): Take the vectorization factor as a
7918 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7919 (vect_analyze_data_refs): Change min_vf from an int * to a
7921 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7922 vfm1 as a poly_uint64 rather than an int. Make the same change
7923 for the returned bound_scalar.
7924 (vect_gen_vector_loop_niters): Handle polynomial vf.
7925 (vect_do_peeling): Likewise. Update call to
7926 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7927 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7929 * tree-vect-loop.c (vect_determine_vectorization_factor)
7930 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7931 (vect_get_known_peeling_cost): Likewise.
7932 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7933 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7934 (vect_transform_loop): Likewise. Use the lowest possible VF when
7935 updating the upper bounds of the loop.
7936 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7938 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7939 polynomial unroll factors.
7940 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7941 (vect_make_slp_decision): Likewise.
7942 (vect_supported_load_permutation_p): Likewise, and polynomial
7944 (vect_analyze_slp_cost): Handle polynomial vf.
7945 (vect_slp_analyze_node_operations): Likewise.
7946 (vect_slp_analyze_bb_1): Likewise.
7947 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7948 than an unsigned HOST_WIDE_INT.
7949 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7950 (vectorizable_load): Handle polynomial vf.
7951 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7953 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7956 Alan Hayward <alan.hayward@arm.com>
7957 David Sherwood <david.sherwood@arm.com>
7959 * match.pd: Handle bit operations involving three constants
7960 and try to fold one pair.
7962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7964 * tree-vect-loop-manip.c: Include gimple-fold.h.
7965 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7966 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7967 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7968 Add a path that uses a step of VF instead of 1, but disable it
7970 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7971 and niters_no_overflow parameters. Update calls to
7972 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7973 Create a new SSA name if the latter choses to use a ste other
7974 than zero, and return it via niters_vector_mult_vf_var.
7975 * tree-vect-loop.c (vect_transform_loop): Update calls to
7976 vect_do_peeling, vect_gen_vector_loop_niters and
7977 slpeel_make_loop_iterate_ntimes.
7978 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7979 (vect_gen_vector_loop_niters): Update declarations after above changes.
7981 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7983 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7984 128-bit round to integer instructions.
7985 (ceil<mode>2): Likewise.
7986 (btrunc<mode>2): Likewise.
7987 (round<mode>2): Likewise.
7989 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7991 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7992 unaligned VSX load/store on P8/P9.
7993 (expand_block_clear): Allow the use of unaligned VSX
7994 load/store on P8/P9.
7996 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7998 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8000 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8001 swap associated with both a load and a store.
8003 2018-01-02 Andrew Waterman <andrew@sifive.com>
8005 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8006 * config/riscv/riscv.md (clear_cache): Use it.
8008 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8010 * web.c: Remove out-of-date comment.
8012 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8014 * expr.c (fixup_args_size_notes): Check that any existing
8015 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8016 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8017 (emit_single_push_insn): ...here.
8019 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8021 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8022 (const_vector_encoded_nelts): New function.
8023 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8024 (const_vector_int_elt, const_vector_elt): Declare.
8025 * emit-rtl.c (const_vector_int_elt_1): New function.
8026 (const_vector_elt): Likewise.
8027 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8028 of CONST_VECTOR_ELT.
8030 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8032 * expr.c: Include rtx-vector-builder.h.
8033 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8034 directly on the tree encoding.
8035 (const_vector_from_tree): Likewise.
8036 * optabs.c: Include rtx-vector-builder.h.
8037 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8038 sequence of "u" values.
8039 * vec-perm-indices.c: Include rtx-vector-builder.h.
8040 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8041 directly on the vec_perm_indices encoding.
8043 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8045 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8046 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8047 * rtx-vector-builder.h: New file.
8048 * rtx-vector-builder.c: Likewise.
8049 * rtl.h (rtx_def::u2): Add a const_vector field.
8050 (CONST_VECTOR_NPATTERNS): New macro.
8051 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8052 (CONST_VECTOR_DUPLICATE_P): Likewise.
8053 (CONST_VECTOR_STEPPED_P): Likewise.
8054 (CONST_VECTOR_ENCODED_ELT): Likewise.
8055 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8056 (unwrap_const_vec_duplicate): Likewise.
8057 (const_vec_series_p): Check for a non-duplicated vector encoding.
8058 Say that the function only returns true for integer vectors.
8059 * emit-rtl.c: Include rtx-vector-builder.h.
8060 (gen_const_vec_duplicate_1): Delete.
8061 (gen_const_vector): Call gen_const_vec_duplicate instead of
8062 gen_const_vec_duplicate_1.
8063 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8064 (gen_const_vec_duplicate): Use rtx_vector_builder.
8065 (gen_const_vec_series): Likewise.
8066 (gen_rtx_CONST_VECTOR): Likewise.
8067 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8068 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8069 Build a new vector rather than modifying a CONST_VECTOR in-place.
8070 (handle_special_swappables): Update call accordingly.
8071 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8072 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8073 Build a new vector rather than modifying a CONST_VECTOR in-place.
8074 (handle_special_swappables): Update call accordingly.
8076 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8078 * simplify-rtx.c (simplify_const_binary_operation): Use
8079 CONST_VECTOR_ELT instead of XVECEXP.
8081 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8083 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8084 the selector elements to be different from the data elements
8085 if the selector is a VECTOR_CST.
8086 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8087 ssizetype for the selector.
8089 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8091 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8092 before testing each element individually.
8093 * tree-vect-generic.c (lower_vec_perm): Likewise.
8095 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8097 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8098 * selftest-run-tests.c (selftest::run_tests): Call it.
8099 * vector-builder.h (vector_builder::operator ==): New function.
8100 (vector_builder::operator !=): Likewise.
8101 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8102 (vec_perm_indices::all_from_input_p): New function.
8103 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8104 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8105 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8106 instead of reading the VECTOR_CST directly. Detect whether both
8107 vector inputs are the same before constructing the vec_perm_indices,
8108 and update the number of inputs argument accordingly. Use the
8109 utility functions added above. Only construct sel2 if we need to.
8111 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8113 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8114 the broadcast of the low byte.
8115 (expand_mult_highpart): Use an explicit encoding for the permutes.
8116 * optabs-query.c (can_mult_highpart_p): Likewise.
8117 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8118 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8119 (vectorizable_bswap): Likewise.
8120 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8121 explicit encoding for the power-of-2 permutes.
8122 (vect_permute_store_chain): Likewise.
8123 (vect_grouped_load_supported): Likewise.
8124 (vect_permute_load_chain): Likewise.
8126 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8128 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8129 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8130 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8131 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8132 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8133 (vect_gen_perm_mask_any): Likewise.
8135 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8137 * int-vector-builder.h: New file.
8138 * vec-perm-indices.h: Include int-vector-builder.h.
8139 (vec_perm_indices): Redefine as an int_vector_builder.
8140 (auto_vec_perm_indices): Delete.
8141 (vec_perm_builder): Redefine as a stand-alone class.
8142 (vec_perm_indices::vec_perm_indices): New function.
8143 (vec_perm_indices::clamp): Likewise.
8144 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8145 (vec_perm_indices::new_vector): New function.
8146 (vec_perm_indices::new_expanded_vector): Update for new
8147 vec_perm_indices class.
8148 (vec_perm_indices::rotate_inputs): New function.
8149 (vec_perm_indices::all_in_range_p): Operate directly on the
8150 encoded form, without computing elided elements.
8151 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8152 encoding. Update for new vec_perm_indices class.
8153 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8154 the given vec_perm_builder.
8155 (expand_vec_perm_var): Update vec_perm_builder constructor.
8156 (expand_mult_highpart): Use vec_perm_builder instead of
8157 auto_vec_perm_indices.
8158 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8159 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8160 or double series encoding as appropriate.
8161 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8162 vec_perm_indices instead of auto_vec_perm_indices.
8163 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8164 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8165 (vect_permute_store_chain): Likewise.
8166 (vect_grouped_load_supported): Likewise.
8167 (vect_permute_load_chain): Likewise.
8168 (vect_shift_permute_load_chain): Likewise.
8169 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8170 (vect_transform_slp_perm_load): Likewise.
8171 (vect_schedule_slp_instance): Likewise.
8172 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8173 (vectorizable_mask_load_store): Likewise.
8174 (vectorizable_bswap): Likewise.
8175 (vectorizable_store): Likewise.
8176 (vectorizable_load): Likewise.
8177 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8178 vec_perm_indices instead of auto_vec_perm_indices. Use
8179 tree_to_vec_perm_builder to read the vector from a tree.
8180 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8181 vec_perm_builder instead of a vec_perm_indices.
8182 (have_whole_vector_shift): Use vec_perm_builder and
8183 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8184 truncation to calc_vec_perm_mask_for_shift.
8185 (vect_create_epilog_for_reduction): Likewise.
8186 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8187 from auto_vec_perm_indices to vec_perm_indices.
8188 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8189 instead of changing individual elements.
8190 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8191 the vector in d.perm.
8192 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8193 from auto_vec_perm_indices to vec_perm_indices.
8194 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8195 instead of changing individual elements.
8196 (arm_vectorize_vec_perm_const): Use new_vector to install
8197 the vector in d.perm.
8198 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8199 Update vec_perm_builder constructor.
8200 (rs6000_expand_interleave): Likewise.
8201 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8202 (rs6000_expand_interleave): Likewise.
8204 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8206 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8207 to qimode could truncate the indices.
8208 * optabs.c (expand_vec_perm_var): Likewise.
8210 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8212 * Makefile.in (OBJS): Add vec-perm-indices.o.
8213 * vec-perm-indices.h: New file.
8214 * vec-perm-indices.c: Likewise.
8215 * target.h (vec_perm_indices): Replace with a forward class
8217 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8218 * optabs.h: Include vec-perm-indices.h.
8219 (expand_vec_perm): Delete.
8220 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8221 (expand_vec_perm_const): Declare.
8222 * target.def (vec_perm_const_ok): Replace with...
8223 (vec_perm_const): ...this new hook.
8224 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8225 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8226 * doc/tm.texi: Regenerate.
8227 * optabs.def (vec_perm_const): Delete.
8228 * doc/md.texi (vec_perm_const): Likewise.
8229 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8230 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8231 expand_vec_perm for constant permutation vectors. Assert that
8232 the mode of variable permutation vectors is the integer equivalent
8233 of the mode that is being permuted.
8234 * optabs-query.h (selector_fits_mode_p): Declare.
8235 * optabs-query.c: Include vec-perm-indices.h.
8236 (selector_fits_mode_p): New function.
8237 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8238 is defined, instead of checking whether the vec_perm_const_optab
8239 exists. Use targetm.vectorize.vec_perm_const instead of
8240 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8241 fit in the vector mode before using a variable permute.
8242 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8243 vec_perm_indices instead of an rtx.
8244 (expand_vec_perm): Replace with...
8245 (expand_vec_perm_const): ...this new function. Take the selector
8246 as a vec_perm_indices rather than an rtx. Also take the mode of
8247 the selector. Update call to shift_amt_for_vec_perm_mask.
8248 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8249 Use vec_perm_indices::new_expanded_vector to expand the original
8250 selector into bytes. Check whether the indices fit in the vector
8251 mode before using a variable permute.
8252 (expand_vec_perm_var): Make global.
8253 (expand_mult_highpart): Use expand_vec_perm_const.
8254 * fold-const.c: Includes vec-perm-indices.h.
8255 * tree-ssa-forwprop.c: Likewise.
8256 * tree-vect-data-refs.c: Likewise.
8257 * tree-vect-generic.c: Likewise.
8258 * tree-vect-loop.c: Likewise.
8259 * tree-vect-slp.c: Likewise.
8260 * tree-vect-stmts.c: Likewise.
8261 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8263 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8264 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8265 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8266 (aarch64_vectorize_vec_perm_const): ...this new function.
8267 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8268 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8269 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8270 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8271 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8272 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8273 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8275 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8276 check for NEON modes.
8277 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8278 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8279 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8280 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8282 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8283 the old VEC_PERM_CONST conditions.
8284 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8285 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8286 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8287 (ia64_vectorize_vec_perm_const_ok): Merge into...
8288 (ia64_vectorize_vec_perm_const): ...this new function.
8289 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8290 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8291 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8292 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8293 * config/mips/mips.c (mips_expand_vec_perm_const)
8294 (mips_vectorize_vec_perm_const_ok): Merge into...
8295 (mips_vectorize_vec_perm_const): ...this new function.
8296 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8297 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8298 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8299 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8300 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8301 (rs6000_expand_vec_perm_const): Delete.
8302 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8304 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8305 (altivec_expand_vec_perm_const_le): Take each operand individually.
8306 Operate on constant selectors rather than rtxes.
8307 (altivec_expand_vec_perm_const): Likewise. Update call to
8308 altivec_expand_vec_perm_const_le.
8309 (rs6000_expand_vec_perm_const): Delete.
8310 (rs6000_vectorize_vec_perm_const_ok): Delete.
8311 (rs6000_vectorize_vec_perm_const): New function.
8312 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8313 an element count and rtx array.
8314 (rs6000_expand_extract_even): Update call accordingly.
8315 (rs6000_expand_interleave): Likewise.
8316 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8317 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8318 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8319 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8320 (rs6000_expand_vec_perm_const): Delete.
8321 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8322 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8323 (altivec_expand_vec_perm_const_le): Take each operand individually.
8324 Operate on constant selectors rather than rtxes.
8325 (altivec_expand_vec_perm_const): Likewise. Update call to
8326 altivec_expand_vec_perm_const_le.
8327 (rs6000_expand_vec_perm_const): Delete.
8328 (rs6000_vectorize_vec_perm_const_ok): Delete.
8329 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8330 reference to the SPE evmerge intructions.
8331 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8332 an element count and rtx array.
8333 (rs6000_expand_extract_even): Update call accordingly.
8334 (rs6000_expand_interleave): Likewise.
8335 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8336 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8338 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8340 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8342 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8343 vector mode and that that mode matches the mode of the data
8345 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8346 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8347 directly using expand_vec_perm_1 when forcing selectors into
8349 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8353 * optabs-query.h (can_vec_perm_p): Delete.
8354 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8355 * optabs-query.c (can_vec_perm_p): Split into...
8356 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8357 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8358 particular selector is valid.
8359 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8360 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8361 (vect_grouped_load_supported): Likewise.
8362 (vect_shift_permute_load_chain): Likewise.
8363 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8364 (vect_transform_slp_perm_load): Likewise.
8365 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8366 (vectorizable_bswap): Likewise.
8367 (vect_gen_perm_mask_checked): Likewise.
8368 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8369 implementations of variable permutation vectors into account
8370 when deciding which selector to use.
8371 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8372 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8373 with a false third argument.
8374 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8375 to test whether the constant selector is valid and can_vec_perm_var_p
8376 to test whether a variable selector is valid.
8378 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8380 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8381 * optabs-query.c (can_vec_perm_p): Likewise.
8382 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8383 instead of vec_perm_indices.
8384 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8385 (vect_gen_perm_mask_checked): Likewise,
8386 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8387 (vect_gen_perm_mask_checked): Likewise,
8389 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8391 * optabs-query.h (qimode_for_vec_perm): Declare.
8392 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8393 (qimode_for_vec_perm): ...this new function.
8394 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8396 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8398 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8399 does not have a conditional at the top.
8401 2018-01-02 Richard Biener <rguenther@suse.de>
8403 * ipa-inline.c (big_speedup_p): Fix expression.
8405 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8408 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8411 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8415 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8416 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8417 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8418 cond_taken_branch_cost 3->4.
8420 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8422 PR tree-optimization/83581
8423 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8424 TODO_cleanup_cfg if any changes have been made.
8427 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8428 convert_modes if target mode has the right side, but different mode
8432 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8433 last argument when extracting from CONCAT. If either from_real or
8434 from_imag is NULL, use expansion through memory. If result is not
8435 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8436 the parts directly to inner mode, if even that fails, use expansion
8440 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8441 check for bswap in mode rather than HImode and use that in expand_unop
8444 Copyright (C) 2018 Free Software Foundation, Inc.
8446 Copying and distribution of this file, with or without modification,
8447 are permitted in any medium without royalty provided the copyright
8448 notice and this notice are preserved.